An update on AMD 790FX motherboards

We weren’t particularly impressed with AMD’s 790FX chipset when we took our first look at it last month. The Gigabyte GA-MA790FX-DQ6 motherboard we tested had numerous issues, including BIOS quirks and relatively high power consumption. The chipset itself appeared flawed, as well, exhibiting slow USB and PCI performance and compatibility problems between Windows Vista and Serial ATA drives running in AHCI mode.

To gather a broader impression of the 790FX, we’ve rounded up new motherboards based on the chipset from Asus and MSI to see how they compare with Gigabyte’s initial offering. We’ve also dug deeper into the SB600’s support for AHCI in Vista, discovering some troubling performance characteristics along the way. Keep reading for a quick look at the current 790FX motherboard landscape.

New boards from Asus and MSI

With a street price hovering around $265, Gigabyte’s GA-MA790FX-DQ6 comes in at the very high end of the 790FX motherboard spectrum. There, it’s joined by Asus’ M3A32-MVP Deluxe WiFi-AP Edition, which can be had for about $35 less. At the other end of the map we find MSI’s K9A2 Platinum selling for as little as $156—clearly the bargain of the three. Time constraints prevent us from giving these boards a full work-up. We’ve run them through our full chipset test suite, though, and have a complete set of results to report.

Asus’ M3A32-MVP Deluxe WiFi-AP Edition

These boards are pretty much what you’d expect from their respective manufacturers. The Asus is packed to the gills with onboard peripherals, including an 802.11g Wi-Fi module. Asus manages to squeeze four PCIe x16 slots onto the board while ensuring that monster graphics cards won’t obscure access to onboard ports, as well.

Asus bundles an intricate memory cooler in the box

Adding to the recent trend toward elaborate chipset coolers, Asus bundles a memory cooling module with the board that links memory modules to the chipset heatpipe network. This adjustable memory heatsink is designed to be used in lieu of traditional DIMM heat spreaders, but it only works with two modules at a time.

MSI’s K9A2 Platinum motherboard

As one might expect, the MSI board is quite a bit more subdued, lacking Wi-Fi and fancy memory cooling. MSI still manages to squeeze in four PCIe x16 slots, though, and they’re spaced to allow for four double-wide graphics cards. However, longer double-wide cards installed in the second PCIe x16 slot will obscure acess to some, if not all, of the board’s SATA ports. Note that the K9A2 Platinum locates the north bridge chip slightly to the left of the CPU socket rather than directly below it.

Specifications compared

To allow for quick comparison, we’ve summarized the specifications of the M3A32-MVP Deluxe and K9A2 Platinum below.


Asus M3A32-MVP
Deluxe WiFi-AP Edition

MSI K9A2 Platinum

CPU support
Socket AM2/AM2+
Phenom, Athlon X2 processors
Socket AM2/AM2+
Phenom, Athlon X2 processors

North bridge
AMD 790FX AMD 790FX

South bridge
AMD SB600 AMD SB600

Interconnect
PCIe x4 (2GB/s) PCIe x4 (2GB/s)

Expansion slots
4 PCI Express x16

2 32-bit/33MHz PCI

4 PCI Express x16

1 PCI Express x1

2 32-bit/33MHz PCI


Memory
4 240-pin DIMM
sockets

Maximum of 8GB of DDR2-667/800/1066 SDRAM

4 240-pin DIMM
sockets

Maximum of 8GB of DDR2-667/800/1066 SDRAM


Storage I/O
Floppy disk

1 channel ATA/133

4 channels Serial ATA with RAID 0, 1, 10 support

2 channels Serial ATA with RAID 0, 1 support via Marvell 6121

Floppy disk

1 channel ATA/133

4 channels Serial ATA with RAID 0, 1, 10 support

2 channels Serial ATA with RAID 0, 1 support via Promise T3

Audio 8-channel HD audio
via SB600 and Analog Devices AD1988B codec
8-channel HD audio
via SB600 and Realtek ALC888 codec
Ports 1 PS/2 keyboard

6 USB
2.0 with headers for 4 more

1 RJ45 10/100/1000 via Marvell 88E8056

1 802.11g Wi-Fi via AzureWave AW-NA830

1 1394a Firewire via Agere L-FW3227 with header for 1 more

1 eSATA via Marvell 6121


1 analog front out

1 analog bass/center out

1 analog rear out

1 analog surround out

1 analog line in

1 analog mic in

1 digital coaxial S/PDIF output

1 digital TOS-Link S/PDIF output

1 PS/2 keyboard

1 PS/2 mouse

4 USB
2.0 with headers for 6 more

1 RJ45 10/100/1000 via Realtek RTL8111B
1 1394a Firewire via
VIA VT6308P with header for 1 more

2 eSATA via Promise T3


1 analog front out

1 analog bass/center out

1 analog rear out

1 analog surround out

1 analog line in

1 analog mic in

1 digital TOS-Link S/PDIF output

Given the price difference between the boards, the MSI’s specs certainly look promising. About the only integrated peripheral it lacks is Wi-Fi, and it even manages one more expansion slot than the M3A32-MVP. The boards use different chips to provide auxiliary storage, Firewire, and Gigabit Ethernet, though. As you’ll see in a moment, there are clear differences in peripheral performance between them.

