Original Link: http://www.anandtech.com/show/4051/nvidias-geforce-gtx-570-filling-in-the-gaps
NVIDIA's GeForce GTX 570: Filling In The Gapsby Ryan Smith on December 7, 2010 9:00 AM EST
NVIDIA can be a very predictable company at times. It’s almost unheard of for them to release only a single product based on a high-end GPU, so when they released the excellent GeForce GTX 580 last month we knew it was only a matter of time until additional GTX 500 series cards would join their product lineup.
Now less than a month after the launch of the GTX 580 that time has come. Today NVIDIA is launching the GeForce GTX 570, the second card to utilize their new GF110 GPU. As the spiritual successor to the GTX 470 and very much the literal successor to the GTX 480, the GTX 570 brings the GTX 580’s improvements to a lower priced, lower performing card. Furthermore at $350 it serves to fill in the sizable gap between NVIDIA’s existing GTX 580 and GTX 470 cards.
So how does NVIDIA’s latest and second greatest stack up, and is it a worthy sibling to the GTX 580? Let’s find out.
|GTX 580||GTX 570||GTX 480||GTX 470|
|Texture Address / Filtering||64/64||60/60||60/60||56/56|
|Memory Clock||1002MHz (4008MHz data rate) GDDR5||950MHz (3800MHz data rate) GDDR5||924MHz (3696MHz data rate) GDDR5||837MHz (3348MHz data rate) GDDR5|
|Memory Bus Width||384-bit||320-bit||384-bit||320-bit|
|FP64||1/8 FP32||1/8 FP32||1/8 FP32||1/8 FP32|
|Manufacturing Process||TSMC 40nm||TSMC 40nm||TSMC 40nm||TSMC 40nm|
The GTX 570 is likely the closest thing we’ll see to a GF110 version of GTX 480 – or any other GF100 card for that matter. With the higher yields afforded by the GF110 design and TSMC’s process improvements, we’ve already seen NVIDIA go for a fully operational GF110 design in the GTX 580, so the GTX 570 works from there. The end result is a melding of the GTX 480’s shader count with the GTX 470’s ROP count and memory bus, and with a clockspeed a bit over GTX 480 and well over GTX 470, performance is much closer to the GTX 480 than the GTX 470.
With 15 of 16 SMs enabled, the GTX 570 matches the GTX 480 at a total of 480 active CUDA Cores and 60 texture units. The core clock is 732MHz, 32MHz (4.5%) over the GTX 480 in order to make up for the reduced ROP/memory blocks and to take advantage of GF110’s lower leakage at higher clocks (as a minor aside, why the strange clocks lately? Look to the PLL). Meanwhile the memory system uses the same 320bit (64bit x 5) memory bus & 10 memory chip configuration we saw on the GTX 470, however this time the memory clock is up to 950MHz (3.8GHz data rate), 113MHz (13.5%) over the GTX 470. Memory clocks are also marginally faster than the GTX 480 by 26MHz, but this isn’t nearly enough to make up for the narrower memory bus. Finally we have the ROPs, which share an existence with both the core and memory subsystems and split the difference – it’s the same 60 ROPs and 640KB of L2 cache as the GTX 470, but because the ROPs run on the core clock they’re running 125MHz (20.5%) faster than the GTX 470.
Since it’s based on GF110, GTX 570 also shares the same architectural enhancements we first saw in the GTX 580. This means GTX 570 can retire twice as many FP16 texels per clock as GTX 480, and it also features NVIDIA’s improved Z-culling system. For the GTX 570 this helps to further close the potential performance gap between the GTX 570 and GTX 480 that results from the lower ROP count and narrower memory bus. Do note however that compared to the GTX 470 the overall improvements are asymmetric: we’re looking at around a 30% theoretical improvement in shading/compute/texture performance, but only a 13.5% improvement in memory bandwidth, so unlike the GTX 580 and its balanced approach, the difference on the GTX 570 is going to be greater on shader-bound games and applications, and lesser when we’re memory bandwidth limited.
As the GTX 470’s successor, the GTX 570 generally fits in the same power and noise profile as the GTX 470. NVIDIA puts the TDP at 219W – a mere 4W over GTX 470 – highlighting the fact that NVIDIA has gone for maximizing performance within their selected power profile for the GTX 570, versus increasing performance but also decreasing power consumption to the GTX 580. The card is otherwise identical to the GTX 580 – the GTX 570 uses the same PCB, the same vapor chamber cooler, and the same shroud as the GTX 580.
