GeForce GTX 980 Ti cards compared

The GeForce GTX 980 Ti is arguably the best video card you can buy right now, by most standards. AMD’s Radeon R9 Fury X has arrived and shown itself to be competitive but not quite dominant. In most games, the 980 Ti is faster.

Sure, you could pony up a grand for a GeForce Titan X with 12GB of memory in order to get the “ultimate,” but the 6GB of GDDR5 aboard every GTX 980 Ti appears to be all that anyone needs in current games. More importantly, perhaps, the Titan X is limited to Nvidia’s reference design, while the GTX 980 Ti comes in a bunch of different flavors, with slick aftermarket coolers, custom board designs, and tweaked clock speeds. If you’re looking for the finest, most extreme single-GPU video card on the market, look no further than cards like these:

Pictured above are GTX 980 Ti cards from the likes of Asus, MSI, EVGA, and Gigabyte. They’re all different, but they all have their sights set on this guy:

That’s Nvidia’s GeForce GTX 980 Ti reference design, with the familiar stock cooler that Nvidia has been shipping for several years. The custom GTX 980 Ti cards we’ve assembled all cost about 10 to 20 bucks more than the reference version, and they all aim to improve upon the vanilla card in multiple ways. Have they succeeded? More importantly, which one of them might be the best choice for you? Let’s take a look.

EVGA GeForce GTX 980 Ti SC+

The SC+ is one of five different GTX 980 Ti cards that EVGA is currently selling at online retailers. That thought might be a little daunting, but don’t worry. The SC+ is selling for $679 at Newegg, just ten bucks more than cards based on the reference design, and it’s likely to be the best value among EVGA’s offerings. The others will set you back more and include exotic options like liquid cooling. For most folks, the SC+ will probably be the card to buy.

This beast is a nice step up from the reference design, starting with base and boost clock frequencies of 1102 and 1190MHz, roughly 10% above the reference defaults of 1000 and 1076MHz. Believe it or not, those clock speeds make the SC+ the most modestly clocked card among this bunch, mostly because Nvidia builds quite a bit of headroom into its reference clocks in order to make products like these possible.

Of course, default clocks aren’t terribly important if you’re willing do a little overclocking. The SC+ could help facilitate that, ahem, questionable practice with the help of EVGA’s ACX 2.0+ custom cooler. This cooler features twin fans and a total of five different heatpipes snaking across above the GPU. As you can see below, three of those heatpipes run straight through the length of the card. EVGA says the lack of bends in the pipes improves heat transfer efficiency by six percent. We’ll have to see how that works out in practice.

The ACX 2.0+ cooler also participates in one of my favorite innovations of recent years in video cards. Its fans are thermally controlled and don’t spin at all when the GPU’s temperature is below 60° C. That means the card should be essentially silent during desktop work and even during light gaming. I believe every card in this group, with the exception of the reference board, has a similar semi-passive fan policy, so it’s not unique to EVGA. But yeah, it’s nice.

Another of this EVGA card’s virtues is its compact size: at 10.5″ long and no taller than the PCIe slot covers, the dimensions of the SC+ are virtually identical to the reference card’s. That means this puppy should slide into a broad range of PC enclosures without any drama. The SC+ also has a six-plus-eight-pin power input config. The rest of the aftermarket cards we have on hand require dual eight-pin inputs, so EVGA wins points for potentially working with a broader array of power supplies.

Unlike the reference card, the SC+ ships with a handsome backplate meant to protect the rear of the card from an errant screwdriver tip.

EVGA bundles the SC+ with two notable pieces of software beyond the drivers. The firm’s Precision X utility allows for extensive overclocking, tweaking, and monitoring, and it’s one of my favorite utilities of its type. Also, right now, this card is available at some retailers with a bundled copy of Metal Gear Solid V: The Phantom Pain, which I was surprised to learn is not a game about a knee replacement.

 

Gigabyte GTX 980 Ti G1 Gaming

With a 1152MHz base clock and a potent 1241MHz boost speed, Gigabyte’s take on the GTX 980 Ti is pretty aggressive, although its memory clocks are bone stock. You’ll pay a little more for this card right now; it’s listed for $689.99 at Newegg, although it’s currently out of stock. Truth be told, GeForce GTX 980 Ti cards generally seem to be in high demand. As a result, all of the cards we’re testing may be subject to limited availability and potential markup from retailers while supplies are tight.

 

It’s not hard to see why this thing would be in high demand. We’ve appreciated the merits of Gigabyte’s graphics cards for several generations. The company’s triple-fan Windforce cooler is back for another, erm, spin, this time with some silver cladding added atop the black shroud to give it an updated look.  I think the new look works for it, but I think I still prefer the all-black look overall.