Neither the Asus nor the MSI board exhibited any BIOS irregularities during testing. The M3A32-MVP does provide a slightly more extensive array of overclocking options, though.


Asus M3A32-MVP
Deluxe WiFi-AP Edition

MSI K9A2 Platinum

Bus speeds
HT: 200-600MHz in
1MHz increments

DRAM: 400, 533, 667, 800MHz

PCIe: 100-150MHz in 1MHz increments
HT link: 200, 400, 600, 800, 1000MHz

HT: 200-600MHz in
1MHz increments

PCIe: 100-150MHz in 1MHz increments
HT link: 200, 400, 600, 800, 1000MHz


Bus multipliers
CPU: 5x-13x in 1x increments (Athlon X2 5200+) CPU: 5x-13x in
0.5x increments (Athlon X2 5200+)
HT link: 1x-13x in 1x increments

Bus dividers
NA HT:DRAM: 1:1,
1:1.22, 1:1.66, 1:2
Voltages CPU: 0.8-1.6875V in
0.0125V increments

DRAM: 1.8-2.5V in 0.02V increments

PCIe: 1.1-1.4V in 0.02V increments

NB PCIe: 1.8-2.1V in 0.1V increments

SB: 1.2-1.4V in 0.02V increments

HT: 1.2-1.5V in 0.1V increments
CPU NB: 0.8-1.6875 in 0.0125V increments

CPU: 1.318-1.587V in
0.038V increments

DRAM: 1.8-3.1V in 0.05V increments

NB: 1.125-1.525V in 0.025V increments
HT: 1.125-1.525V in 0.025V increments


Monitoring
Voltage, fan
status, and temperature monitoring
Voltage, fan
status, and temperature monitoring

Fan speed control
CPU CPU

The Asus BIOS’s advantage comes in the voltage department, where it provides access to a much greater range of CPU overvolting options. It’s also possible to tweak a greater number of M3A32-MVP voltage variables. However, the K9A2 is better equipped for memory overclocking thanks to its higher memory voltage ceiling. MSI’s processor fan speed control is a little better than what you get with the Asus board, too, thanks to the ability to set a temperature target and fan speed percentage arbitrarily.

Benchmark results

We used an identical test configuration to that of our 790FX chipset review. The latest BIOS revisions were used for the Asus and MSI boards—revision 0603 and 1.1, respectively. As you’ll see, performance doesn’t vary much from board to board until you get into peripheral testing. We’ll save our commentary for the conclusions.

Memory performance

STARS Euler3d computational fluid dynamics

Worldbench

Gaming

Serial ATA performance

USB performance

Firewire performance

Ethernet performance

PCI Express performance

PCI performance

Power consumption

Overclocking

340MHz on the M3A32-MVP

In testing, we were able to get the M3A32-MVP Deluxe stable with a 340MHz HT clock before the system refused to post.

330MHz on the K9A2 Platinum

The K9A2 Platinum fared nearly as well, sailing up to a 330MHz HT clock before flaking out. In both cases, the boards achieved their highest HT clocks with stock voltages. Additional juice didn’t help them scale any higher.

AHCI issues

A documented compatibility problem between Windows Vista and the 790FX chipset’s SB600 south bridge component currently complicates running drives in AHCI mode. As Microsoft notes:

When you try to install Windows Vista on a portable computer that uses an ATI SB600 Series chipset, each driver may take a long time to install. Therefore, the Windows Vista installation process may take several hours.

That’s if it works at all. One can install Vista by putting the SATA controller into IDE mode, but AHCI mode is necessary to take advantage of Native Command Queuing and SATA device hot-swapping, so it’s not a trivial capability to lose.

With a hotfix from Microsoft, we were able to get drives running in AHCI mode on both boards. However, doing so quickly all but requires an auxiliary storage controller. Vista must be installed with the south bridge running in IDE mode. You then have to move the hard drive to a secondary controller, switch the south bridge to AHCI mode, and apply the hotfix and necessary drivers before swapping the drive back. This problem should be resolved by Vista Service Pack 1, but without slipstreaming, that won’t necessarily make the installation process any easier.

The 790FX chipset’s AHCI problems don’t end with installation, though. There are also some performance oddities that we can illustrate nicely with IOMeter. In the results below, the 790FX is shown with the SB600 running in IDE mode, in AHCI mode with Windows Vista’s drivers, and in AHCI mode with 2.5.1540.47 drivers provided by AMD. Incidentally, those drivers come from Silicon Image, which is apparently responsible for the SB600’s Serial ATA controller.

