Nvidia’s GeForce GTX 670 graphics card

The new PC hardware releases have been coming at us so fast lately, we haven’t been able to keep up. Yes, today marks the introduction of yet another new graphics card, as Nvidia fills out its GeForce GTX 600 series. After posting my GeForce GTX 690 review just yesterday, I had half a mind to delay this review and give myself a breather. But then I realized, hey, that’s why God made caffeine.


Nvidia’s GeForce GTX 670 reference design. Source: Nvidia

Besides, we wouldn’t want to miss this one, because the GeForce GTX 670 is sure to be the most broadly attractive product yet in the 600 series. As you might have cleverly deduced, the GTX 670 is a cheaper, somewhat down-spec version of the mighty GeForce GTX 680, which is itself a pretty darned desirable video card.

GPU

base

clock

(MHz)

GPU

boost

clock

(MHz)

Shader

ALUs

Textures

filtered/

clock

ROP

pixels/

clock

Memory

transfer

rate

Memory

interface

width

(bits)

Peak

power

draw

GeForce GTX 670 915 980 1344 112 32 6 GT/s 256 170W
GeForce GTX 680 1006 1058 1536 128 32 6 GT/s 256 195W

There’s no magic to the GTX 670’s appeal. Nvidia has dialed back the specifications just a bit versus the GTX 680 and sliced 100 bucks off of the price. Math is fun; do it with me: the GTX 680 lists at $499, so the GTX 670 will sell for $399. That’s getting into territory where a lot more folks might feel inclined to justify the expense. The GTX 670 has a chance to be pretty popular.

Those chances are bolstered by the fact that Nvidia’s cuts to the GTX670’s specifications aren’t likely to hurt too much. They’ve nuked one of the GK104 chip’s eight SMX units and reduced clock speeds, but the cuts don’t go terribly deep. The GTX 670 still has all 32 ROP units intact, preserving its anti-aliasing power, and its memory interface is still 256-bits wide at 6Gbps, just like the GTX 680. (Frankly, memory bandwidth and ROP rate are two of the places where we’d expect the GK104 GPU to be relatively weak, in the grand scheme of things.) Also helpful: the GTX 670’s boost clock is 980MHz, not far from the GTX 680’s boost clock of 1058MHz. Although the gap between the two cards’ base clocks is a bit larger, the boost clock is probably the more important of the two; it’s closer to the GPU’s typical operating speed.

All in all, we suspect the GTX 670 may not trail the GTX 680 much at all in real-world performance. Looks to us like Nvidia is positioning itself pretty aggressively against the Radeon HD 7950.

Having said that, we should point out the GTX 670 is still made up of some relatively lightweight hardware. The card’s max power draw is rated at only 170W—versus 200W for the Radeon HD 7950—and, well, have a look at the thing.

Source: Nvidia


Source: Nvidia

The board is only about 6.75″ inches long, shorter than the world’s tolerance for Nickelback. The cooling shroud protrudes to 9.5″, like a pair of platform shoes, attempting to keep the GTX 670 from looking like something you’d find in a cut-rate Dell.

Nvidia does expect video card makers to do some nice things with its teensy-weensy reference PCB design, though, including shorter cards intended for small-form-factor enclosures and possibly some single-slot coolers. Board makers will also take things in the other direction, of course, offering higher-clocked variants with beefier cooling.


Source: Nvidia

Here’s a picture of a Zotac GTX 670 AMP! card that is purportedly making its way to us right now via an unfortunately slow shipping method. Zotac tells us this puppy will have base and boost clocks of 1098 and 1176MHz, respectively, with 6.6GT/s memory. The performance of this thing is almost certain to eclipse a stock-clocked GeForce GTX 680’s, given those numbers. Since Zotac’s list price for the GTX 670 AMP! is $449, we think the stock GTX 680 might want to start looking for new employment. The cooler looks like the same snazzy one on Zotac’s GTX 680 AMP!, from which we’ve seen good things. We’re eager to test this card once, you know, it makes it here.

Peak pixel

fill rate

(Gpixels/s)

Peak bilinear

filtering

(Gtexels/s)

Peak bilinear

FP16 filtering

(Gtexels/s)

Peak shader

arithmetic

(TFLOPS)

Peak

rasterization

rate

(Mtris/s)

Memory

bandwidth

(GB/s)

GeForce GTX 560
Ti
29 58 58 1.4 1800 134
GeForce GTX
570
29 44 44 1.4 2928 152
GeForce GTX 580 37 49 49 1.6 3088 192
GeForce GTX
670
29 102 102 2.5 3660 192
GeForce GTX 680 32 129 129 3.1 4024 192
Radeon HD 6970 28 85 43 2.7 1780 176
Radeon HD 7870 32 80 40 2.6 2000 154
Radeon HD
7950
26 90 45 2.9 1600 240
Radeon HD 7970 30 118 59 3.8 1850 264

Even if you crunch the numbers based on the card’s bone-stock base clock frequency, the GTX 670 looks to be considerably more powerful than Nvidia’s prior-gen top-o’-the-line offering, the GeForce GTX 580, in every category but memory bandwidth (a tie) and ROP rate (where the 580’s faster). Against the Radeon HD 7950, the comparison becomes complicated in a hurry because both cards have pronounced advantages in certain categories. Anyhow, it’s best not to compare these numbers across different architectures and pretend they’ll reliably predict real-world performance. We won’t linger, now that our specs table quotas have been fulfilled.

We should address a couple of tricky issues instead.

First, although the GTX 670 looks like it might be something folks would want to buy, there is a question lingering over its head: will there be a sufficient supply of cards in the market? Since its launch, the GTX 680 has been flashing in and out of stock at online retailers very quickly, as shipments of cards arrive and then sell out quickly. Will the GTX 670 suffer the same fate? Nvidia tells us no, that it’s confident the supply of GTX 670s will be better. The presence of custom boards from Nvidia’s various partners on day one is likely to help. We’ll have to watch the availability situation in the coming weeks, though. The supply and demand picture for video cards is notoriously difficult to predict.

Second, how does one pull off a video card review in a single work day? The answer, my friend, is cutting corners like a government contractor. That’s precisely what we’ve done in order to produce this quick first look at the GeForce GTX 670. We have egregiously stolen the performance results from our GTX 690 review, added scores for the GTX 670 and the Radeon HD 7950, and called it good. One consequence of this highly questionable decision is that all of our performance numbers come from a triple-display setup running at the ridiculous resolution of 5760×1200. Thus, you’re going to see the GTX 670 struggling to provide acceptable performance in some cases on the following pages. These results should still give you a sense of how this new card compares to its key competitors, along with some really expensive dual-GPU products from the last couple of generations. At least there won’t be much doubt whether our test scenarios are truly GPU-bound.

Our testing methods

As ever, we did our best to deliver clean benchmark numbers. Tests were run at least three times, and we’ve reported the median result.

Our test systems were configured like so:

Processor Core i7-3820
Motherboard Gigabyte
X79-UD3
Chipset Intel X79
Express
Memory size 16GB (4 DIMMs)
Memory type Corsair
Vengeance CMZ16GX3M4X1600C9
DDR3 SDRAM at 1600MHz
Memory timings 9-9-11-24
1T
Chipset drivers INF update
9.3.0.1019

Rapid Storage Technology Enterprise 3.0.0.3020

Audio Integrated
X79/ALC898

with Realtek 6.0.1.6526 drivers

Hard drive Corsair
F240 240GB SATA
Power supply Corsair
AX850
OS Windows 7 Ultimate x64 Edition

Service Pack 1

DirectX 11 June 2010 Update

Driver
revision
GPU
base

core clock 

(MHz)

Memory

clock

(MHz)

Memory

size

(MB)

GeForce
GTX 570
ForceWare
301.34
732 950 1280
GeForce
GTX 590
ForceWare
301.24
608 854 3072
GeForce
GTX 670
ForceWare
301.34
915 980 2048
GeForce
GTX 680
ForceWare
301.33
1006 1502 2048
GeForce
GTX 680 + Zotac GTX 680
ForceWare
301.33
1006 1502 2048
GeForce GTX
690
ForceWare
301.33
915 1502 4096
Radeon
HD 6990 AUSUM
Catalyst
12.4 + 12.3 CAP 1
880 1250 4096
Radeon
HD 7950
Catalyst
12.4 + 12.3 CAP 1
800 1250 3072
Radeon
HD 7970
Catalyst
12.4 + 12.3 CAP 1
925 1375 3072
Radeon
HD 7970 + XFX HD 7970
Catalyst
12.4 + 12.3 CAP 1
925 1375 3072

Thanks to Intel, Corsair, and Gigabyte for helping to outfit our test rigs with some of the finest hardware available. AMD, Nvidia, and the makers of the various products supplied the graphics cards for testing, as well.