The G1 Gaming is unique among 980 Ti cards because it sports a second DVI output courtesy of Gigabyte’s Flex Display arrangement. If you have multiple fancy monitors that happen to be a bit older, the G1 Gaming might be the path of least resistance, since it would require one less DVI adapter.

Gigabyte claims this Windforce 3X cooler can dissipate up to 600W of heat, which is about twice what it would ever be asked to do aboard this card. Then again, I think these fans probably only ever reach about 50% of their peak speed, so it all makes a sort of sense. As with the other cards, the fans spin down when the GPU’s temperature is low enough, plunging them into silence.

You can see the five heatpipes running across the generously-sized copper plate above the GPU in the picture. Despite its beefy nature, the Winforce 3X is actually relatively compact, as these things go. The cooling shroud extends just a quarter-inch above the PCIe slot covers, way less than some of its competitors. Instead, the shroud steals some additional space lengthwise, making the G1 Gaming roughly 11.1″ long, over a half-inch longer than the reference boards. For a lot of today’s ATX enclosures, the added length won’t matter, but you might want to do a quick measurement before you buy.

That “Windforce” lettering atop the card lights up, as in past products, but this time, there’s a new twist: custom lighting. The deep blue hue on older cards was nice, but it didn’t match well with the red and white hues that dominate PC components these days. Gigabyte has taken notice and blessed its GTX 980 Ti with a rainbow of possibilities. I’m weirdly jealous not to have this color-matching option in the Damagebox. Users can choose from seven different colors using Gigabyte’s OC Guru II tweaking utility. Those “silent” and “stop” lights are always white, though, and come on whenever the fans spin down.

The utility can also switch the LEDs to always off, always on, or “auto” mode, in which the lights glow only when the GPU is busy.

In addition to OC Guru II, the G1 Gaming currently comes with the same Metal Gear Solid V bundle as the EVGA card, at least at certain stores.

 

MSI GTX 980 Ti Gaming 6G

MSI video cards are easy to spot with their distinctive red-and-black color schemes, dragon-themed iconography, and more pipes than a Colorado dispensary.

The GTX 980 Ti Gaming 6G will currently set you back $679 at Newegg, yet it has one of the most aggressive combinations of cooling and clock speed in this bunch. The Gaming 6G’s base and boost clocks are 1178 and 1279MHz, respectively, and its memory runs at 7.1Gbps, a tick above the stock 7Gbps speed.

Heat removal comes courtesy of the flashy Twin Frozr V cooler. The two fans on this puppy are independently controlled, so each one only spins as fast as needed—and they stop rotating entirely when not required. Notice that there’s a shiny area with a kink in it in every other fan blade. MSI says the mix of two blade types allows for better dispersion of air across the heatsinks. Makes sense, I suppose.

Both the Gaming 6G board and the Twin Frozr cooler are larger than the 980 Ti reference design. The heatpipes curve up above the PCIe slot cover tops by about 1.25″, and the board’s about a half-inch longer than the reference one at roughly 11″.  This combination of length and height ought to fit well in the majority of today’s popular mid-tower ATX cases, but it could prove difficult to squeeze into mATX enclosures or anything that’s less than spacious.

Pull off the cooler, and you’ll see that MSI’s card “only” has four heatpipes, but a couple of those pipes make an S-curve and snake into the heatsink fins in two places. I don’t want to give away too much, but this arrangement and the dual blade types really do seem to work well together, as we’ll soon see.

The entire rear expanse of the Gaming 6G is covered by a protective backplate. That little MSI dragon logo you can see peeking up between the SLI connectors and the heatpipes is illuminated by a white LED, and MSI’s Gaming app lets users choose from a range of lighting effects like “breathing,” flashing, and static illumination.

In addition to that Gaming app, MSI offers its excellent Afterburner tweaking utility for overclocking, a six-month subscription to XSplit Gamecaster for live streaming via Twitch, and that familiar MGS V bundle at select retailers.

 

Asus Strix GTX 980 Ti OC Edition

If you’ve been reading TR the past few weeks, this sight may look familiar to you.

That’s a video card adorned with Asus’ extra-beefy DirectCU III cooler. We’ve seen this same cooler deployed on the Radeon R9 Fury and 390X. Here, the new Asus cooler design makes it way to the GeForce side of the aisle aboard the Strix GTX 980 Ti. The Strix is a relative newcomer to the scene and appears to be out of stock most places, but it’s listed for $669.99 on Amazon.

Thanks in part to that humongous cooler, the Strix has easily the highest default clock speeds of any card in this group, with a 1216MHz base and 1317MHz boost, along with 7.2-Gbps memory.