With AMD’s official drivers, performance doesn’t scale at all as the load increases—even IDE mode offers higher transaction rates. Windows Vista’s native drivers seem to have it right; SB600 performance ramps more aggressively than even the nForce 590 SLI.

AMD’s official drivers also deliver much slower response times, suggesting that Vista’s native drivers may be the way to go.

Except for one little problem: CPU utilization. Vista’s native drivers may offer the AHCI performance we’re used to seeing, but they exact a huge CPU utilization penalty in the process. Silicon Image’s AMD’s drivers offer competitive CPU utilization, at least, but we’ve already seen that their performance is lacking.

At the moment, running the SB600 in AHCI mode requires choosing between poor performance or high CPU utilization. Neither is acceptable in our opinion, especially since competing chipsets seem to have no problems getting it right. Let’s hope AMD can address this issue with its upcoming SB700 south bridge.

Conclusions

After running the M3A32-MVP Deluxe and K9A2 Platinum through our gauntlet of chipset and motherboard tests, it appears our initial prognosis for the 790FX chipset was largely accurate. The Asus and MSI boards we’ve looked at today aren’t any faster in application tests than Gigabyte’s GA-MA790FX-DQ6, which puts them right in line with competing nForce 590 SLI boards. However, the SB600’s AHCI issues and slower USB performance remain. The SB600 appears to have some PCI performance quirks, as well, with the Gigabyte and MSI boards delivering much lower PCI throughput than our nForce 590 SLI platform. The Asus board fared a little better in our PCI throughput test.

Indeed, the M3A32-MVP had a performance edge over the K9A2 Platinum through much of our peripheral suite. The Asus board isn’t particularly exemplary on that front, but the MSI board suffers from a Gigabit Ethernet chip with atrocious CPU utilization and a Firewire controller that’s a little slow. To its credit, the K9A2 does deliver lower power consumption than the other 790FX boards. However, it only draws about 12 fewer watts than the nForce 590 SLI, which is a relatively modest advantage.

If you absolutely must buy a 790FX motherboard today, the Asus M3A32-MVP Deluxe and MSI K9A2 Platinum are better alternatives to the Gigabyte GA-MA790FX-DQ6. For those looking for a high-end board packed with fast peripherals, the Asus is the best option. Those looking to save a few dollars are much better off with the K9A2 Platinum.

That said, we still recommend waiting for the second wave of 790FX boards to hit with updated SB700 south bridge chips. By then, AMD should have B3-stepping Phenom chips available that correct the TLB erratum, as well.

Comments closed
    • sigher
    • 12 years ago

    I wish somebody sued or something and forced manufacturers to either not tie NIC’s to old/slow PCI or force them to not call it gigabit, after all at the very best when ONLY using the NIC on the PCI lanes you have 1/3rd of a Gb and not 1Gb, and if you use a PCI soundcard or something simultaneously then it sinks even more.

    Perhaps more modern chipsets would in general run much smoother and better if they didn’t have to deal with trying to handle old PCI.
    I wonder what would happen if TR disabled all PCI and related stuff in the BIOS and ran some test.

      • evermore
      • 12 years ago

      Gigabit = 1000Mbits per second (or 2000 in full-duplex)

      32-bit, 33MHz PCI = 1064Mbits per second total

      So technically PCI is completely fast enough to handle Gigabit speeds, even if only half-duplex. The manufacturers aren’t responsible if you stick other devices on the bus that use some of the bandwidth. Audio isn’t enough to use much bandwidth, you’d need to be running a drive controller or USB/Firewire card or similar to use enough to make a difference.

      Variations between chipsets can also be quite large, regardless of the bus they’re on.

      PCI does perform more poorly than PCIe (or Intel’s CSA) or directly integrated LAN controllers, but it’s not because of pure bandwidth limits. However it’s still far far faster than Fast Ethernet, and a good upgrade for the average consumer. As PCIe becomes more commonly used, PCI network chips will get cheaper for a while and help make for lower cost boards. People who are really concerned about getting full Gigabit throughput ought to be paying attention to the way it’s implemented on the products they purchase.

      Eliminating PCI entirely probably wouldn’t make much difference to overall performance of the rest of the system. It’s not DOING anything that other components don’t. You can’t disable the PCI controller entirely through the BIOS, although you might be able to make the board not send the clock signal to unoccupied slots.

      What I’d prefer is that chipset-makers integrate Gigabit LAN into all chipsets, so board makers don’t need to use PCI or PCIe controllers, they just need a PHY chip for the physical interface. SiS integrates Gigabit, VIA integrates 10/100, Intel does GigE in ICH9 but had none at all in ICH6/7/8, Nvidia’s used native Gigabit entirely since nForce5 (maybe nF4) except nF520, ATI/AMD hasn’t had LAN integrated ever.

    • pluscard
    • 12 years ago

    So, I guess the conclusion is the chipset is great, but it limited by the dated southbridge.