Unless otherwise specified, image quality settings for the graphics cards were left at the control panel defaults. Vertical refresh sync (vsync) was disabled for all tests.

We used the following test applications:

Some further notes on our methods:

  • We used the Fraps utility to record frame rates while playing a 90-second sequence from the game. Although capturing frame rates while playing isn’t precisely repeatable, we tried to make each run as similar as possible to all of the others. We tested each Fraps sequence five times per video card in order to counteract any variability. We’ve included frame-by-frame results from Fraps for each game, and in those plots, you’re seeing the results from a single, representative pass through the test sequence.
  • We measured total system power consumption at the wall socket using a Yokogawa WT210 digital power meter. The monitor was plugged into a separate outlet, so its power draw was not part of our measurement. The cards were plugged into a motherboard on an open test bench.

    The idle measurements were taken at the Windows desktop with the Aero theme enabled. The cards were tested under load running Skyrim at its Ultra quality settings with FXAA enabled.

  • We measured noise levels on our test system, sitting on an open test bench, using an Extech 407738 digital sound level meter. The meter was mounted on a tripod approximately 10″ from the test system at a height even with the top of the video card.

    You can think of these noise level measurements much like our system power consumption tests, because the entire systems’ noise levels were measured. Of course, noise levels will vary greatly in the real world along with the acoustic properties of the PC enclosure used, whether the enclosure provides adequate cooling to avoid a card’s highest fan speeds, placement of the enclosure in the room, and a whole range of other variables. These results should give a reasonably good picture of comparative fan noise, though.

  • We used GPU-Z to log GPU temperatures during our load testing.

The tests and methods we employ are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.

The Elder Scrolls V: Skyrim

Our test run for Skyrim was a lap around the town of Whiterun, starting up high at the castle entrance, descending down the stairs into the main part of town, and then doing a figure-eight around the main drag.

We set the game to its “Ultra” presets with 4X multisampled antialiasing. We then layered on FXAA post-process anti-aliasing, as well, for the best possible image quality without editing an .ini file. We also had the high-res texture pack installed, of course. Although it’s not pictured above, the total display resolution was 5760×1200.

Frame time
in milliseconds
FPS
rate
8.3 120
16.7 60
20 50
25 40
33.3 30
50 20

If the plots above aren’t familiar to you, then you’re probably not acquainted with our new methods of testing video cards. Rather than explain everything yet again in this review, let me point you here and here for some explanations.

Even those not acquainted, though, can see that the GTX 670 appears to handle our triple-display Skyrim workload quite nicely. Frame latencies are mostly below the 25 millisecond mark; that means performance remains steadily above 40 FPS.

These results are nice and neat. Regardless of whether focus on traditional FPS averages or on the more important question of frame latencies, the GTX 670 performs well, just trailing the GTX 680, as expected. The GeForce cards take the top four spots in this test, too. The Radeons don’t perform too poorly, with the exception of the 6990, but the Nvidia GPUs consistently have an edge.

Batman: Arkham City

We did a little Batman-style free running through the rooftops of Gotham for this one.

Those latency spikes are something we’re used to seeing from Arkham City—doesn’t seem to matter much whether you’re dealing with one GPU or two.

Every card produces some long-latency frames in this test scenario, but the size and frequency of the latency spikes on the GeForce cards is smaller than on the Radeons. As a result, even though the Radeon HD 7970 produces more frames per second than the GTX 670, the new GeForce actually offers smoother gameplay—all of our latency-oriented metrics agree on that point. The trouble for the Radeons, both the 7950 and the 7970, is that 5-10% of the frames they produce take a relatively long time, as the latency-by-percentile plot illustrates. The performance “tail” on the GeForce cards is much smaller and tamer.

Battlefield 3

We tested Battlefield 3 with all of its DX11 goodness cranked up, including the “Ultra” quality settings with both 4X MSAA and the high-quality version of the post-process FXAA. Our test was conducted in the “Operation Guillotine” level, for 60 seconds starting at the third checkpoint.

The tables turn a bit in BF3. Although the GTX 670 performs pretty well for a single card driving a six-megapixel monitor array at this game’s “Ultra” settings, the Radeon HD 7950 is simply a little bit faster. Neither card is all that fast, but neither one has any issues with big latency spikes, either. Subjectively, they’re both remarkably playable given their FPS averages in the mid-twenties. I’d probably dial back the quality level on both of ’em in order to ensure really fluid gameplay, though.

Crysis 2

Our cavalcade of punishing but pretty DirectX 11 games continues with Crysis 2, which we patched with both the DX11 and high-res texture updates.

Notice that we left object image quality at “extreme” rather than “ultra,” in order to avoid the insane over-tessellation of flat surfaces that somehow found its way into the DX11 patch. We tested 60 seconds of gameplay in the level pictured above, where we gunned down several bad guys, making our way across a skywalk to another rooftop.

The GTX 670 outguns the Radeon HD 7950 in every respect here, especially in terms of consistent delivery of low-latency frames. The Radeons’ weakness on that front is due to a problem we explained here; it’s related to the high-latency frames you can see on every card right at the beginning of the test run.

Power consumption

These power results are a bit odd since we’re driving three monitors throughout our tests. That causes the Radeons to draw quite a bit more power at idle than we’d usually expect, and AMD’s ZeroCore power feature doesn’t seem to work, either, robbing the Radeon 7000-series cards of their usual advantage when the display is in power-save mode.

The GTX 670 doesn’t appear to be affected by similar problems, and its power draw at idle is nice and low. Remember, that’s 68W at idle for a whole system built on Intel’s beefy X79 platform. When running a game—Skyrim, in this case—our test system draws a little more power with a GTX 670 installed than it does with a 7950 installed. That’s a surprise given the two cards’ specs. We’re curious to try this again with a different game or two in the future.

Noise levels and GPU temperatures

You know how I said the GeForce GTX 690’s smooth fan action feels like money? Yeah, well the GTX 670’s cooler sounds cheap. That’s why the 670 is relatively noisy at idle, even though it has very little power—and thus heat—to dissipate. The GTX 670’s blower emits a rough, friction-filled whine. It’s not especially loud, and heck, the thing manages not to register too high on decibel meter under load. Tucked inside of a case that insulates sound well, you might barely notice it. Still, this thing isn’t Nvidia’s best work. I’d be looking toward a card with an aftermarket cooler, most likely.

Conclusions

Since we completely mailed in this review and tested with punishing image quality settings at a six-megapixel resolution, I’m hesitant to draw too many strong conclusions. Still, I think the GeForce GTX 670’s story is pretty simple. The 670 trails its big brother, the GeForce GTX 680, by only an eyelash in most of our performance results, closely enough that you’d have a very hard time telling the difference between them in everyday use. Not only that, but it delivers lower 99th percentile frame times than both the Radeon HD 7950 and the 7970 in three of our four game tests. Meanwhile, the 670’s power consumption is quite tame.

If I were buying a video card right now for myself, I’d order up a GeForce GTX 670. You really can drive three two-megapixel displays with it and play games at reasonable frame rates, even though you’d want to go easier on the image quality options than we did. For those folks with a four-megapixel display or less, the 670 should deliver a nice dose of gaming nirvana. Now that the 670 is here, I fail to see the point of spending more on a GeForce GTX 680 or a Radeon HD 7970, unless it has higher-than-stock clocks or something.

We haven’t had time to overclock the 670, or to test an aftermarket card with a hopefully better cooler, or to put it up against some prior-gen options like the GeForce GTX 570, or to test at 2-4 megapixels, or much else… yet. All of those things will come, in due time. With those extra factors in play, it’s possible a higher-clocked version of the Radeon HD 7950, like XFX’s Black Edition, could swing the competitive balance in AMD’s favor. The GTX 670 and the 7950 aren’t terribly far apart from one another, in the grand scheme. But right now, based on everything we’ve seen, Nvidia appears to have the upper hand in raw price-performance terms. Funny how it comes down to that at the end of the day, so often, with these incredibly complex products.

I’m not nearly as wordy on Twitter.

Comments closed
    • CBHvi7t
    • 7 years ago

    How does this compare to older cards like the GF 450 GTS or HD 3870?
    What if my display is just 1/5 as big?
    I got an AMD HD 3870 that is still ok for many of my games (e.g. Crysis 1, HAWX 2) but I want something with the new features.

    • moose17145
    • 7 years ago

    The only thing I would like to see added to the new videocard reviews with the three monitors, is a single and dualy monitor setup for the power testing. I have a dual *monitor* setup on my pc, and would like to know if the AMD cards can enter their ZeroCore power feature with only two screens attached. likewise I think seeing the power consumption with only a single screen attached should be fairly relevant as well since I am sure a large enough number of people do only have a single screen attached to their system. I’m no expert since I do not write and do performance reviews for pc hardware, but I wouldn’t think disconnecting one or two monitors and taking two more readings should add too much more time to finish the review.