NVIDIA is putting the MSRP for the card at $349, a price that in recent weeks has been vacant as neither NVIDIA or AMD had a product to put between the 480/470 and 5970/5870 respectively. Coming from the top-end of the market this is more or less a nice price drop for GTX 480-like performance, but it also means the Radeon 5870 and GTX 470 are the GTX 570’s value threats – the 570’s a good bit faster, but they’re nearly $100 cheaper. The only other competition for the GTX 570 for now will be the GTX 460 1GB SLI and the Radeon HD 6850 CF.
Today’s launch should be a hard launch. Going in to the GTX 580 launch we had our doubts that NVIDIA could have so many GF110 products ready on such short notice, but they were able to prove us wrong there and we’re willing to take them at face value on this. Based on their own estimates and the lower price of the GTX 570 we’d expect some cards to sell out, but availability shouldn’t be an issue.
Finally, with the launch of the GTX 570, NVIDIA’s lineup will be shifting. GF110 is a very effective replacement for GF100 and NVIDIA will be looking to phase out GF100 cards as quickly as they reasonably can. The GTX 470 will still be around for quite some time (all indications are that NVIDIA still has a lot of GF100 chips left) but GTX 480’s days are numbered.
|Winter 2010 Video Card MSRPs|
|$470||Radeon HD 5970|
||$250||Radeon HD 5870|
|$240||Radeon HD 6870|
|$180-$190||Radeon HD 6850|
Meet the GTX 570
As we quickly touched upon in our introduction, the GTX 570 is a complete reuse of the GTX 580 design. NVIDIA used the same PCB, cooler, power throttling chips, and shroud as the GTX 580; our reference card is even clad nearly identical livery as the GTX 580. Indeed the only hardware difference between the two cards from the outside is that the GTX 580 uses 6+8pin PCIe power sockets, while the GTX 570 uses 6+6pin PCI power sockets.
By using the same design as the GTX 580 this no doubt has let NVIDIA bring the GTX 570 to the market quickly and cheaply, but it also means that the GTX 570 inherits the same design improvements that we saw on the GTX 580. This means the cooling for the GTX 570 is provided by a vapor chamber-based aluminum heatsink, backed by NVIDIA’s reinforced blower. For SLI users this also means that it’s using NVIDIA’s angled shroud that is supposed to allow for better cooling in tight spaces. As we’ve already seen on the GTX 580 this design can give the GTX 470 a run for its money, so it shouldn’t be a surprise when we say that the GTX 570 is similarly capable. Overall the only notable downside to this design is that because NVIDIA is using the GTX 580 design it’s also inheriting the GTX 580’s 10.5” length, making the GTX 570 an inch longer than the GTX 470.
As with the GTX 580 the situation with custom GTX 570s will be nebulous. NVIDIA is taking tighter control of the GTX 500 series and will only be approving designs that are equal or superior to the reference design. This isn’t a bad thing, but it means there’s less latitude for custom designs, particularly if someone wants to try lobbing an inch off of the card to make it the same length as the GTX 470. Interestingly, triple-slot coolers are also out – we found out last week that NVIDIA is vetoing them on the GTX 580 (and no doubt the GTX 570) as they aren’t suitable for use in SLI mode with most motherboards, so any custom designs that do appear will definitely be more conservative than what we’ve seen with the GTX 400 series.
Since NVIDIA is reusing the GTX 580 PCB, I/O is identical to the GTX 580. Here it’s covered by the usual NVIDIA configuration of 2x DVI ports and 1x mini-HDMI port, with the second slot occupied by the card’s exhaust. This also means the card can only drive 2 of the 3 ports at once, meaning you’ll need an SLI configuration to take advantage of NVIDIA/3DVision Surround. Meanwhile HDMI 1.4a for 3D video purposes is supported by the card’s mini-HDMI port, but audio bitstreaming is not supported, limiting audio output to LPCM and DD+/DTS.
Due to a last-minute fiasco with our second GTX 570, we’ll only be looking at single-GPU performance today. We’ll update our article with GTX 570 SLI performance as soon as we have a replacement card.