In order to achieve that feat, Asus has given the Strix an outsized custom circuit board with a massive 12+2-phase power delivery section. The card is nearly as tall as the cooler, and the cooler sticks up 1.7″ above the top of the PCIe slot cover. At 12″ long, the Strix is both the longest and tallest card of this bunch—and it’s not really close.

Asus isn’t kidding about the DirectCU thing, either. The copper heatpipes really do come into contact with the surface of the GPU, although only three of the card’s five pipes appear to touch the chip in this case.

 

If you require a light show from your graphics card, Asus has you covered. The Strix logo atop the cooler glows and pulses like a heartbeat. Around back, the board is clad with a metal backplate, just like the other guys.

Asus supplies its own config and overclocking utility, known as GPU Tweak II, and it also includes a full year of XSplit Gamecaster service with the purchase. Asus looks to be participating in the MSG V bundle at Newegg, as well.

 

Performance

We’ve explored the performance of the GeForce GTX 980 Ti extensively, using advanced metrics, in our initial review of the GPU and, most recently, in our Radeon R9 Fury review. I won’t repeat that sort of in-depth testing here, but I do want to take a look at the relative performance of these cards. To do so, I tested with the built-in benchmarks in Shadow of Mordor and Tomb Raider. Both games were configured to use their “Ultra” image quality presets. These simple FPS-based tests should be sufficient for comparing against different implementations of the GM200 graphics chip.

All of these cards deliver a substantial boost, at the default clock speeds, compared to the stock GeForce GTX 980 Ti reference card. Given that these are actual products you can buy for $10-20 more than the 980 Ti’s base price, that’s kind of a big deal.

The Asus Strix captures the top spot among the 980 Ti offerings we’re comparing, with the Gigabyte and MSI cards firmly in second and third place, respectively. Although the MSI card has slightly higher base and boost clocks (1178/1279MHz) than the Gigabyte (1152/1241MHz), the Gigabyte looks to be a little faster in these games. Since the GPU’s actual clock speed is determined by Nvidia’s GPU Boost algorithm, the question of clock speed is more complex than base and boost speeds alone.  We’ll explore that dynamic more in a moment.

While we’re here, though, I can’t help but notice that these hot-clocked GTX 980 Ti cards upset the order of things at the top of the GPU stack. All of them are faster than our GeForce Titan X. Also, these cards improve Nvidia’s competitive position compared to the Radeon R9 Fury X. In Shadow of Mordor at 4K, the reference card is just a hair slower than the Fury X, but the Asus Strix outguns the Fury X by ~14%. This outcome is a natural consequence of the fact that Nvidia builds some headroom into its GPUs for its board partners to exploit.

Power consumption

Pushing a big GPU like the GM200 to its limits does impose some power costs. When running Crysis 3, our GPU rig’s total power consumption rises by at least 30W with any of these aftermarket cards installed in place of the reference GTX 980 Ti. The Asus and Gigabyte cards require another 20W beyond that, pushing their power use beyond that of a Radeon R9 Fury X. The total of 378W still seems frugal next to the 449W of power draw from the R9 390X-based system, though.

Noise and temperatures

All of these custom GTX 980 Ti cards spin their fans down at idle, so they’re all essentially silent at the Windows desktop or with the display in power-save mode. Any differences in the readings on the decibel meter likely have to do with the noise floor here in Damage Labs fluctuating slightly. With the right setup, we might be able to measure any slight coil whine or chatter happening on each card, but we’re talking about incredibly subtle stuff at idle.

The 35-dBA reading for the R9 Fury X, by the way, is likely driven by its pump whine. Some Fury X cards have this problem, and ours is among them.

Welp, every one of the aftermarket cards is quieter than the reference GTX 980 Ti under load. The big revelation here is the MSI Gaming 6G, which stands out by being incredibly quiet. In fact, the air-cooled MSI card generates less noise than the liquid-cooler Radeon R9 Fury X, which is quite a feat. The Asus and EVGA offerings aren’t far behind.

Gigabyte’s default cooling policy is much more aggressive than the other cards’. That results in more noise and markedly lower temperatures than the rest of the pack. Even then, the G1 Gaming registers lower on the decibel meter than both the reference GTX 980 Ti and the Titan X.

 

Overclocking

The process of overclocking a GeForce GTX 980 Ti isn’t especially difficult since it all happens in software, but it’s not terribly straightforward. Nvidia’s GPU Boost algorithm factors in multiple variables in order to determine the best possible clock speed from moment to moment, and a number of things can limit clock speeds: temperatures, power limits, voltage peaks, frequency caps, and so on. Current GeForce cards come with two rated clock speeds: a “base” clock that is a guaranteed minimum and a “boost” clock that represents a typical speed while gaming. (Yep, the boost clock is not a maximum. The GPU can range higher than that under the right conditions.)