    But where does that leave us as compared to nVidia’s SLI chipset.

    Which is the better choice, today?

      • Flying Fox
      • 12 years ago

      Chipset = North + South bridges. And you look at it as a whole.

      So you should probably say “the Northbridge is great”. However, a platform is only at its best with its weakest link, and the SB600 doesn’t look so hot at the moment. 🙁

    • eitje
    • 12 years ago

    something to try: vlite (http://www.vlite.net/) can presumably let you roll your hotfix(es) directly into the installation.

      • evermore
      • 12 years ago

      Is the Vista version of nlite §[<http://www.nliteos.com/,<]§ which works pretty well. Personally I'm itching for the time to wipe Vista from my new laptop and put XP on it.

    • evermore
    • 12 years ago

    I have an idea. Buy a mainboard with some x1 PCIe slots, and an x4 slot, in addition to the x16, and use those to bypass the southbridge for everything important. Stick a PCIe SATA card in the x4 slot (wouldn’t want to limit it with x1), a USB/Firewire card in an x1, maybe a sound card, make sure the board has PCIe Gigabit LAN. All set. Of course, you’d have to find a board priced as if there were no southbridge at all, and none of those PCIe cards are particularly inexpensive. At least the system would perform better.

      • eitje
      • 12 years ago

      q[

    • leor
    • 12 years ago

    lots of n00bs to the site i guess, TR bashed the crap out of intel during the netburst era, which i thought was cool cause netburst was a flaming turd.

    any perceived anti AMD bias might be because intel currently has a better product?

    • Rza79
    • 12 years ago

    /[

      • Dissonance
      • 12 years ago

      With the exception of IOMeter, all 790FX testing was conducted with the south bridge in IDE mode.

        • Rza79
        • 12 years ago

        2 weeks ago i made a system for a client with a GA-MA790FX-DQ6.
        He had 3 Raptor HDD’s. 1 74GB GD and 2 150GB ADFD’s.
        When all HDD’s were connected to the SB600, you would get data corruption. I replicated the setup on a Asus M2R32-MVP and the same problem persisted. In the end I connected 2 HDD’s to the JMicron controller and only 1 to the SB600. In this configuration I had no data corruption.
        So the SB600 has bad signal and when many HDD’s are connected, it becomes apparent. Or it’s not compatible with Raptors (or one of them) but I didn’t have the time anymore to try with an other HDD.
        Next week I have to make a system with an Asus M2R32-MVP and that one has normal HDD’s. I’ll see if i have the time to test the write performance.
        Did you only test with the Raptor or also with other HDD’s?

      • Damage
      • 12 years ago

      l[https://techreport.com/articles.x/13628<]§ This was a follow-up article intended to see whether the problems we encountered initially with the Gigabyte board and a pre-release version of the Asus persisted.

        • Rza79
        • 12 years ago

        Yeah my bad. Forgot about the ‘update’ part.

      • ssidbroadcast
      • 12 years ago

      q[

      • evermore
      • 12 years ago

      10% is a significant performance difference. Enough that we have a word for it, decimate, because it is so heavy a toll. Of course as with executing every tenth soldier, it’s a dramatic difference but not enough to cripple the performance entirely. With soldiers, it’s made up for by the survivors being WAY more inclined to work hard; with technology, it might be made up for by higher performance in other areas, or lower cost, but that’s not the case with the SB600.

    • pluscard
    • 12 years ago

    Quote from Digitimes:
    “Although the Radeon HD 3800 series was launched three weeks later than the GeForce 8800 GT, Radeon 3800 demand has started to pick up, bringing the market shares of Nvidia and AMD from 90% and 10%, originally, to 70% and 30%.”

    My point is we need to think critically, and even strategically, when reading product reviews. AMD is making real headway with the affordable 3800 cards. While the Phenom might not be ready for prime time, both the Radeon 3800 and 790FX chipsets are clearly winners.

    A review that concludes “…right in line with competing nForce 590 SLI boards.” while suggesting the 790 has “issues” is clearly steering people to the nvidia direction, which would likely mean away from the Radeon as well.

    While TechReport writers may not be quite this devious, I can assure you Intel marketing certainly is, and they’re likely planting sentiments with many well read review sites.

      • Dissonance
      • 12 years ago

      We haven’t “suggested” that the 790FX has issues, we’ve clearly illustrated exactly what those issues are.

      • evermore
      • 12 years ago

      790 might be a winner (though clearly not the be-all end-all of performance), but the platform as available today is not. You can’t buy a mainboard with only a northbridge, so you must evaluate the whole package, and the SB600 is crippling.

      Even if TR were biased, so are you, so what’s your point?