    Other than wanting to see some power consumption numbers on the radeons when they are able to enter into their ZeroCore state, I thought it was an excellent review! Thank you for all the hard work and keep it up!

    *I have a TV (1366×768 res) hooked up to my computer as my second “monitor”. Mainly I just use it to play netflix or watch tv series I have saved while I work on school work or doing whatever.

    Edited for spelling / grammar

    • thebeastie
    • 7 years ago

    Would appreciate at seeing bf3 benchmarks at lower resolution, I play games in nvnidia 3dtv mode on my Sony HMZ-T1 and games struggle when using 3d in low resolutions..

    • beck2448
    • 8 years ago

    As a result, even though the Radeon HD 7970 produces more frames per second than the GTX 670, the new GeForce actually offers smoother gameplay—all of our latency-oriented metrics agree on that point.

    This is the key observation for me.

      • ish718
      • 7 years ago

      The GTX 670 also consumes around 50w less power than the HD 7970 at full load. AMD really tweaked their drivers for BF3, the only game where they have a distinct advantage…

    • sweatshopking
    • 8 years ago

    stop arguing. THE ANSWER IS SIMPLE. IT’S THE BEST CARD [i<] [b<] EVER [/i<] [/b<]

      • Airmantharp
      • 8 years ago

      I agree-

      popped out my HD6950’s, popped in the Evga GTX670 SC, and first impressions are lasting. Fast and quiet. We’ll see about BF3 MP tonight.

    • APWNH
    • 8 years ago

    The BF3 results are very inconsistent with those of Anandtech’s reviews and Tom’s hardware reviews. All the other sites have the 670 soundly beating the 7950 and even the 7970 but that is clearly not the case here.

    I wonder what’s going on.

      • travbrad
      • 8 years ago

      It’s because those sites tested at very different resolutions. Most of them tested at 1080p or 2560×1600, not 5760×1200.

      The 7900 series’ 3GB of memory and higher memory bandwidth help a lot at such an extreme resolution.

      • Damage
      • 8 years ago

      See here:

      [url<]https://techreport.com/discussions.x/22922?post=636209[/url<]

    • dpaus
    • 8 years ago

    [quote<]The board is only about 6.75" inches long... (but) The cooling shroud protrudes to 9.5", like a pair of platform shoes[/quote<] More like the world's first e-peen extender. It's unfortunately going to be a few weeks too late, but this would have been a fine addition to my HTPC/gaming system..... Except that the stupid extender makes it too long for my CoolerMaster case.

      • Meadows
      • 8 years ago

      Not all versions of the card will feature that extra length.

        • dpaus
        • 8 years ago

        Thankfully; but I don’t want to wait any, um, longer, so I’m going with [url=Google gets license to test drive autonomous cards on Nevada roads]the PowerColor 7850[/url] (the fastest card I could find that would fit)

    • spigzone
    • 8 years ago

    “as Nvidia fills out its GeForce GTX 600 series.”

    Problematic wording with several cards left to go to ‘filll out’ the 600 series.

      • Silus
      • 8 years ago

      Not problematic at all, you just have a reading comprehension disorder. “As NVIDIA fills out” indicates an on-going process…as in, it’s not finished.

      Also, NVIDIA posted yet another quarterly profit:

      [url<]http://www.reuters.com/article/2012/05/11/nvidia-idUSL1E8GBHN520120511[/url<]

    • rrr
    • 8 years ago

    Awesome card, undercutting 7970 and even nVidia’s own 680.

      • Arclight
      • 8 years ago

      In my country it’s pretty close in price to the GTX 680. So much so that if you have enough money to spend on a 670 there is no reason not to buy the GTX 680 or 7970, depending on brand preferance.

      To be a clear winner it should be priced lower, imo.

        • rrr
        • 8 years ago

        Either availability is extremely shitty in your country or sellers are highway robbers. In mine it is ~20% cheaper than 680, and slightly cheaper than 7970 – and it’s not a major market by any stretch.

          • Arclight
          • 8 years ago

          Leanning towards highway robber. Plus we got to deal with $=Euro sh*t and the fact that we got 24% VAT over here. Also we don’t get [s<]VAT[/s<] tax refunds back at the end of the fiscal year like in the U.S.

            • Prion
            • 8 years ago

            > Also we don’t get VAT back at the end of the fiscal year like in the U.S.

            Huh? I’ve never lived anywhere that refunded VAT/sales tax on personal purchases.

            • Arclight
            • 8 years ago

            Yeah i looked it up and i was wrong. I was thinking of “tax refund ”

            [quote<]A tax refund or tax rebate is a refund on taxes when the tax liability is less than the taxes paid. Taxpayers can often get a tax refund on their income tax if the tax they owe is less than the sum of the total amount of the withholding taxes and estimated taxes that they paid, plus the refundable tax credits that they claim. (Tax refunds are money given back at the end of the financial year.)[/quote<] Edit: there is no such thing in my country......and i heard many on the forums, residents in the US (i presume), who were claiming to buy new PC parts when they get their "tax refund" (or incorrectly using the term "tax return").

        • entropy13
        • 8 years ago

        The GTX 670 over here is around $430, while the GTX 680 is around $535. The HD 7970 is around $540.

      • ptsant
      • 8 years ago

      Well, performance is certainly excellent, but the real price is not exactly in line with suggested price. In switzerland, the 670 can be had for 460-470 francs while the 7970 starts at 460 and goes up to 520-530 francs with immediate availability (for the record, the 690 is also immediately available for a mere 1111 francs). Obviously, higher prices correspond to overclocked custom models. When it is in direct competition with the 7970, I’m not sure that the 670 is as desireable as it sounds in theory.

    • jackbomb
    • 8 years ago

    [url<]http://img521.imageshack.us/img521/8845/gtx670.jpg[/url<]

    • Musafir_86
    • 8 years ago

    -Hmm, Scott, no compute (GPGPU) test result at all?

    Regards.

      • Airmantharp
      • 8 years ago

      If you look at the chart comparing compute capabilities, you quickly realize how silly it would be for him to waste his time.

      Not trying to be rude, but on a time limited review of a card known to have very limited GPGPU capabilities, I think we can agree to wait for the more extensive write-up that fills the gaps.

    • Laykun
    • 8 years ago

    Absolutely no mention of how much video memory the card has in the entire article? I must have missed something. I know that the memory bus width being the same can imply the same amount of memory but that’s not always a given.

      • Airmantharp
      • 8 years ago

      Just a quick peruse, but I think you might be right!

      [i<][b<]We[/i<][/b<] all know that it has 2GB of VRAM, but I can understand why that might be an issue for some.

      • benjaminsun
      • 8 years ago

      not to nitpick but there’s a chart on page 2 listing memory of all cards tested including 670

        • flip-mode
        • 8 years ago

        This.

    • marraco
    • 8 years ago

    99th percentile frame times makes TechReport the best review all over Internet.

    Keep doing a good job, and I’m waiting for SLI benchmarcks.

    I heard rummors that the 670 can be modded to 680 with a 680 BIOS. Any info?

      • Duck
      • 8 years ago

      I’d be surprised if the disabled bits were not fused off. It’s all moot anyway because yields are so bad there’s no way a fully working GPU will go into a 670GTX when they fail so badly at supplying the 680GTX.

    • DancinJack
    • 8 years ago

    The cover picture of this article is really good.

      • Duck
      • 8 years ago

      Damage is an artist.

    • flip-mode
    • 8 years ago

    If it’s not in stock in the U.S. it’s not real. [spoiler<]No offense to other countries, many of which I'd like to visit or spend some extended time in, but honestly, if it's not in stock in the number 1 consumer market then there is a legitimate vaporware argument to be made. Maybe this is Nvidia's way of being able to claim "availability"[/spoiler<] FWIW there seem to be plenty available at Newegg at the moment. But GTX 680s are still unobtainium - not that it matters now that the GTX 670 essentially obsoltetes it.

      • Airmantharp
      • 8 years ago

      But it IS in stock in the U.S. I just bought one!

        • flip-mode
        • 8 years ago

        Yep, they are in stock at the moment. But the bigger issue is that you have an addiction that needs to be dealt with. You really need to stop buying a new high end video card (or dualies) every 6 months.

          • Farting Bob
          • 8 years ago

          He’s not the only one. On bit-tech.net forums there is someone who went from 2x 580’s to 2 7970’s to 2x 680’s. I cannot fathom spending that much money just to get about 10 more fps (if that) every few months. Im sure they are tempted to get a 690 or 2 now as well.

          • Airmantharp
          • 8 years ago

          Well, if the card is $419 shipped, then I just need to get that for both of my cards. Shouldn’t be too hard, really, since they’re the much-sought-after OEM cards.

