|
|
| Overview and Specifications | |||||
ATI has been particularly aggressive with their FireGL workstation graphics card lineup over the past six months. While their flagship R600-architecture was the basis for the solid (but ultimately underwhelming) Radeon HD2900XT release for the gaming market, our tests have shown that this particular GPU architecture actually is much more competitive in the workstation market. While the R600 may not be the best choice for a high-end gaming rig, it is indeed a powerful chip for high-end OpenGL work, and as such, has led to a revival of sorts for ATI’s FireGL lineup. With an excellent performing workstation GPU in tow, ATI finally took the steps to get their latest generation of FireGL cards out to the market fast – much faster than we’ve seen from them in the past. ATI now has a number of R600-based workstation products out on the market, and is competing quite well with NVIDIA in terms of both price and performance. ATI’s cards generally offer nearly all the bang for the buck as the QuadroFX lineup, but ATI has some unique models as well which give them their own following of users. One of their newest models, the FireGL V8600, has finally hit the market. This is the card we’ll be looking at today. The FireGL V8600 follows on the heels of its father, the big daddy FireGL V8650. The FireGL V8650 is ATI’s current ultra high-end product, featuring ATI’s R600 GPU in all its 320-shader processor glory, paired with 2 GB of high-speed GDDR memory - it's a beast of a card – and sells for well over $2,000 on the street. The FireGL V8600, on the other hand, follows this same mentality but is running with a smaller frame buffer – a "mere" 1 GB of memory onboard. In most situations, the extra 1 GB of memory on the FireGL V8650 isn’t being used, so ATI has lopped off this extra memory and is now offering a trimmed down version which is far more competitive in terms of price. Considering that most mid to high-end applications don’t even come close to using a full 2 GB frame buffer, this 1 GB card should perform with a smidgen of ATI’s much more expensive V8650, but will cost a lot less. Obviously, this is good news for any potential buyer. With 1 GB of memory, the FireGL V8600 has an MSRP $900 less than the FireGL V8650 ($1,899 vs. $2,799), and is already selling for around $1,500 on the street. Of course, all of this is nicely timed, as NVIDIA’s QuadroFX 3700 board also just shipped with a very similar MSRP ($1,699) and is just now hitting the review circuit. ATI is targeting the FireGL V8600 right at NVIDIA’s new card, touting its double-sized 1 GB of memory and full 320-shader processor architecture at roughly the same price point. Let’s see what the better option is, shall we? ATI's massive FireGL V8600 1 GB graphics card in all its glory.
|
| Cooling and Power Consumption | ||||||||
The cooling system which is housed on the FireGL V8600 is simply massive. ATI equips this card with a (seemingly) military-grade cooling system which uses copper heatpipes, copper thin-fins, a centrifugal blower, and a series of heavy-duty aluminum rods to cool the power conversion modules. The cooling system makes the card extremely heavy compared to most other graphics cards on the market, which is likely part of the reason why the card is equipped with a plastic holding bracket on the right side, which can provide additional support from compatible cases and decrease the weight load on the I/O shield and PCI Express connector.
However, once you start to put load on the GPU, the fan speed kicks up and immediately becomes noticeable. Again, we’re not saying that the noise level is intolerable, but it makes its presence known and can be slightly distracting if you’re trying to focus on the task at hand. The cooling system also takes far longer to spin-down after an intensive task to idle mode compared to all the members of the QuadroFX lineup we’ve seen thus far. It looks like ATI has addressed these issues with the new FireGL V7700 product, which uses a newer cooling system and a lighter heat producing GPU, but for their high-end products like the V8600 and V8650, you simply have to get used to a bit of noise. Sad, but true. We will say, however, that the raw weight and complexity of ATI’s cooling system is quite impressive, and we would have faith that this cooling system can handle un-friendly thermal environments with ease.
![]()
The FireGL V8600 is not extremely friendly in terms of power consumption, as seen by our charts. This card somehow managed to test slightly higher compared to the FireGL V8650 in terms of overall power consumption, outpacing all of the other cards in our test. Keep in mind, this card is competing against the QuadroFX 3700 which consumes far less power. Under load, the 3700 is consuming about 100 watts less than the V8600 card, which is nothing to scoff at. The FireGL V8600 even needs more power than NVIDIA’s monstrous G80-based QuadroFX 5600 1.5GB card. |
| Testbed and Cinebench | ||||||||
Just as with our other workstation-class GPU articles – we’ve tested these cards with Windows XP (32-bit). Both NVIDIA and ATI claim XP to still be the dominant workstation-class operating system, and performance is fine tuned for this operating system in order to reflect that.