      • Rza79
      • 12 years ago

      Obviously what your pointing out is that *[

      • green
      • 12 years ago

      if you go looking for bias then you’re obviously going to find it

      you quoted this:
      §[<http://www.digitimes.com/mobos/a20071228PD207.html<]§ while i applaud amd for bringing out a ~$200 lower power graphics solution on a 55nm process putting them a step ahead of nvidia, it doesn't magically fix the issues with current 790fx boards (or phenom cpus) and quoting that article does nothing but show that all you're really here to do is promote amd in an attempt to persuade us to support them for the sake of competition you say we need to think critically and strategically. yet when the review is critical of poor sata performance you're up in arms saying it is biased. while the 790fx chipset may be ready from prime-time, all implementations thus far ARE NOT. this is made clear in the article that it is the sb600's fault, not the 790fx after reading the review and looking at the benchmark numbers, why would i bother getting a current 790fx chipset when i can do just as well with a 590? for something that is claimed to be for an "enthusiast platform" the performance doesn't really stand out from the competition the "issues" encountered with the sata are a very big deal it's bad enough that hdd's are the slowest part of the virtual memory sub-system but the sb600 essentially cripples the board with a very slow write speed resulting in slow paging and hence slow performance it's like removing the speed limits off highways then forcing people to drive cars with 2cyclinder-2stroke engines and they want to claim this as part of an "enthusiast platform"? then there's 4-way crossfire. it is NOT something for the enthusiasts. i would rather shoot myself than get 4 video cards which in the end would only be used to enhance games performance having 4-way crossfire does not make this an enthusiast platform. it makes it an e-peen platform so why in the world would i buy a CURRENT 790fx boards when i know there's an "issue" that WILL affect my computer's performance, that an sb700 chipset is coming that will have better IO performance, and that to take advantage of this board i would need to buy a phenom that, irrespective of whether or not i would actually encounter the issue, amd has said all mb's must apply the patch which results in slightly* worse performance? there are reasons why many of us are NOT early adopters of new products. amd's latest line-up is a shining example of why to avoid it. if you feel you need to help amd then by all means go ahead and buy amd. but you have the gall to claim that intel marketing is responsible for what you perceive as bashing of amd in this review? i would say you are either a deep-rooted amd supporter, or have a personal (aka financial) stake in amd amd has had a terrible year and there's no point denying the fact amd will do significantly better next year across all their products but that still doesn't make up for the fact the current 790fx boards poor implementations * i say slightly as the largest difference was in firefox2 which is slow in the first place (compared to earlier versions)

        • pluscard
        • 12 years ago

        That’s a pretty long winded first post to TR.

        Posting from Santa Clara, I suppose?

          • evermore
          • 12 years ago

          The obvious response to that is that you’re apparently posting from Sunnyvale.

          • green
          • 12 years ago

          it’s funny that
          i actually had the word “fanboy” ready to go in my original post instead of “deep rooted amd supporter”
          i decided to change it to give the benefit of the doubt. should have gone with my first instinct it seems

          the bottom line is the motherboard is fine for the run-of-the-mill consumer
          most mb’s out there are because they don’t need all the bells and whistles
          but this is part of the “spider” platform which is for enthusiasts
          and i doubt enthusiasts will happy with the sata performance

          and it’s not like everything is lost. thinking about it, all it really needs is a driver update
          the “issue” really boils down to the amd supplied driver capping the speed
          alternatively you could go with the vista driver but it comes at the cost of (relatively) high cpu usage
          all it needs is a driver that gives the vista driver speeds with the amd driver cpu usage
          but until that comes out i doubt any ‘enthusiast’ would appreciate having to choose one or the other

          p.s. no. it wasn’t my first post.

    • evermore
    • 12 years ago

    Because of price/performance/value for my needs, I haven’t had an Intel system since my first PC, a 486 in 1995. But I just changed my X2 4200+/nforce570 (with a slight overclock to 2.4GHz, the best I could get without spending more time tweaking, and 2.5GHz was able to run but not stable) to a Core 2 Duo E6420/P35, which I’ve got overclocked to 2.8GHz with minor tweaking (3.2GHz was possible with no tweaks, and ran Windows, but didn’t always boot).

    The CPU cost $167, the P5K Deluxe WiFi-AP mainboard was $200. CompUSA’s going out of business sales are really nice for some items.

      • Shinare
      • 12 years ago

      Are you kidding? I went to CrapUSA last night to give them one last final horah and I was floored by the prices that they were STILL asking after their going out of business discounts were applied. If they are giving up because their prices were too high to compete, you think they would have learned from that lesson when trying to clear out their inventory. I almost ordered a 4GB microSDHC for my blackberry yesterday from newegg but thought I would check CompUSA’s going out of business “sale” first. If I could get within $10 of the newegg price of $39 shipped I would have paid the extra plus tax just to have it last night. When I got there and asked how much the 4GB’s cost they told me $89.99!!! After my guffaw I walked around and saw a $149 1GB memory kit (2x512MB), a $140 500W power supply, a CrapUSA branded copper HSF for $30, all after the supposed %10-30 discount. Unbelievable. I went home and immediatly ordered from Newegg. I like to support my local businesses and all, but not for double to quadruple the prices of online.