          Realistically, I won’t be losing any average performance, will be gaining playability, at basically the same cost, while setting myself up to drop a second one in whenever they drop in price.

          • Airmantharp
          • 8 years ago

          Also, my reasoning really extends to my original mission when building this system- that being that I wanted to run the latest games on high at 2560×1600, since I’d finally saved up for a 30″ panel after 5+ years of ogling at them.

          So yeah, it is sort of a disease, but I don’t really feel that it’s negative- once I’m happy with the performance, I will probably keep the system static for 2+ years, if not more.

        • DancinJack
        • 8 years ago

        What are you doing with your CF setup?

          • Airmantharp
          • 8 years ago

          On sale!

          Or at least it will be listed for sale when I get a chance (need to take pictures).

          If you want details/pricing, they’re the AMD built units with the BIOS switch (never unlocked, but unlockable) and the stock blowers, fully stock and never overclocked, and are in pristine condition having had only been run inside my fully-filtered Define R3. Just send me a PM if interested.

      • Firestarter
      • 8 years ago

      Who’s saying that the US is the #1 consumer market? Do you have the numbers on the Asian market? Do you know how many graphics cards are sold on a yearly basis in Europe?

      I guess I’m taking offense to your assertion of the importance of the USA.

        • Airmantharp
        • 8 years ago

        Sadly, it’s something we do here, whether it’s correct or incorrect in the situation in question. It’s essentially bred in.

        • cegras
        • 8 years ago

        The ‘asian’ market is HK and the other small island nations nearby. It does not compare to the US.

      • Jigar
      • 8 years ago

      I am sorry, you have just been topped by china, so keep quiet.

      [url<]http://online.wsj.com/article/SB10001424053111903461304576525852486131230.html[/url<]

    • Airmantharp
    • 8 years ago

    I think I found my card.

    Going to put the HD6950 2GB cards up for sale, combined they should be able to cover one of these. In my Fractal Design Define R3, it should be just as quiet, and significantly cooler. It’ll be slower, but smoother, performance wise, and I’ll be able to add a second card when the 700-series gets nearer. Sounds like a win!

    • Ushio01
    • 8 years ago

    Just ordered a Palit one for £318 in a £1200+ upgrade.

    Bye bye money.

    • clone
    • 8 years ago

    their has got to be a better way…… I just read the full Anandtech review of the GTX 670 and I came away thinking HD 7970 is the best card on the market winning more often than not against both GTX 680 & GTX 670 while only consuming a little more power while the GTX 670 is the better value card vs HD 7950.

    then I drop by TechReport and I’ve almost no idea what I just read???…. I like the added info but wow has it grown into an effort to discern data from TR’s testing results.

    sorry.

    p.s. tested resolutions at 5700 X 1200…. really?

    truly hardware has completely surpassed software on the graphics side.

      • Duck
      • 8 years ago

      I sort of agree. But TR at it’s worst is still better than pretty much all other sites at their best. So while TR might be getting some criticism over this, it’s only really because standards are usually so high that the only way left to go is down.

      • Ushio01
      • 8 years ago

      What Anandtech review did you read? They reviewed 10 games the 7970 won in 2 at all resolutions and at some resolutions with 2 others while the 680 cleen sweeped in the other 6 games.

        • clone
        • 8 years ago

        the one where HD 7970 won the first 4 games at 5760 X 1200 resolution, I don’t really care about “all resolutions” because the lower resolutions were all well above playable, anyway after the 4 victories the HD 7970 lost the next 6 games but then it dominated the synthetic tests.

        all the while selling for more than a GTX 670 but notably less than a GTX 680 which still can’t be found making it the best card.

        p.s. the differences in power consumption are vapor to me given the high end cards today consume around the same amount of power of a GTX 260 did just a few years ago.

          • Ushio01
          • 8 years ago

          There is no such anandtech review.

            • clone
            • 8 years ago

            actually their is but you can see whatever you want vs the results.

            • Firestarter
            • 8 years ago

            Can you post a link?

            • clone
            • 8 years ago

            sure can.

            [url<]http://www.anandtech.com/show/5818/nvidia-geforce-gtx-670-review-feat-evga[/url<] AMD HD 7970 wins: Crysis 2 5760 res min frames win. Metro 2033 5760 res min frames win Dirt3 5760 res min frames win Shogun total war2 5760 res win.... no min frame test, that's the first 4 games all AMD HD 7970 wins where it matters the rest is fluff comparing playable frames. Nvidia GTX 680 wins: Batman Portal 2 Battlefield 3 StarCraft II Skyrim Elder Scrolls Civ 5 then onto synth results where AMD dominates. AMD HD 7970 wins: Civ 5 SmallLux 3dmark pixel 3dmark texture fill tessalation normal tessalation max Nvidia GTX 680 wins: AESencryptDecrypt Fluid sim Unique heaven. factoring those results the minimal power consumption difference, the AMD's better OC scaling, AMD's notably lower price and the availability of the HD 7970 and I give the nod to AMD. with performance, price & availability being the most important aspects.

      • flip-mode
      • 8 years ago

      I call troll bait. TechReport’s article is perfectly intelligible for an average middle-schooler.

        • CampinCarl
        • 8 years ago

        I dunno, that 1000/frametime conversion is just /hard/.

        • clone
        • 8 years ago

        call it whatever you want, I go to Anandtech and see a graph and have my answer I get to TechReport, read the text and get to the charts….and have to stop and discern the settings from a separate box, then I scroll down to have a look and then back up to double check the resolution, then I have to stop and concentrate on the stutter graphs comparing high points and their frequency then I go to the “average” graph & then down to the “lag” graph but wait I’ll need to go back up to the screen shot with the settings to triple check because by now I’ve got 3 windows open comparing 2 other websites reviews.

        finally I go back to the selected resolution then reassess the lag graph against the breakdown of a second graph and see if I can distinguish the lag points, then I wonder if it’ll be noticeable then I go back to the average frame graph…..that’s the first game’s charts and now I’m onto the 2nd games set of charts at which point I start the process over.

          • Arclight
          • 8 years ago

          That’s because you get more usable info that you need to consider…..ofc it takes more time. As shown by the “lag” having the best average does not translate to having the best and smoothest game experience.

          While yes it’s not as fastto read and interpret, it is a closer representation of real life performance.

            • clone
            • 8 years ago

            never said the data was worthless, just a pain to absord, my request isn’t to dumb down the info but to find a better way to relate it.

          • flip-mode
          • 8 years ago

          Your laziness is contemptible. You prefer to know and understand less because you have to scroll and concentrate. Amazing. Amazing that you’re not even ashamed of it; nay, you publicly complain about your “option” (reading this site is optional) to have the most thorough and accurate graphics card analysis available /anywhere/. Pfft; good bye.

            • clone
            • 8 years ago

            your anger is sad, your opinion flawed and I’ve not once asked for less data despite the accusation and “pfffft good riddance.”

            on a side note I’m hoping to push TR into finding a better way to relate the data, what’s even sadder than your anger is that instead of striving for improvement you prefer status quo and condemn a request for innovation.

          • MrJP
          • 8 years ago

          Perhaps they should boil it down to a picture of a GTX 670 with “teh winner” and a 7950 with “teh suck” underneath. Would that be more time-efficient for you?

            • clone
            • 8 years ago

            not really, haven’t said anything about not wanting the data but instead am requesting a better way be found for it to be related.

            • MrJP
            • 8 years ago

            OK, I apologise for misinterpreting what you were saying, but what exactly are you proposing? At the moment it sounds like you’re saying you want just as much information, but somehow compressed into fewer plots. I think you need to come up with a specific idea before you can be too critical about how its done at the moment.

            • clone
            • 8 years ago

            I have some ideas.

            the first problem with the layout is the scroll fest, I can’t imagine what it’s like for someone with a 1366 X 720 laptop, I’m using 1920 X 1200 and it’s terrible.

            take the GTX 670 review as a test case, their is a screenshot which is fine but then a comparative mountain of space is taken up with a screenshot of the settings used…. just tell us in the graphs, a single line of text and it’s done, far more useful and saves a lot of space.

            then their are 3 separate stutter graphs at full size below one another, put them side by side and have then shrunk so when the cursor passes over top they expand to full size, worse still 2 paragraphs are below the stutter graphs explaining what they are…… if a graph requires a multi paragraph explanation it should be simplified.

            the answer is already on the side of the graph btw….. frame time in milliseconds, no need for additional paragraphs although they could put “individual frame times in milliseconds” if they liked and if the info is needed put it in the review just prior to the charts not in between the graphs where it’s “getting in the way” as readers stop to read then realize they don’t need to.

            let’s move on to the “Average FPS” chart, add maximum and most notably minimum into the mix like other websites do, the average frame chart can be the most worthy and likely should be the first chart with that additional data as it presents the overall result.

            the 99th percentile chart and the frame latencies by percentile are uninformative because “A” they are just repeating what the first 2 graphs have mentioned and “B” while repeating that data they aren’t showing when and where which is what the previous graphs did, they are redundant….. if it’s felt the % is important put a stat line under the above graphs saying something along the lines of “time spent beyond = “X” % of the time that way the % complements.

            if I see a graph like the HD 6990 stutter graph I’ll know immediately that micro stuttering is an issue throughout.

            then you have the “time spent beyond 50ms which is a nifty graph because 50ms is micro stuttering territory so keep it and with my reccomendations it’d be placed directly below the stutter graphs complementing what they’ve already said.

            then you have time spent beyond 16.7ms which can’t be seen…. keep it, lose it no matter but limited in value at best and misleading at worst because where a video card may have a consistent and smooth 25ms frame rendering or 40fps which is certainly better than playable it’d look on the 16.7 graph like the very worst candidate in the list.