As you might be gathering, actual GPU clock speeds will vary depending on the workload. Before overclocking these cards, I decided to establish a baseline by seeing what speeds they run in a very demanding program: MSI’s Kombustor, specifically the “Furry PQTorus” test based on FurMark. This application represents a worst-case thermal scenario for modern GPUs and pushes things much harder than any game I’ve seen. We’ve used Kombustor in past reviews for this very same purpose.

Here’s how each of the cards handled in FurMark.

  GPU

base

clock

(MHz)

GPU

boost

clock

(MHz)

Memory

speed

(GT/s)

Kombustor

GPU

voltage

Kombustor

GPU

clock

(MHz)

Reference GTX 980 Ti 1000 1076 7000 1.018 1189
Asus Strix GTX 980 Ti OC 1216 1317 7200 1.030 1240
EVGA GTX 980 Ti SC 1102 1190 7000 0.999 1050
Gigabyte GTX 980 Ti Gaming G1 1152 1241 7000 1.003 1177
MSI Gaming GTX 980 Ti 6G 1178 1279 7100 1.043 987

Right away, you can see that things get a little weird when we push these big GPUs with an atypical and very intensive workload. The delivered clock speeds and voltages vary widely.  Only the reference card, with the tamest tuning of the bunch, runs at a speed higher than its boost clock. You’ll know from the previous page that these numbers aren’t a very good predictor of in-game performance. I’ll probably want to switch to a more typical workload next time I decide to overclock a batch of big GPUs.

Regardless, Kombustor is still a pretty good stress test, if nothing else. I was able to use it for that purpose while seeking the max possible clock speeds for each of the cards.

I decided to use EVGA’s Precision X tool for overclocking this time around. Here’s what I was able to achieve by prodding each of these cards for a while. Of course, since this overclocking and is in no way guaranteed, your mileage may vary. The peak possible clock speed at a given voltage will vary from chip to chip.

  GPU

clock

offset

(MHz)

Boost +

offset

(MHz)

Memory

speed

(MT/s)

Kombustor

GPU

voltage

Kombustor

GPU

clock

(MHz)

Kombustor

GPU

temp. (°C)

Reference GTX 980 Ti +270 1346 8000 1.012 1101 84
Asus Strix GTX 980 Ti OC +60 1377 8000 1.055 1328 83
EVGA GTX 980 Ti SC +140 1330 8000 0.999 1177 77
Gigabyte GTX 980 Ti Gaming G1 +120 1361 8000 1.074 1354 74
MSI Gaming GTX 980 Ti 6G +60 1339 8000 1.043 1215 78

The first thing you might notice in the table above is that third column, “boost + offset.” That’s effectively the peak speed I was able to wring out of each product, and there’s not much distance between the slowest and fastest options. Cards like the MSI with higher default clocks would only tolerate smaller offsets, while conservatively tuned boards like the reference model handled larger offsets well. At the end of the day, GM200 chips tend to have similar limits, regardless of which card they’re on.

One reason they’re so similar is that Nvidia has limited the peak voltage. Each and every one of these cards will allow you to raise the default GPU voltage by 87 mV and no more. I simply took each card to its max when I tested. I’d need more voltage to push the GPU offsets any higher without stability problems.

As for memory speeds, well, I generally pushed from the default of 7Gbps to 8Gbps and stopped. Each of the cards was able to handle that speed, but things got dicey in a hurry when I pushed past that mark. Since a number of GDDR5 chips are involved on each board, what you can get on the hairy edge of memory speeds will likely vary even more from one board to the next.

When overclocked using the settings above, all of the 980 Ti cards limited their clock speeds in Kombustor for the same reason: hitting their power limits. The reference board briefly hit a temperature limit instead, until I raised its temperature tolerance by a few degrees. Then they were all power limited. You can raise the power limit a bit in Precision or similar utilities, and I maxed it out for each contestant. Most cards topped out at 110% of their default, but the MSI allowed a 120% setting. The Gigabyte went to 130%. Regardless, power was always the limiting factor for clock speeds in Kombustor.

I should say that it’s entirely possible that my simple formula of using the peak voltage with the peak clock offset might not be the most optimal config in every case. One could possibly extract more performance from a power-limited GPU by keeping its voltage lower. But finding that magic sweet spot involves serious tweaking and tedium, and I just didn’t have the extra hours to dedicate to that process for every card. At least next time, I’ll use something else generate the GPU load, so we can see more typical clock behaviors.