      Crash and burn CompUSA. You certainly deserve it.

        • evermore
        • 12 years ago

        Like I said, SOME items. Plus they may be different depending on which store you go to. This is in Nashua, NH. 20% off tech parts brings many mainboards and CPUs well below Newegg’s prices, but laptops are only 5% off their outrageous pricing so that’s not much value. Video cards and memory as well are only 10% off which makes them continue to be highly overpriced.

        I certainly would never have called CompUSA’s prices reasonable on most items.

        • bfellow
        • 12 years ago

        I went to COMPUSA and liquidation sale of 10-15% off items isn’t “liquidation” at all!

          • evermore
          • 12 years ago

          Well, you might piss yourself when you see the prices…

    • Kurotetsu
    • 12 years ago

    And pluscard’s posts just continue to highlight an extremely annoying trend I’ve been seeing on hardware sites:

    Any review site that doesn’t automagically declare any crap that AMD puts out to be the most powerful hardware in the universe MUST have a pro-Intel/anti-AMD agenda. The only hardware reviews they consider ‘legitimate’ are ones that laud AMD non-stop, from start to finish, regardless of the actual quality and performance of the product.

    It really is disgusting to see. My first two system builds were AMD, and they were rock solid and awesome. Yet for some bizarre reason I haven’t developed this utterly irrational wank-habit for their brand. I’m not sure why…

      • RambodasCordas
      • 12 years ago

      Well he has great points. In fact I completely agree with him.

      In fact the Spider (or 790FX) is an excellent chipset with goodies not addressed yet by Intel or Nvidia.

      I can’t understand why sites like this praise the P35 over P965 or P965 over P945 where the only thing Intel did was address the Core 2 compatibility and a minor FSB speed bump.
      AMD brings a new Bus connection (HT3), new PCIe connection (PCIe 2.0), 4 Gpus capability on a single chip and gets bashed.
      Intel/Nvidia just rebrands the same product and its “god” creation. That is the part that I really don’t understand.

    • WaltC
    • 12 years ago

    I don’t think TR has an anti-AMD bias at all.

    Like in any other publication, however, different writers have differing opinions on various subjects–some of those opinions are well-formed and some are not–and sometimes you see these imperfections coming through. But that only makes TR exactly like every other publication you might care to read….;)

    Speaking of opinions and how they slant coverage of various issues, the “TLB erratum” stories were handled in a level-headed way by some sites and exaggerated to to an extreme by other sites. Personally, I think the only site to date that I’ve read that placed that particular subject in the context it deserved was [H] –I think Kyle framed and explained the issue exactly as it should have been presented. But let’s not forget that when Intel had some P4 erratum a few years ago, sites like AnandTech went way overboard with it at the time to the extent of so badly distorting the issue that it could scarcely be understood in the context in which it actually existed. OTOH, the Core 2 erratum which both Intel and Microsoft to date have publicized and documented has been almost completely ignored.

    The problem, of course, is that sometimes people being imperfect as we all are, we fear making mistakes. Some of us are less fearful of that than others, though, which is why some people presented the “TLB erratum” in its proper context–that is, a minor erratum affecting a very small number of people and which AMD is correcting in the next revision of the chip, whereas other people covered the erratum as if it meant the entire cpu was ruined, worthless, and not fit for consumption by human beings…;)

    It’s always best to be widely read, imo, as you stand a better chance of getting a clearer picture of just about any situation than you do if you restrict yourself to single source of information. But I’m sure this is news to nobody…;)

      • spuppy
      • 12 years ago

      Wow, are you really comparing an errata that may cause a memory leak in some circumstances, which has since received a microcode update at no cost to performance, to another that crashes systems and fixing the error causes a 10-20% drop in performance?

      Really? Are you really?

        • VILLAIN_xx
        • 12 years ago

        Really. He really is not. The post was about Intel/AMD erratum coverege and not the forgiveness of performance loss.

    • pluscard
    • 12 years ago

    I’m not sure I should be surprised, that once again, TechReport outdoes it self in bashing AMD/ATI product.

    Every other review I’ve read said nothing touches the 790fx chipset for performance, despite the current phenom being the weak link.

    Techreport takes the opportunity to bash not once, but twice.

    I’m sorry, but TechReport appears to have a clear Intel bias at this point.

    Btw, AMD will be around for a very long time.

      • evermore
      • 12 years ago

      Did the other reviews test the same things, or did they just run a few game benchmarks? Package benchmarks like 3DMark and game benchmarks don’t test the things that TR has shown to be below-average in performance. Disk performance doesn’t change much in benchmarks that aren’t specifically testing it, and USB or Firewire performance isn’t tested at all for anything but data transfer tests. Ethernet performance also is only a data throughput test.