            Batman Arkham city in the same review is even worse because the screenshot is of the complete settings page, fancy? sort of, needed? not at all.

            • flip-mode
            • 8 years ago

            Some of those are not bad ideas, per se, but some ideas I completely disagree with. But really, the article is perfectly legible and intelligible as is, and that’s the biggest problem with your complaints. The second biggest problem with your complaints is that they’re all completely subjective and what do you do if most people disagree with you? What if people prefer scrolling to mousing-over? What if people WANT to see the 99th percentile chart and the frame latencies by percentile? I think those are excellent, personally, and I want to see them. What if most people find it much easier to understand the game settings if they just see them in a screenshot? I’m just curious, do you keep complaining that the article is difficult to read or do you actually start to suspect that reading is just difficult for you?

            • clone
            • 8 years ago

            the goal isn’t a dictatorship, the goal was a discussion, a constructive one hence the reason I didn’t default to “what a mess” in my first post, the current format is a mess with some redundancies in it.

            the 99% charts aren’t required, if you love them then love them but know that the data translated requires but a single line of text and not a chart to convey the exact same information.

            how can anyone require game settings to be displayed in a screenshot versus a line of text over top of each graph? …. so when they are reading a chart and have forgotten the settings all that’s required is to see the top of the chart versus scrolling all the way back to the top and then having to discern the settings from the screenshot which doesn’t just have settings but has a wad of fluff in it?

            the problem with your “I’m just curious” is that it provides 2 flawed options both black (stupid) in a world of grey, is the article difficult to read…. no, I can read, is the article a pain in the ass to read? absolutely.

            I’ve answered this in an earlier response so I’ll just quote it.

            [b<]I get to TechReport, read the text and get to the charts....and have to stop and discern the settings from a separate box, then I scroll down to have a look and then back up to double check the resolution, then I have to stop and concentrate on the stutter graphs comparing high points and their frequency then I go to the "average" graph & then down to the "lag" graph but wait I'll need to go back up to the screen shot with the settings to triple check because by now I've got 3 windows open comparing 2 other websites reviews. finally I go back to the selected resolution then reassess the lag graph against the breakdown of a second graph and see if I can distinguish the lag points, then I wonder if it'll be noticeable then I go back to the average frame graph.....that's the first game's charts and now I'm onto the 2nd games set of charts at which point I start the process over.[/b<] Flip-Mode your response is disingenuous, yet another effort to stifle innovation & creativity for the sake of an argument... no real solutions "per se" just a lot of criticism... "what if", "what if", "what if",.... just more criticism Flip.

            • flip-mode
            • 8 years ago

            LOL, the only thing I’m trying to stifle is unnecessary change. There’s always the risk that when you fix something that’s already working you can actually break it. HTML is pretty easy so I imagine you could mock up a page using your suggestions. Maybe it would be great. If not, no loss, since it’s already great.

            • clone
            • 8 years ago

            “unnecessary change”…. lol.

            thanks for admitting your goal is to dictate what gets discussed.

            • flip-mode
            • 8 years ago

            I call BS again. It’s obvious that your issue is with the amount of data, otherwise you’d have a specific suggestion for how to present it better. Instead, you offered Anandtech as an example, and what does Anandtech do? Less data. They test with more games and more cards but deliver less actual performance data for each game run, nor do they completely expose the testing method. So, yes, you directly implied that you’d like less data.

            • clone
            • 8 years ago

            your lack of insight is typical and expected Flip-Mode, it’s why I usually dismiss your posts even when it appears they may have some truth to them.

            it’s not worth the time to see given the usual silly.

            p.s. while I mentioned I was at Anandtech prior to reading TechReport that is not a claim that Anandtech’s reviews are perfect, only an idiot would imply that just like it took an idiot to imply that I was demanding less data when my opening & later comments were “their has to be a better way”…. “I like the added info”…. what I did come away with from Anandtech was an impression of the cards relative performance until I got to the comparative “mess” that is the Tech Report’s layout.

            • Meadows
            • 8 years ago

            You can still simply look at TR’s FPS charts and skip everything else, if you’re that dysfunctional.

            • clone
            • 8 years ago

            sigh…. very late to the party, equally uninsightful, try reading Meadows because apparently your hands are typing, you should look into that.

            • Meadows
            • 8 years ago

            I’ve read all your crap and I still don’t see the problem. You act as if all the graphs were serially forced down your throat.

            • clone
            • 8 years ago

            “FANBOY”, “TROLL”, “CRAP”, “BS”, “DYSFUNCTIONAL”…. stupid comments by default.

            no thought just tears.

            all this from “their must be a better way” and later in the same post “I like all the data but”…. god forbid a balanced comment lead to a balanced response.

            • flip-mode
            • 8 years ago

            Why are you surprised that is what happens when you call a perfectly intelligible article “an effort to discern” and “[the] mess that is Tech Report’s layout” and “with TR’s it’s a scroll fest page filled ponderous time”? A pat on the back? LOL. Not going to happen.

            • clone
            • 8 years ago

            the best part is that you’re defending the practice of making stupid comments by default.

            to the authors at TR who composed this article, as is always the case with prolonged discussion on the web the dialogue loses it’s moderate tone… the article is ok albeit I’d argue a bit of a mess but my effort was to generate some discussion on how it might be improved, that’s been lost under silly, sorry about that.

            I know I’m right on this because instead of balance, it’s personal attacks, no defense, no subjective, just unrelated, ineffective, uninsightful personal attacks, vapor.

            because it’s not perfectly intelligible, because it is a mess, because it is a scroll fest, because that perfectly laid out article even by TR’s standards requires 2 paragraphs to explain what the article is saying.

            their has to be a better way…. a simpler way and btw I’ve offered up some suggestions above which is constructive unlike your continued worthless silly.

            • flip-mode
            • 8 years ago

            You seem to be missing the fact that I’m not defending stupid remarks, rather I’m criticizing the manner in which you voiced your criticisms. Or is it only OK for you to voice criticisms but not be criticized yourself? Your first few posts that started the conversation had no specific suggestions – just a vague criticism that just had troll-smell all over it. Finally, one of your recent posts (to which I replied) had some specifics to it.

            If you’re going to give criticism then you have to be willing to take it; though, perhaps if you delivered your initial criticism a little more carefully and with more specificity you would not have gotten so much blowback. Your post that listed specific suggestions was pretty good, even if I generally disagree with most of them; there was nothing about the post /itself/ that deserved any criticism.

            • clone
            • 8 years ago

            I’m not going to dance for anyone Flip, neither should you.

            the reality is you default to stupid comments, you’ve defended said actions by claiming a right to default to stupid comments whenever you disagree with a comment…. in this case the claim is because it wasn’t worded the way you’d prefer it it’s ok to become an idiot.

            if you feel the criticism is too vague defaulting to idiot savant accusations “troll”, “lazy”, “contemptible” and then following them up with “I call BS” or “liar” for short and then repeatedly trying to justify your stupid…. only hurts not helps your position.

            if you feel the criticism is too vague instead of defaulting to destructive comments try constructive….. not hard really, quite logical but so much less childish and so many love to be caustic children on the web.

            • flip-mode
            • 8 years ago

            You want to throw a fit because I said “BS”, “troll”, and “lazy”, and then you turn right around and say “stupid”, “uninsightful”, “idiot”, “worthless silly”, etc. Well now you can complain about me calling you a hypocrite.

            You’ve backed away from discussing the actual article and your problems with it and you are now consumed with throwing a tantrum (which is somewhat amusing for me to watch) over what you perceive to be the terrible treatment that you have suffered. You’re not going to die of persecution because I called you lazy (and based on what you said in the post I replied to I stand by that response). Calm down, toughen up, and quit throwing yourself a pity party.