Of course, the true test is how these overclocked configurations perform in games.

Turning up the clock speeds and voltage/power limits for these 980 Ti cards adds another 10% or so to their performance. Asus and Gigabyte are once again at the top of the pack when overclocked.

There’s a reason the Asus and Gigabyte cards are able to achieve the highest stable boost + offset clock frequencies and thus the highest frame rates: they use more power than everything else in order to stay stable. I’m not sure whether they’re secretly throwing more voltage at the GPU or what, but the difference in power draw versus MSI and EVGA is about 18W at stock speeds and ~30-50W when overclocked.

Even when topping the performance charts, the Asus Strix is among the quietest cards while overclocked. That’s a nice combination of attributes. The Gigabyte card achieves similar performance, but it makes a more noise in its efforts to keep GPU temperatures lower.

 

Conclusions

These are $650+ graphics cards, and as you might expect, they are all very good products. There aren’t any poor choices in this group, and each one of them has its own particular virtues. Which one is right for you will depend on what you value. Let’s review briefly, and you’ll see what I mean.

We should start with the GeForce GTX 980 Ti reference design from Nvidia, which many of these brands still sell under their own names. Although each of the aftermarket cards improve on the original in specific ways, Nvidia’s board and cooler design set the standard for everyone else. The reference cooler is relatively noisy—but only relatively so—and it doesn’t spin down its blower to zero like the other coolers represented here. But it does direct all of the GPU’s heat out of the back of the PC case, something the others only do partially, if at all. The reference card also lacks a backplate to provide extra protection on the back of the board.

Nvidia chose to omit the backplate and use a blower for a specific reason: multi-GPU configs. The blower removes heat from the enclosure, and the backplate provides room for another card’s air intake.  If you’re planning on packing two or more—especially more—graphics cards into the same system for SLI, then I’d recommend giving the reference boards serious consideration.

For everyone else, EVGA’s GTX 980 Ti SC+ is a clear upgrade from the reference card for single-GPU use. The ACX 2.0+ cooler is every bit as compact as the reference design, yet the SC+ is among the quietest cards we tested. The EVGA card is also the only aftermarket contender that requires only a six-plus-eight-pin power input, so it should work with a broader range of power supply units than the other cards. In a way, EVGA has set its own baseline for GTX 980 Ti cards with the SC+. The firm also offers more expensive 980 Ti variants, but this one is an easy choice that gives you a little something more with minimal hassle.

Gigabyte’s GTX 980 Ti G1 Gaming takes a distinctly different approach. It’s one of the two fastest cards we tested overall. The Windforce cooler gives this card a long, lean profile; it’s not much taller than the reference cards, which will make it a good fit in enclosures without much Z height. This card is a little noisier than most thanks to its default fan profile, but the Windforce cooler is clearly very capable. The G1 Gaming runs cooler than anything else we tested, yet it’s still quieter than the reference cards. Also, I’m smitten with Gigabyte’s multi-colored lighting—not because I like to taste the rainbow, but because it can be made to match or complement the other LEDs in your build. The G1 Gaming also has that second DVI port, which could matter to some folks.

The quietest card of the bunch is easily the MSI GTX 980 Ti Gaming 6G. The MSI also manages the second-lowest temperatures under load, behind only the Gigabyte—and it maintains its quiet, cool profile when overclocked. In short, the Twin Frozr V cooler is excellent, part of a long tradition of MSI goodness. Despite relatively high default clock speeds, though, the Gaming 6G isn’t quite as fast in the benchmarks as the entrants from Asus and Gigabyte. Still, I have admit: this board might be my choice for my own personal system. The performance differences at stake here are minor, and I appreciate the Gaming 6G’s quiet running and flashy good looks.

Asus Strix GTX 980 Ti OC

July 2015

That said, the Asus Strix GTX 980 Ti would have to merit serious consideration. The Strix is the fastest card we tested, both at its stock speeds and when overclocked, and it’s second only to MSI in going easy on the decibel meter. Asus’ new DirectCU III cooler is deadly effective, but you’ll pay for it in case volume. The hulking Strix makes the other cards look tiny. Provided it will fit comfortably into your PC’s case, though, there’s very little downside. Judging by current listings, the Strix costs no more than the other cards, and Asus throws in a full year of XSplit Gamecaster in addition to Metal Gear Solid V. That’s the most generous software bundle on offer here.

At the end of the day, the Strix is the best all-around contender in this bunch, and I’m making it my pick for a rare, coveted TR Editor’s Choice award. Just be sure you measure your case before ordering one. If your system can’t swallow a 12″ long card that juts up 1.7″ above the PCIe slot covers, consider going with one of the other options instead. You really can’t go wrong with any of them.