      Heck I can barely find any actual reviews, and none of them did in-depth tests like TR.

      The problems seem to be largely with the SB600. Maybe they’ll fix it with the SB700, but ATI’s 0 for 2 on good southbridges so far.

      TR isn’t biased, except in a bias towards better performance rather than crap.

      • Flying Fox
      • 12 years ago

      I’m not sure I should be surprised, that once again, you outdo yourself in bashing TR’s bashing of DAAMIT’s products.

      When one product is clearly inferior you call it like it is. So what do you think they should do to be called “not biased”? Call them the best thing since sliced bread even the thing is a dog? That is a really objective approach… NOT.

      Btw, TR did not say “AMD is toast” or “AMD is dead”. It seems that /[

        • n00b1e
        • 12 years ago

        Don’t bother replying, nothing will get through his impregnable fanboy shield.

      • Fighterpilot
      • 12 years ago

      TR was very harsh on Intel products right up until Intel released a really good CPU….the Conroe family.
      Then(correctly) they praised them and now its is AMD’s turn to take criticism of their current line up of CPUs which do not perform as well as the competition.
      That is how it should be.
      The glowing terms that are used in NVidia reviews here are justified by their excellent performance graphic cards.
      ATI has taken a lot of heat for their underperforming GPUs in recent years,however the current HD series cards are really rather good.
      Although I’d like to see a little more praise for the feature set which the new range has rather than just frames per second performance I think TR tells it like it is and is probably the most objective and trustworthy hardware review site on the Internet.
      Hang in their pluscard…AMD will be back…they have some damn fine competition to beat tho.

        • pluscard
        • 12 years ago

        TechReport clearly has an Intel/Nvidia bias. You can feel it in the sentiment of the article, and which components of the product they tend to feature.

        The HD3870 is a value card, that is available in a 4 way crossfireX configuration with the 790fx chipset, and the motherboards in this review. The pricing will obviously fall as AMD ramps production. Nvidia has no 4 way SLI to compete. Further, crossfire is superior to sli even in the 2 way configuration.

        The 790FX has support for the fastest memory, has the highest bandwidth, and supports up to 4 HD3870s. It’s a direct upgrade for X2 customers who want to upgrade their graphics now, and add the B3 Phenom when it’s available.

        This article is not critical of the 790FX’s performance which it says is “in line” with nvidia sli boards (despite spanking nvidia in most of the benchs). Rather they focus on usb, firewire and ahci drivers.

        They pound the 790 for cpu utilization while the sli board is at least twice as bad during serial ATA.

        This is by no means a “clean” review – it out and out says not to buy it, and suggests the inferior nvidia product, which limits you to nvidia inferior SLI.

        The worst thing in the world is a site that is subtle in it’s bashing. I

          • flip-mode
          • 12 years ago

          You are the biased one. You take exception to perfectly legit benchmarks and entire articles. You defend/support inferior products. Such constitutes a fanboy. People revealing lame products for what they are is not a problem; fanboys that obfuscate the lameness of their preferred brand’s products is a problem.

          And since you brought it up, the 790 is not a good upgrade for someone with an AM2 CPU. The P35 paired with a cheap E21xx is, which, on average, will overclock to performance levels beyond what, on average, you can get an Athlon X2 to. You can get a P35 and an E21xx for, on average, less than the price you can get a 790x mobo for.

            • donkeycrock
            • 12 years ago

            just wanted to say that i had a Amd X2 1.8 oc’ed to 2.4 Ghz:paid ($150 for cpu crappy mobo and decent ram), Just got a Intel E2160 1.8 oc’ed to 3.5 Ghz:paid ($205 for cpu and awesome mobo), the difference is night and day. I was a big AMD fan for the last 4 years, but i have to say that intel has made me a believer.

            I don’t think TR is bias, i think they are being accurate. If you want to get the most bang for your buck. Intel is superior. TR /[

          • Dissonance
          • 12 years ago

          As I’m sure you’re well aware, 4-way CrossFire drivers are not yet available.

          Also…

          /[

          • Flying Fox
          • 12 years ago

          If you want to call liking better performing parts as “bias”, then go ahead, continue hiding in whatever shell you are in…

          • radix
          • 12 years ago

          How many people are using SLI/Crossfire? This is mainly for people that want to throw money away. Take a look at the Steam statistics for graphic cards, I think it’s like less than 5% of all users have SLI or Crossfire on their machines, heck it’s probably much less than that, I am shooting high so you don’t complain. And as Dissonance pointed out, there are no 4 way Crossfire drivers available yet. And how can TR be biased? Are you reading the numbers correctly on the benchmarks? I’m sorry if you bought a lot of AMD shares before their price started to go down, calling TR biased when clearly only you are biased won’t help. Right now it’s wait and see if AMD can come up with something competitive quick, and I really want them to, a monopoly will do no good to customers.

      • cripplecore
      • 12 years ago

      You are one of the WORST AMD pumpers from the Yahoo AMD financial forums. Please stay there.