            • clone
            • 8 years ago

            not throwing a fit Flip, I love it when ppl put words on a screen and believe by doing so they must be true. you are guilty of everything you accuse btw but please continue to tell me how it’s right to make stupid comments by default.

            providing proof of hypocrisy is not a tantrum, I mentioned at the end that ppl love to behave as children on the web and here you are.

            Flip you can say whatever, you never once discussed the introduced topic, you readily admitted you didn’t want it discussed by anyone, you readily admitted that you wanted to dictate that no change is required as decided by you, you were a hypocrite long before you mentioned it.

            you came in fighting and now look…. typical, lol.

            were done, note: yeah I know but I condensed & some wording was altered then I figured I’d add were done which bumped it up then I added “lol” which bumped it up again.

            • flip-mode
            • 8 years ago

            Most of that post was unintelligible (oh, the irony). But I did try to directly discuss the main topic. See here:
            [url<]https://techreport.com/discussions.x/22922?post=636918[/url<] It seems excessive to rehash that post, but what the heck, we're well into excessive territory here. My "bottom line" so to speak was that I think you're asking for unnecessary change, and most of what you ask for is nothing more than your individual preference which may or may not be in conflict with the individual preferences of others and may or may not be more work to implement in HTML. Take mouse-over versus scrolling - I often prefer scrolling because it might let me see two graphs at once while mousing over only lets you see one at a time. Then you want to ditch a couple of graphs that I really like, and I figure the authors might like them too since they bothered to put them in there. So then I suggested you mock something up in HTML, but you say you are "not going to dance for anyone". Which is odd to say since it seems to me like you're asking TR to dance for you.

            • clone
            • 8 years ago

            finally something topic related, mature and reasonable…… .after how many stupid attacks, empty accusations and childish justifications and no your link was disingenuous, as for my previous, only a moments thought is required to comprehend the response which was deliberately composed to require that, if you don’t understand, no surprise and no biggie.

            [b<]I think you're asking for unnecessary change, and most of what you ask for is nothing more than your individual preference which may or may not be in conflict with the individual preferences of others and may or may not be more work to implement in HTML.[/b<] why anything else was said and I mean all of the stupid posts on top of the empty text is a boring mystery. p.s. their is no doubt HTML can handle what I've mentioned just fine, noticed the practice used by other websites.

            • flip-mode
            • 8 years ago

            [quote<]p.s. their is no doubt HTML can handle what I've mentioned just fine, noticed the practice used by other websites.[/quote<] That quote represents the whole issue here, which is your reading comprehension. Read my post again and note that I didn't even imply any doubts about what "HTML can handle". What I really said doesn't need repeating or clarification; go read it again. Reading comprehension has been your problem from the outset, and then as soon as you got criticized for it your reading comprehension actually got worse, as you were too busy calling people stupid and furiously protesting anyone who has suggested that the problem was your reading skills rather than how the article was written. Reading comprehension. Get some, bro.

          • cynan
          • 8 years ago

          Damage made it easy for you: The Average FPS graph (practically the only info you get at other sites – though some add min/max FPS as well) is the first figure on each page. Stop there.

            • clone
            • 8 years ago

            I place more weight on the stutter graphs actually with the average frames graphs coming in 2nd and the lag graphs a distant 3rd.

            my comments aren’t about getting rid of data but instead finding a better way to relate it.

            atm their is no quick and easy way to breeze through a review, instead it’s ponderous.

            the GTX 670 review was an admittedly rushed review that lacked data limiting it’s value but even with the extensive reviews I typically read them 2 or 3 times, once quickly because I’m reading numerous reviews to see if other reviews conflict then I’ll revisit and look more indepth.

            with TR’s it’s a scroll fest page filled ponderous time … not because the reviews aren’t information filled excellence but because of the way they are compiled.

      • Laykun
      • 8 years ago

      To be honest average and minimum framerates simply don’t express how the card performs when it comes to smoothness of gameplay.

      Minimum and average framerates express performance for sure, it says it can produce X frames per second. However the important thing about gaming isn’t ONLY average and mimimum, it’s about consistency. Consistent well timed frame delivery will offer a superior experience, and you’re paying money for the experience not the power.

      What’s more important in my opinion is the subjective results from the tests, particularly with SLi and crossfire. From my experience with triple crossfire I have witnessed high frame rates yet jerky stuttery in games, this is produced by poorly timed frame delivery and inconsistent frame rates, and as such I find it important to know whether a card suffers from inconsistency problems and stuttering even though it produces the best avg FPS.

      Companies that build video cards need to focus on delivering excellent fluid game experiences, not maxing out arbitrary numbers on benchmarks and games, I don’t play benchmarks. Currently I believe nvidia leads in this regard.

      You’ll find that the AnandTech review mostly focuses on describing the data they’ve collected, without actually describing the in game experience, we can all read graphs and I doubt many people need the filler commentary they put with the graphs.

        • Airmantharp
        • 8 years ago

        I agree-

        A good place for this is the [H]. They still need an editor, but I lean on their reviews for a ‘closer to the gamer’ perspective.

          • Airmantharp
          • 8 years ago

          Yes, yes, down-rate the [H].

          I could really give a rats about family feuds. They still do good work over there, and definitely add something unique to the community.

    • redwood36
    • 8 years ago

    Hi im looking to upgrade from a GTX285 sometime in the future and have lately stuck with nvidia. As such Im looking at the 670. However looking at anandtech test scores for its capacity for graphics computing is see that AMDs offering is seemingly miles ahead of the 67/8/90. I handle large cmyk 600 dpi files at 14×17 (comic pages) and on my current card I have to cut them down to 300 dpi so that my computer can edit them comfortably. will the 670 be able to keep up with full 600 dpi cmyk image files or should i go for AMD and take what seems to be a cut in power to my favorite games (skyrim, diablo 3 given how they performed with starcraft, and left 4 dead 2). I am also keenly interested in increasing my performance in witcher 2 (which btw i requested before a little look into how these cards perform on, being that its one of the few non fps titles that offer a lot of eye candy with dx11 etc options)
    However if any of this means I have to sacrifice on my capacity to handle the large images, I will definitively have to go for AMD.

      • Yeats
      • 8 years ago

      What software, and what are your computer specs? You should make a post in the forum if you need advice.

        • redwood36
        • 8 years ago

        right, might go to the forum, but i just thought this was nice and informal and quick. im running a i7 920 with 6gb ram on an asus 96t deluxe v2.
        and in general the adobe suite. for the comic pages its photoshop cs5.

      • xeridea
      • 8 years ago

      Nvidia went to crap with GK104 as far as GPGPU is concerned, while AMD made huge leaps. If you do any sort of non gaming with your cards, skip Nvidia this gen.

        • Airmantharp
        • 8 years ago

        Or catch (at least) the hyped-up GTX560 Ti 448 with the Fermi GPU in it, if one happens to go on sale. the GF110 still has all of Nvidia’s GPGPU chops.

        But for current-gen GPGPU, AMD takes the cake.

        • redwood36
        • 8 years ago

        im really getting that impression. However im not sure if my work really qualifies as intense GPGPU work. normally thats for 3d and movie processing, comics are comparatively low tech.

          • Airmantharp
          • 8 years ago

          To be honest, I have no idea how a GPU has anything to do with limiting the DPI you can work with. To me, it sounds like it’d be CPU bound, or at least VRAM bound (assuming you have plenty of RAM).

    • syndicatedragon
    • 8 years ago

    I’d be really nice to see the price/performance scatter charts for all of these cards.

      • khands
      • 8 years ago

      We’ll probably get that when the full review shows up.

    • Chrispy_
    • 8 years ago

    Is anyone surprised that the board is really short?

    This is a die-harvested GK104 that failed to make the grade as a GTX 680.
    I still see the GK104 as an evolution of the GF104 that powered the GTX 460, that’ sub-200 mid-range fermi, and GTX 460 never needed a big board. Whilst Kepler has more logic than than GF104, it’s also on a substantially smaller 28nm process.

    Given that two of these chips comfortably sit on an 11″ GTX 690, it’s refreshing to see that they’ve squeezed it all down for those of us with mATX cases.

    • Unknown-Error
    • 8 years ago

    Thank you Scott for the review!

    PS: Looks like we can forget about the GTX 680. The difference is pretty small compared the 680 and USD 100 less? Not to mention gives the 7970 a nice spanking on several occasions.

    • Tristan
    • 8 years ago

    This card is very short. There is not only shortages on 28 nm chips, but also shortages on PCB.
    Waiting for 660…

      • Silus
      • 8 years ago

      Next is probably GTX 660 Ti, which rumors point to being another cut down version of GK104 (1152 SPs maybe?)

      GTX 660 should debut only once we know more about GK106.

    • jjj
    • 8 years ago

    Just stopped reading when you mentioned you are using only 1 res …

      • cygnus1
      • 8 years ago

      panzy

    • flip-mode
    • 8 years ago

    Yep, just want to say thank you, Scott, for putting in so much work over the last little while. Seriously. I hope you are now taking a few days off.