Comments closed
    • The Truth
    • 4 years ago

    How could you guys possibly run an article like this and leave out the Zotac 980Ti Amp! Extreme Edition? I benchmarked the MSI and EVGA cards right alongside the Zotac 980Ti Extreme and neither card could hang. We’re talking 1350 on the core, 2,000MHz on the memory. Temps remained in the mid 50’s to low 60’s.

    Hashtag: gameover.

    • ronch
    • 4 years ago

    The last time I read a video card comparison article here on TR, Asus was given the Ed’s Choice Award. Now it’s Asus again?

    I have to say, Asus probably has the best engineers in Taiwan.

    • itachi
    • 4 years ago

    Nice love the idea of reviewing multiple cards, I knew G1 was the card to go though the legend is made, look at these temps.

    That said do you guys know if it’s true that with EVGA you can remove the cooler and still benefits from warranty ? if so I might be interested to get an EVGA instead for future watercooling :).

    • Dhruv990
    • 4 years ago

    No love for zotac? :p I want an amp extreme, so fast it is. And damn those prices, these cards are cheap in USA the Asus (reference model) sells for 957$ in my country, and that price is after discount!

    • Supra93
    • 4 years ago

    Great roundup but you forgot probably the best card released yet… The ZOTAC GeForce GTX 980 Ti AMP! Extreme

    I’d be interested in how it stands up against ASUS Strix!

    • Bensam123
    • 4 years ago

    This is going to seem weird, but after looking at numerous GPUs and after market coolers, I think it would be quite prudent to test the coolers themselves without the GPU. As each GPU is different and each GPU has a different set voltage for each chip, it becomes impossible to test the GPUs at say the same heat dissipation running at the same fan speed.

    Perhaps not even the same fan speed, but rather they should be measured for dissipation at a certain decibel level. I think that would probably be the absolute fairest way to decide which card has the ‘best’ cooler. Although you then run into the card itself having a crappy fan profile, but I’m sure that’s not as big of a issue as the cooler itself.

    So maybe using a heatplate to simulate a heatsource and then a fan level at a certain decibel rating.

    Also props should be given for manufacturers that use heatsinks for the VRMs and/or memory. Also should be noted if they make contact with the whole heatsink or they have their own. I assume the GPU heatsink is better at dissipating heat then little heatsinks underneath it that ‘leech’ air off of it.

      • Firestarter
      • 4 years ago

      You’re right, but the GPUs themselves aren’t heatplates themselves either. Testing with a heatplate might give a good idea of the cooler’s relative performance when given the perfect workload, but GPUs (and CPUs for that matter) are not a perfect workload, parts of them are hotter than others. Testing all coolers with a heatplate would ignore the way that the base of the cooler interacts with the GPU, which given the impact that poor thermal paste application can have is a rather important part of the thermal equation.

      It might be most fair to compare all coolers strapped to a single CPU sample, but that’s impossible and still not representative of the performance you get with the custom board designs. So to actually give a really good and fair idea of how the cards stack up, you’d have to take a good random sample of each model (say about 5 to 10), test them all and make a nice average out of the results without outliers. That would be a prohibitive amount of testing of course.

      Anyway, I agree with you that the heatplate method of testing would be very useful and enlightening, but it cannot be used instead of direct measurements of the cards as a unit.

        • Bensam123
        • 4 years ago

        I disagree. As long as it makes contact over the same area as a GPU die, it wouldn’t matter. Saying ‘all chips are different, therefore not comparable’ when we normalize for the variable that throws them off isn’t applicable. It doesn’t even make sense. The whole reason you normalize variables is so you have a baseline to compare them to, just the same way you don’t test video cards on different resolutions because they operate differently on different resolutions.

        Workload wouldn’t change anything. Maybe a small section of the chip would get hotter then another part, but the for all intents and purpose, the variance caused by that and it happening to be in a less ‘conducive’ portion of the heatsink where it vastly changes the way the heatsink cools would change next to nothing. Especially when you put this in light of testing against different fan speeds and different voltage chips operating at different max frequencies.

        I mean look at the bottom of those heatsinks. They’re giant platters. As long as the heat source is remotely the same size of the chip it’s not going to matter one bit. You could use a copper block for that too to transfer heat.

        Testing random samples from the same boards STILL is not testing the heatsinks compared to each other, you’re still testing the heatsink AND the cards. Some manufacturers choose worse binned chips because they’re cheaper and that’s not testing the heatsink itself and would not be comparable to other cards heatsinks, which is probably the biggest selling point of these cards. Which has the better heatsink? That determines which cards will be the quietest, has the most thermal headroom, and will more then likely OC the best across multiple samples.