      • mako
      • 12 years ago

      You must be new here.

    • My Johnson
    • 12 years ago

    Off topic just barely…

    So, when is Intel going to drop the bomb, and sell quad core for under $200?

      • ssidbroadcast
      • 12 years ago

      They have no chance to survive make their time.

      • flip-mode
      • 12 years ago

      Not until Intel has a reason too (BTW, killing AMD is a disincentive to doing what you suggest).

      • pluscard
      • 12 years ago

      You can buy the AMD quad today from newegg for $199.

      Interestingly enough, Pricewatch.com lists the cheapest Intel quad for $277, but to go from 2.4ghz to 2.66ghz, you’ll have to pay $899.

      For $1079, you step up to 2.93ghz, but at that point you might as well pay $1099 for the 3.0ghz model.

      I guess it cost a lot more to make those higher clocked models.

      Or something like that…

        • evermore
        • 12 years ago

        A 2.3GHz Phenom is $239 at Newegg, which would be a more appropriate comparison, as TR’s benchmarking shows even that is at best a tiny bit ahead of Q6600, more often just even with the Q6600, without the TLB patch, and in some cases vastly worse (usually but not always with the patch). A 2.4GHz Phenom would likely be more expensive than a Q6600, while only offering performance about on-par (until we see what fixed versions can do) and still losing significantly in some cases. Even a Phenom ES 2.6GHz is barely able to outdo the Q6600, and still gets spanked in many tests.

        I don’t expect any AMD chip to magically appear and wipe the floor with Intel in EVERY test, anymore than I would expect the reverse to happen. Some tests do better on one architecture, some on the other, but we expected Phenom to at least outperform Core2 occasionally, and not have such huge losses in so many tests. Other than pure memory bandwidth testing, Phenom doesn’t have a lot to recommend it.

        Don’t try to turn this into an argument that Intel is charging too much so we should support AMD. The ONLY reason AMD charges so little is because that’s the only way it can make sales in many cases. Intel charges more for higher clock speeds for valid reasons in terms of market demand AND manufacturing costs. AMD prices the same way, but also has to price based on competing against the “default” processor that people will usually select, Intel.

        Incidentally, 2.66GHz Q6700 is only $540 on Newegg. At least try not to compare apples to caviar. The $899 price is for a Core 2 Extreme QX6700.

    • bigfootape
    • 12 years ago

    Any chance of performing the IOMeter tests in XP?

    • srg86
    • 12 years ago

    TR said: “By then, AMD should have B3-stepping Phenom chips available that correct the TLB erratum, as well.”

    I’m afraid that if this is correct:

    §[<http://www.xbitlabs.com/news/cpu/display/20071226214456_Advanced_Micro_Devices_Faces_Further_Delays_of_Phenom_Processors.html<]§ Then the B3 Stepping is further delayed into Q2 2008. AMD just seems to be a laughing stock at the moment.

    • Dposcorp
    • 12 years ago

    Guys, if I can go all AMD and get my moneys worth, why not?
    How about a newer 790FX board w/ the new SB, B3 Phenom, and a 3870 or better for around $300.
    That aint bad and I could re-use my DDR2.

    If they drop the phenom and video card in price, it could happen

      • cygnus1
      • 12 years ago

      i don’t know about $300, but $400 or $500. all of that for under $500 would still be a pretty good deal, under $400 would be a steal, but $300 is just crazy talk.

        • Dposcorp
        • 12 years ago

        $300 could work.

        $100 X3 Phenom, $75 board, $125 3870.

        Or close to it.

        Prices drop when new things come out.

          • flip-mode
          • 12 years ago

          Yeah, um, those prices you posted are not going to happen for quite some time, 9-12 months would be my guess.

    • Dposcorp
    • 12 years ago

    Thanks for the update.
    I really wanna go all AMD for my next major upgrade, so I guess i’ll wait a bit more.

      • yfital
      • 12 years ago

      I’m really not sure if AMD is the way to go, low end or high end.
      Got mixed feelings.
      If i would have upgraded, i would probably go Nvidia/Intel, i feel safer with them right now…

        • tesmar
        • 12 years ago

        He probably is doing it out of brand loyalty.

        • RambodasCordas
        • 12 years ago

        Safer?! Didn’t understand that part.
        What’s wrong with the 3850/3870 cards? The fact they already have DX10.1 for example? I bet if it was the other way around everyone would be pointing out how Ati was outdated or something like that.
        And the X2 still competes greatly with the Celeron/E2xxx/E4xxx and with some of the E6xxx.
        The Intel quads are practically unbeatable but if AMD prices then a little lower why not?

      • willyolio
      • 12 years ago

      i was planning the same thing, but now it looks like my next rig will have no AMD at all.

      i hope the next graphics refresh will at least give me reason to buy one part from them…

Pin It on Pinterest

Share This