      • CampinCarl
      • 8 years ago

      Agreed. Even though I would love to see lower-res results for the 670, I think someone else could handle it…Scott can take a vacation!

    • DPete27
    • 8 years ago

    Yes please

      • cchsballa0655
      • 8 years ago

      I’d say more like slightly below the 7970 lol….if we are talking normal resolutions.

    • Deanjo
    • 8 years ago

    It should make for lots of room in the case once they come out with a water cooled version.

    • tfp
    • 8 years ago

    Thank you for putting the review out right away I think it shows everything needed even if it’s missing standard resolutions and doesn’t have the largest list of games tested. A follow up review will be icing on the cake.

    I’m happy to read a review of the card here first.

    • ssidbroadcast
    • 8 years ago

    [quote<]The board is only about 6.75" inches long, shorter than the world's tolerance for Nickelback[/quote<] It's not that short.

      • MadManOriginal
      • 8 years ago

      twss

      • Silus
      • 8 years ago

      And when using the TMI scale it’s longer 🙂

    • yogibbear
    • 8 years ago

    Anand’s review is up for anyone that wants < 5760 resolutions… though results are generally the same across the board except where on game favours either team slightly.

    Battlefield 3 heavily advantages the 670 over the 7970. Other games fare differently. If you play BF3 a lot this is probably the card to have. Otherwise YMMV.

    The flipside would be if you play a lot of Total War: Shogun 2… then at the moment the 670 doesn’t enjoy that game at all and AMD is on to a winner across the board.

      • Firestarter
      • 8 years ago

      Ugh, at 1920×1200 their BF3 results really are quite devastating for the Radeons 🙁

      edit: Does anyone know how Anandtech tests BF3? Maybe I’m blind but I can’t find whether they test the SP, MP or some cutscene or something.

      • Washer
      • 8 years ago

      The TR review doesn’t contain 1920×1200 results but I’m beyond confused why TR’s 5760×1200 numbers put the Radeons above the GeForces and Anandtech has results that go significantly the other way.

        • Damage
        • 8 years ago

        It’s easy to become confused if you assume that vastly different testing conditions will always produce the same results. 😉 That’s not always the case. Testing a game at different settings, on different levels, on different test systems, with different techniques will produce… different results.

        I’d tell you more about the possible sources of the differences, but it’s impossible to know how AT tested by looking at that review. There’s no mention of the map used, whether it’s the SP or MP campaign, of what tool they used to capture frame rates, what duration of gameplay they captured, how many samples they took of the same test run, whether they actually played to game or just benchmarked a cut scene. At least, if it’s there, I’ve missed it.

        We did find that the GTX 680 was a little quicker than the 7970 in this same test scenario when we tested with different drivers at four megapixels for our initial GTX 680 review. The 7970 seems to do relatively better here, either because of driver changes or because of the move to 5760×1200. Since the 7970 does have quite a bit more memory bandwidth, the relative improvement would seem to make sense.

        Anyhow, although we didn’t see this problem in the level we tested for today, I would caution you against considering FPS averages like they’re reporting a good indication of overall BF3 performance. Multiple TR editors on multiple occasions in multiple levels of BF3 with various driver revs have found that GeForces exhibit some big frame latency spikes. See here:

        [url<]https://techreport.com/articles.x/22573/6[/url<] [url<]https://techreport.com/articles.x/22151/6[/url<] [url<]https://techreport.com/articles.x/22473/5[/url<] The GeForce cards produce nice FPS averages, but their playability is always actually not as good. An FPS number won't tell you that.

          • Washer
          • 8 years ago

          Results wasn’t the best word, I think trends is probably more what I had in mind. Meaning, I would expect the GTX 680 according to AT’s review to be generally faster in BF3 than the HD7970, but by your review I would expect the HD7970 to be the generally better choice.

          I’m a big believer in the TR testing methods, but this significant difference between conclusions drawn from either review makes me really want to see multiple different testing scenarios done in a game. It’s hard to draw any conclusions when in the back of my mind I know the results from a game could be different if a different level or section of a map were tested.

            • Damage
            • 8 years ago

            Well, results are not trends. We try to test in ways that are representative of typical gameplay and to document our methods well so folks know the deal, but we can’t be assured they will predict performance everywhere. (Heh, and certainly not at some other website’s offices.)

            That said, I’ll try testing a different area next time out. We can’t do multiple areas for each game in every review. But we can vary over time. Could prove helpful.

            Anyhow, hope you’re less confused now.

          • APWNH
          • 8 years ago

          Thank you for pushing the envelope like this by doing proper frame time profiling. It’s the way it should be done, and it’s just one of the many reasons why TR is the best tech review site. The raw fps numbers posted by the GTX 670 made me question my decision to order an HD7950 a few days ago, but it’s clear to me that the GCN architecture is still very, very competitive.

          Now to go about deciding if I should go 3x1920x1200 or 2560×1600 or 2560×1440…

    • kamikaziechameleon
    • 8 years ago

    so 450 for this… I’m a bit confused as nvidia has a gaping hole in its lineup still with this hitting the 450 price point.

      • yogibbear
      • 8 years ago

      gtx 660 perhaps?

      • Silus
      • 8 years ago

      450 ? The card’s MSRP is $399

    • BlockheadBrown
    • 8 years ago

    The performance is great. The price/performance ratio is not all that great though considering what can be done when you overclock a Radeon 7850. My last card was a GeForce 560 Ti. If anything, I honestly tend to lean towards Nvidia. The OC potential of the 7850 along with the comparatively reasonable price (the 7850 should be about $50 cheaper, IMHO) and a lack of new competitive products from Nvidia in the same price range are what lead me to go red this time around.

    • yogibbear
    • 8 years ago

    After the sexiness that was the GTX 690 build quality… I think I have been desensatized as this thing looks FUGLY!!!! Cannot wait for custom PCB designs similar to my 560 Ti from Gigabyte.

      • NeelyCam
      • 8 years ago

      It’s inside your case – why do you care? Unless, of course, you’re one of those whose PCs have a bunch of LEDs inside a case made of plexiglass

        • yogibbear
        • 8 years ago

        Nah my case is just a big black heavily sound proofed box, but I still like knowing that it would look fancy inside if I took the side door off… makes me feel all warm and fuzzy inside. 🙂 Kinda like owning nice undies and wearing them but not showing people in public….

      • Metonymy
      • 8 years ago

      I think that ‘desensitized’ is probably what you haven’t been…

    • Arclight
    • 8 years ago

    Gotta go Krogoth on this.

    Maybe i was too harsh, i mean at 1080p, for people more into BF 3 than Cysis or Metro, this is clearly the card to have, unless it’s still over ur budget and cheaper cards handle ur fav. games at 1080p just fine.

    But i still dislike the plasticky back and super short PCB. What’s up with that?

    • yogibbear
    • 8 years ago

    It is rather telling that a mid-high end GPU can run even the latest games like Skyrim on 3 screens fully maxxxxxxxxed out!

    Not much to stress test GPUs beyond Metro 2033 & Battlefield 3 these days.

      • Madman
      • 8 years ago

      Say thanks to consoles 🙂 Or not…

        • Duck
        • 8 years ago

        Or the sky-rocketing cost of developing a next gen AAA game that brings the latest hardware to it’s knees….

          • cegras
          • 8 years ago

          How much money goes into asset creation related to high fidelity graphics?

            • Duck
            • 8 years ago

            Loads and loads. The investment needed has exploded exponentially over the years. So much so that they sometimes seem to loose all sight of innovation and fun gameplay in the pursuit of graphical fidelity oneupmanship.

            • cegras
            • 8 years ago

            I’m skeptical. The ratio of games released to revolutionary engine based games is pretty low.

            • Duck
            • 8 years ago

            [url<]http://hothardware.com/newsimages/Item20295/GameCosts.png[/url<]

            • Bensam123
            • 8 years ago

            Access denied.

            • Duck
            • 8 years ago

            Click into the URL bar and hit enter.

            • yogibbear
            • 8 years ago

            Make the x-axis be the date (rather than a bar graph), then overlay inflation, no. game sales, no. console sales. Bet the picture will look different.

            • Bensam123
            • 8 years ago

            Yup and even those on good engines have crappy textures and effects. Using someone elses engine takes out the majority of the work too. Most companies don’t build a engine from the ground up for a game.

            • SPOOFE
            • 8 years ago

            [quote<]sing someone elses engine takes out the majority of the work too.[/quote<] Incorrect. Designing all the models and textures are the majority of the work.

            • Bensam123
            • 8 years ago

            Incorrect. I wont deny that texturing and making models is manpower intensive, but it’s not the same as building an engine like the cryengine, ut3, or source engine from the ground up. There is a reason games with new engines take quite a bit longer to produce then just reskinning a game with an existing engine.