        I’m not saying throw out other card testing either, I’m just saying there should be a more uniform method of testing the heatsinks themselves with the other variables removed and normalized for noise.

        As far as thermal paste mattering? It really doesn’t. They talk about that every time the OC competitions come up on TR. You could ask them to test the thermal paste between the heatsinks too if you really want to, but I don’t think that’s the biggest deciding factor here.

          • Chrispy_
          • 4 years ago

          I think you’re onto something but I don’t think it matters that much.

          The real value of the information gained from a TR test is not really down to the specific quality of the sample – as you say there are differences in ASIC quality, some chips will run cool, some chips hot, some will overclock better etc. This variance cannot really be tested without getting many of each card and averaging the results.

          What’s more important, both to the end-user and in making a meaningful difference to product differentiation are things that aren’t specific to the GPU – so the voltage the OEM chooses to set, the boost limits, the fan profiles, the number of VRM phases etc. If you took each of these 980Ti boards and replaced their OEM coolers with a different, standard benchmark cooler, they’d still all perform differently – producing different levels of power draw, heat, noise, performance etc.

          I feel that whilst the quality of the cooler is a contributing factor to the end result, it is not the only one – and so testing the cooler is nowhere near a complete picture. This is bvbiously even more important for those folks that will be scrapping the included cooler and bolting a waterblock into place.

      • Amgal
      • 4 years ago

      Their site is sort of an eyesore and somewhat difficult to navigate at times, but [url<]http://www.frostytech.com[/url<] does what you describe. All heatsink reviews with a hotplate.

      • Nevermind
      • 4 years ago

      I think it’s better to just evaluate the ‘complete’ product as-sold.

      • Freon
      • 4 years ago

      Interesting idea, and as you said every chip is different. But I wonder if the balance of what you’re trading off is worse in terms of separating yourself from the real world case.

      There’s even more possibly variance on card to card with things like how well it’s mounted, the TIM, and soforth. Just removing and reinstalling a heatsink can change performance.

      The real fix is to buy a handful of each card and see how much they vary. Then that’s still not perfect. You may want to start tracking performance month over month, production run to production run, and report that to consumers and readers…

      The rabbit hole here goes very deep.

    • ultima_trev
    • 4 years ago

    Dat 3 fan Gigabyte Windforce cooler.

    EVGA seems to be slacking compared to the other board partners nowadays. Even though it’s a different product, the only EVGA version of the vanilla GTX 980 that comes with a backplate is the K1NGP1N version, which is so expensive you might as well buy a GTX 980 Ti. Pretty stingy of them to charge extra for a backplate, these high end cards and their heatsinks are huge, the risk of sagging is too damn high for them not to include a backplate.

    I’ve decided to wait until 14nm for a worthy HD 7850 replacement, but if I miraculously came across the means (funds) to build a new rig, I wouldn’t say no to that G1 Gaming GTX 980 Ti (or even a G1 Gaming vanilla GTX 980). Shame they don’t do the triple fan version for the R9 390 cards.

    • trek205
    • 4 years ago

    um what the heck is up with the way you are reporting boost on the oc section? none of that makes any sense as you are not listing the actual real world in game boost on any of those cards when oced. even out of the box most of those cards boost higher than what you are showing for “boost + offset.”

      • Damage
      • 4 years ago

      Boost + offset is a combination of the card’s default boost clock plus the highest stable clock offset. It’s not an in-game value, just a tweaking variable.

      I also report the actual GPU clock speed when overclocked in Kombustor, but like I say in the article, Kombustor/FurMark is too intensive to allow the sort of clocks you’d see in games. I didn’t expect the gap to be that big, but it is with GM200. Next time, I will use a different workload to stress the cards when overclocking.

    • gigafinger
    • 4 years ago

    Typo on Page 5 – “The Asus and MSI offerings aren’t far behind.”

    I believe you meant Asus and EVGA.

    Great article as always!

      • Damage
      • 4 years ago

      Doh! Fixed.

    • Meadows
    • 4 years ago

    So quiet lately without leetgamer.

      • chuckula
      • 4 years ago

      Meadows pours one out for his banned homie.

        • Meadows
        • 4 years ago

        [url<]https://www.youtube.com/watch?v=E-dW3InkBeM[/url<]

      • bjm
      • 4 years ago

      Why does TR only do round-ups of nVidia cards?! This is ridiculous, obviously they are paid to show the improvements manufacturers make over reference designs, but they refuse to do the same for AMD GPUs. They will only review and expand on the faults from–What? There’s an Asus Strix Fury review? Uhhmm… Windows 10 is the start of Linux on the desktop!