As the FireGL V8600 and V8650 share the same clock speeds, it’s not surprising to set them score nearly identically in Cinebench, as this application does not consume more than 1 GB of frame buffer memory. The FireGL V8600 is able to outperform the QuadroFX 5600 card, but gets bested by NVIDIA’s new QuadroFX 3700 in this particular test. |
| 3D Studio Max 2008 and Maya 8.5 | ||||||||
We actually see a slight performance difference between the FireGL V8600 and V8650 in 3D Studio Max, which leads us to believe that this particular benchmark is indeed using more than 1 GB of frame buffer memory during some sequences. For the most part, though, it performs between the V7600 and V8650 products, and overall, not particularly great compared to NVIDIA’s product line in this test. The FireGL product lineup does remain strong in Maya 8.5, as again we see the V8600 card performing roughly at the same level as the more expensive V8650 card. |
| SPEC ViewPerf 10 | ||||
![]() ![]() ![]() ![]() Our first set of Viewperf 10 tests show virtually no difference between the V8600 and V8650 product lineups, and the overall performance charts don’t change. The FireGL V8600 shows excellent numbers in Maya, 3dmax and Ensight, although is bested by NVIDIA’s product lineup in Catia. |
| SPEC ViewPerf 10 (Continued) | ||||
![]() ![]() ![]() ![]() The V8600 gets wailed on in ProEngineer and SolidWorks, although puts up solid numbers in TWX and dominating scores in UGNX. Again, virtually no difference between the V8600 and V8650 cards. |
| GPGPU Computing | ||||
![]() ![]() ![]() For GPGPU computing purposes, frame buffer size has less of a performance impact, so it’s not surprising that the V8600 and V8650 cards are virtually identical here. Sadly though, both of ATI’s high-end FireGL cards are bested by NVIDIA’s high-end product lineup in this particular test, although the FireGL cards aren’t far behind at all. |
| GPGPU Computing (Continued) | ||||
![]() ![]() ![]() Our second round of GPGPU computing tests shows the V8600 and QuadroFX 3700 almost on parallel in terms of computing power, although the QuadroFX 3700 takes a slight lead in all three tests. Here we see absolutely zero performance difference between the FireGL V8600 and V8650 cards. |
| Gaming Performance | ||||||||
The FireGL V8600 tests slightly better in terms of gaming performance compared to the FireGL V8650. While the V8600 is able to outperform the QuadroFX 3700 in 3DMark06, NVIDIA’s card does take the performance lead in Crysis by a small amount. |
| Synthetic OpenGL Effects Benchmarks | ||||||||
Our synthetic OpenGL effects benchmarks show the FireGL V8600/V8650 cards providing excellent performance in our “Fur” test, and slightly better performance compared to the QuadroFX 3700 card in the “Soft Shadows” test. Solid performance from both the NVIDIA and ATI camps. |
| Conclusion | ||||
While the FireGL V8600 is likely the last of the R600-based cards to be released for their current generation workstation lineup, its presence is still greatly appreciated. The FireGL V8600 gives more or less identical performance to that of ATI’s top of the line FireGL V8650 card, but at a much more attractive price point. As the V8600 and V8560 are identical, with the only difference being memory size, the similar performance levels are not incredibly surprising. While we ran into a few tests which showed a performance advantage with the FireGL V8650’s 2 GB memory, in most cases, our FireGL V8600 1 GB and FireGL V8650 2 GB cards typically were within 1-2% of each other in the benchmarks. Considering the FireGL V8600 is 33% less expensive, we’d say that it offers a much better price/performance ratio compared to the high-end V8650 card.
While the newly launched FireGL V8600 shares a similar MSRP to NVIDIA’s new QuadroFX 3700 card ($1,899 for V8600 vs. $1,599 for the 3700), a quick check online shows that the QuadroFX 3700 has dropped substantially thus far, while the V8600 has only dropped by a small amount. At the time of this writing, it’s about $600 less expensive to pick up a QuadroFX 3700 card compared to a FireGL V8600, which makes the FireGL V8600 a tough sell if you look at the benchmarks. While there are some places where ATI’s R600 architecture can dominate Nvidia’s G92 architecture (Maya, UGNX), in the majority of our benchmarks, performance between the FireGL V8600 and QuadroFX 3700 is remarkably close. Considering the street price differences between these cards and that the QuadroFX 3700 runs significantly cooler, quieter, and consumes less power, the FireGL V8600 is a tough card to recommend right now. A few months ago, our outlook would have been different, but Nvidia has definitely tightened the race lately. The FireGL V8600 does have its advantages, though. If you need a large frame-buffer, the V8600 has double the onboard memory of the QuadroFX 3700, which can come in quite valuable depending on your application load. The card also supports Genlock/Framelock, which the FX3700 simply doesn't support.
|