            • Bensam123
            • 8 years ago

            Where and when did these games come out? The only one that remotely fits this argument is Crysis… Do you know how long ago Crysis came out? 2007. That makes this a five year old argument based on one game. Two if you add Crysis 2, but the graphical fidelity in that game is nothing compared to the level Crysis was at when it came out and Crysis 2 is consolized.

            • SPOOFE
            • 8 years ago

            [quote<]Where and when did these games come out?[/quote<] It was called "the '90s". I know, it's hard to see it way back there with those rose-colored glasses on.

            • Bensam123
            • 8 years ago

            I’m glad you’re proud of idioms, but it actually has to apply to what you’re saying…

            I was around during the 90s and playing games… That was during the hay day when games were improving right up till consolization took over. People still tried making things new and different, not just rehashing the same crap with the same gameplay and same graphical quality.

          • Bensam123
          • 8 years ago

          This lame ass argument needs to die in a fire. You can do a lot of things that raise the graphical qualtiy without going into brute forcing high resolution texture packs, which take a lot of time to make. Effects for instance… The only ‘rising cost’ associated with game development now compared to 5-6 years ago is ultra high resolution textures, which take more people and time to make. Games have NOT become more complicated, they’ve become quite a bit less complicated, unless you start talking about MMOs.

          Go play Hard Reset, which was made by an indie company and then come back here with this argument.

            • SPOOFE
            • 8 years ago

            [quote<]This lame ass argument needs to die in a fire[/quote<] You're delusional.

            • Bensam123
            • 8 years ago

            With that sort of counter-argument, apparently I am the one who is.

    • Pez
    • 8 years ago

    Shame there isn’t even a quick-and-dirty overclock, but appreciate the fast work in turning this review around!

    The 670 certainly does look attractive compared to the 680 price/performance wise. This can only bud well for the true mid-range 660, as long as it’s not gimped too much.

    Hmm… might have to put in some overtime.

    • yogibbear
    • 8 years ago

    I was just catching up reading the tweaktown early preview/review shenanigans and then created my new build and noticed my online store had listings for gtx 670’s and a pre-order button!!!! Then I come back here and BAM! Nice!

    • flip-mode
    • 8 years ago

    Sheesh, this thing makes the GTX 680 pointless.

    • ALiLPinkMonster
    • 8 years ago

    I’m starting to think that we will get a GTX 650 that is actually worth buying this time around.

    • Ryhadar
    • 8 years ago

    Wow, look at the length of the stock PCB. Has a card this strong had a PCB this short in recent memory?

    A vendor could easily turn this into a half-height card, I bet, which would be great for my brother’s machine.

    I understand why they opted for the blower style but it is a shame it adds so much length to the card.

      • MadManOriginal
      • 8 years ago

      Yup, I will really like seeing what can be done with custom coolers and such a short PCB. Crads like that will be an easy recommendation for SFF cases. It is a bit pricey these days though when for single-monitor gaming a lesser card will typcially be fine.

        • Ryhadar
        • 8 years ago

        Absolutely my thoughts as well.

        I have a modded SG02 and while I currently have a 10″ Radeon 6870 in there it’s a tight fit. I certainly wouldn’t complain of a faster, shorter card.

      • phileasfogg
      • 8 years ago

      >>> A vendor could easily turn this into a half-height card, I bet, which would be great for my brother’s machine.

      I doubt it. It would be devilishly hard to squeeze in 8 GDDR5 chips into a 1/2-height card.

      Two other problems:

      a. 1/2-height cards have a *maximum* length specification stated in the PCIexpress CEM (Card Electro Mechanical) 2.0 spec : it’s 6.6″ (167.65mm). So, at 6.75″ this card would “make the cut”, IF the shrinkage in height (from 4.376″/111.15mm to 2.731″/68.90mm) does not cause the new card layout to become longer.

      b. 1/2-height cards *also* are subject to an upper bound on power-consumption, as stated in the aforementioned CEM spec. No way that number is ~ 170Watts.

        • cynan
        • 8 years ago

        Who cares about specs almost certainly put in place for storage controller cards, etc, in server environments? Are graphics card OEMs really constrained by these specs? Electronically, I don’t see why a shorter card wouldn’t be able to handle as much power as a taller one. It might add a bit of onus on the consumer to make sure that their smaller enclosure had adequate cooling, but I don’t see that as a deal breaker.

        You may have a point about getting 8 GDDR5 chips with a short enough signal path on a half-height pcb though.

    • flip-mode
    • 8 years ago

    Woah! This one caught me by surprise. Need to find time to sneak-read this during work today!

    • Firestarter
    • 8 years ago

    I bought my HD7950 3 months ago, no regrets! That said, I hope this 670 has better availability than the 680, or it would be a hollow victory.

    • chuckula
    • 8 years ago

    The GTX 670.. unavailable wherever fine graphics cards are sold.

      • Firestarter
      • 8 years ago

      I’m seeing Palit, Gainward, EVGA and MSI cards in stock right now.

        • chuckula
        • 8 years ago

        Wait a little while.. for them to run out.
        Edit: Not even listed on Newegg or on Amazon.. where exactly are these in-stock cards? At least the GTX 680 was listed on launch day even if the stocks ran out quickly.

        • Prospero424
        • 8 years ago

        Where are you seeing them in stock?

          • Duck
          • 8 years ago

          Practically every computer shop in the UK has stock of several different 680GTX SKUs at any time that never sell out completely.

            • Madman
            • 8 years ago

            That’s because they cost almost 2x as much as in US. Almost no one wants to pay THAT much.

            • Duck
            • 8 years ago

            It’s more like +40% rather than twice as much. People don’t buy them, I’m sure. So people don’t spend money and we head into recession again. Our government is full of morons. Hong Kong has super low taxes on the poor and low taxes on the rich. They went from a shanty town to a really nice well developed city in just a few decades. Just copy them, it’s not hard. [/tangent much]

            • willb1967
            • 8 years ago

            Did you know Hong Kong is a tax haven for rich people/multinationals with Asian interests to wash their money?

            • Airmantharp
            • 8 years ago

            Did you know this isn’t related to the thread?

            • Duck
            • 8 years ago

            No? But that sounds ok to me unless I am missing something.

            • willb1967
            • 8 years ago

            Yep, but you’re not the only one. google “tackle tax havens” and have a read/listen it’s a bit shocking.

            PS. The 680’s have always been in stock at overclockers.co.uk but I’m glad I waited for the 670’s

            • Duck
            • 8 years ago

            The rich always find ways to dodge tax. F1 drivers move to Switzerland for example. They wouldn’t have to do that if more countries were like Hong Kong.

            • Bensam123
            • 8 years ago

            It’s right because everyone does it?

          • Firestarter
          • 8 years ago

          Germany, several online shops. They’re slightly cheaper than the HD7950 was at launch, to give you an idea.

            • chuckula
            • 8 years ago

            Once again going to prove my theory that Germans Love [s<]David Hasselhoff[/s<] Nvidia cards.

            • Duck
            • 8 years ago

            If they are in stock then people are not buying them up. Doesn’t sound like love to me.

            • jensend
            • 8 years ago

            Reminds me of what Tim Schafer said about adventure games when pitching his Kickstarter effort: “they exist [only] in our dreams, in our memories… and in Germany.”

      • Airmantharp
      • 8 years ago

      Just picked up an EVGA Superclocked for $419 shipped off the ‘Egg. They had several models available, in stock, including double- and triple-fan coolers.

      • Prospero424
      • 8 years ago

      Yeah, I just grabbed the MSI model off of Newegg. I’m curious to see how long the current stock lasts.

      • Yeats
      • 8 years ago

      My brother bought one on Newegg about 2 hours ago, and GTX 670’s are still in stock there.

    • Dposcorp
    • 8 years ago

    FIRST!!!111 Now off to read

      • Chrispy_
      • 8 years ago

      First to minus fifteen. Congratulations!

        • Metonymy
        • 8 years ago

        Don’t be unkind… some of us just need a little bit more time to ramp up our cooling systems to handle the stress of the actual reading part.

          • Chrispy_
          • 8 years ago

          It’s more that there are no post numbers in the default comments view.
          Claiming first is frowned on anyway, but claiming first in a numberless system is [i<]special needs[/i<] special.

          • NeelyCam
          • 8 years ago

          [quote<]some of us just need a little bit more time to ramp up our cooling systems to handle the stress of the actual reading part[/quote<] I understand - website rendering really trips Brazos chips, and those tiny blower sound like airplane turbines... Maybe consider getting an Intel system..?

      • sweatshopking
      • 8 years ago

      i’ll plus you, my first brother

Pin It on Pinterest

Share This