      Love your friend,
      l33t-g4m3r

        • Meadows
        • 4 years ago

        Oh, you sly one.

    • PerfectCr
    • 4 years ago

    I picked up the EVGA 980 Ti seen in this review (ACX 2.0+, Superclocked, with Back Plate). The card overclocks really well. I settled on +110 on the core, which in most games boosts to an even 1400MHz on the Core. I raised the memory just a bit as well (+100MHz in Precision X). I raised the Power to 110% and Temp to 91C (linked) to allow for maximum headroom. I’ve not seen the core go higher than 78C even after hours of gaming, and fan hovers around 1600RPM. The fan is audible at this level, but nothing I would call remotely close to loud. It’s more of a pleasant low whoosh sound.

    The card is quiet, has great lighting on the side (I am sucker for seeing the model name in lights!). Of course, EVGA is also known for having great support, so keep that in mind also.

      • DancinJack
      • 4 years ago

      I have the same card, and almost exactly! the same OC/clock settings. Sweet!

        • PerfectCr
        • 4 years ago

        Excellent, it’s a great overclocker to be sure!

      • thecoldanddarkone
      • 4 years ago

      I was a little to close to my monitor when I opened that picture.

        • PerfectCr
        • 4 years ago

        Ha sorry about that. 🙂

      • bhd2786
      • 4 years ago

      I also have the EVGA Card mine had a boost clock of 1290 MHz out of the box .. I have not changed it and it just chews through anything I throw at it

    • DataHound
    • 4 years ago

    Wow. I’m currently running 2x MSI GTX 780Ti GAMING in SLI and this is the first set of numbers that makes me think about jumping ship. The Asus card looks amazing.

    Has anyone else gone from 2x 780Ti -> 1x 980Ti? Was the loss in overall FPS worth the single card benefits?

      • Freon
      • 4 years ago

      There’s plenty of luck of the draw at play when you start looking at the overclocking numbers. Even at “stock speeds” there’s a lot going on per card.

      I’ll have to try that Kombuster and see how much my two “identical” Asus Strixx 970’s are. I imagine there’s a fair gap between the two. GPUz shows a different ASIC quality by like 15 or 20%, whatever that means.

      • Westbrook348
      • 4 years ago

      “Overall FPS” is fairly irrelevant. Your min FPS i.e. 99th percentile frame times will be better with the 980Ti. 980Ti beats SLI970 and CF 290X and comes close to SLI980 when you’re looking at minimum frame rates. So to answer your question: yes it’s perfectly fine to lose a little avg FPS in order to improve your worst frame times. Plus you’ll save in electricity, heat, noise, and driver compatibility.

    • DPete27
    • 4 years ago

    So it’s clear that front-back fins (Asus/MSI) requires vertical heatpipes and makes the GPU excessively tall while vertical fins (Gigabyte/EVGA) requires front-back heatpipes which allows the heatsink to stay roughly within the limits of the expansion bracket. Obviously, all that extra height creates problems for fitting into some cases, making it less desirable.

    My question is: inside a case, do the front-back fins REALLY facilitate heat transfer out the expansion bracket vents, or is all the heat just being dumped into the case regardless, making the fin orientation insignificant.

    • Anovoca
    • 4 years ago

    Wow, Scott you finally finished the review.

    Time to sell my Starbucks stock.

    • USAFTW
    • 4 years ago

    Have to say I’m impressed with GM200 and what they’ve been able to do with 28nm and such a large engine. And they didn’t have to go out of their way and make a new DRAM tech to make it happen.
    Next year is gonna be awesome, 16/14 nm finfet GPUs and Zen (if AMD delivers on their schedule and expectations, finger crossed).
    Edit: It seems the reference card is still the one to go for me.

      • ImSpartacus
      • 4 years ago

      Nvidia has been doing great lately. I’m interested to see if they can keep it up.

        • Leader952
        • 4 years ago

        Nvidia is spending very large amounts of cash for R & D so I do think they can keep it up.

        [quote<]Interestingly Nvidia has also managed to claim a spot in the top 10 list, albeit at 10th place. Interestingly the bottom few semiconductor companies all have approximately the same budget, while Intel has more budget than nearly all of them combined. What is perhaps of more interest to our readers is that Nvidia spends approximately 1/3 of its total revenue on semiconductor R&D, needless to say, that is a huge amount. [url<]http://wccftech.com/nvidia-spends-revenue[/url<] [/quote<] [quote<]NVIDIA Research and Development Expense (Quarterly): 339.00M for April 30, 2015 [url<]https://ycharts.com/companies/NVDA/r_and_d_expense[/url<] [/quote<]

Pin It on Pinterest

Share This