The launch drivers for the GTX 570 are 263.09; the performance is identical to the 262.99 launch drivers used for the GTX 580 last month. Meanwhile on the AMD side we’re using Catalyst 10.10e – which in spite of the name are a newer revision than 10.11.
Finally, all tests were done with the default driver settings unless otherwise noted.
|CPU:||Intel Core i7-920 @ 3.33GHz|
|Motherboard:||Asus Rampage II Extreme|
|Chipset Drivers:||Intel 184.108.40.2065 (Intel)|
|Hard Disk:||OCZ Summit (120GB)|
|Memory:||Patriot Viper DDR3-1333 3 x 2GB (7-7-7-20)|
AMD Radeon HD 6870
AMD Radeon HD 6850
AMD Radeon HD 5970
AMD Radeon HD 5870
AMD Radeon HD 5850
AMD Radeon HD 5770
AMD Radeon HD 4870
NVIDIA GeForce GTX 580
NVIDIA GeForce GTX 570
NVIDIA GeForce GTX 480
NVIDIA GeForce GTX 470
NVIDIA GeForce GTX 460 1GB
NVIDIA GeForce GTX 460 768MB
NVIDIA GeForce GTS 450
NVIDIA GeForce GTX 285
NVIDIA GeForce GTX 260 Core 216
NVIDIA ForceWare 262.99
NVIDIA ForceWare 263.09
AMD Catalyst 10.10e
|OS:||Windows 7 Ultimate 64-bit|
Kicking things off as always is Crysis: Warhead, still one of the toughest game in our benchmark suite. Even 2 years since the release of the original Crysis, “but can it run Crysis?” is still an important question, and the answer continues to be “no.” While we’re closer than ever, full Enthusiast settings at a playable framerate is still beyond the grasp of a single card.
Crysis end up setting the stage for the rest of this article. As a GTX 480 replacement the GTX 570 is effectively tied with it at 2560 and 1680, and only at 1920 do we see the GTX 570 fall behind by all of 4%. Meanwhile compared to the GTX 470 it’s 20% faster (and 40% more expensive), while it falls to the GTX 460 1GB SLI by over 10%. Overall the GTX 570 is at near parity with the GTX 480, and should be equally capable of playing just about everything at 1920.
As for AMD’s cards, the 5870 (which was never too far behind the GTX 480) nips at the GTX 570’s heels – at times the GTX 570 is no more than 5% faster and no better than 15% faster, showcasing why the 5870 is a threat to the GTX 570 as a value threat. Meanwhile the 6850 CF is tops here at 1920 by a wide margin for only around $20-$30 more than the GTX 570. As was the case with the GTX 580, a pair of lesser AMD cards is going to offer better gaming performance in exchange for the drawbacks of a multi-GPU setup.
Looking at our minimum framerates, the story is much the same. Outside of 2560 where the extra memory provides a stark advantage for the NVIDIA cards, the GTX 570 and GTX 480 are close together except for 1920 where the 570 falls behind by a bit more than we’d expect. The 5870 isn’t nearly as threatening here as it is with average framerates, but the 460SLI/6850CF configurations are still well ahead.
Up next is BattleForge, Electronic Arts’ free to play online RTS. As far as RTSes go this game can be quite demanding, and this is without the game’s DX11 features.
With no more than 1fps separating the GTX 570 and GTX 480 at all times, this is as close as a match can come. The 570 and 480 are interchangeable here, driving home the point that the GTX 570 will fill the role of a cooler, quieter, and cheaper GTX 480. Furthermore compared to the GTX 470 we’re now a good 25% ahead, well exceeding the memory bandwidth advantage and approaching the theoretical shader performance advantage.
As for the competition, the Radeon 5870 is no threat to the GTX 570. The only real threat is the SLI/CF cards, which enjoy a notable performance advantage but nothing like what we saw with Crysis.
The next game on our list is 4A Games’ Metro 2033, their tunnel shooter released earlier this year. In September the game finally received a major patch resolving some outstanding image quality issues with the game, finally making it suitable for use in our benchmark suite. At the same time a dedicated benchmark mode was added to the game, giving us the ability to reliably benchmark much more stressful situations than we could with FRAPS. If Crysis is a tropical GPU killer, then Metro would be its underground counterpart.
While Metro was an outstanding game for the GTX 580 to show off its performance advantage, the situation is quite different for the GTX 570. Here it once again fulfills its role as a GTX 480 replacement, but it’s far more mortal when it comes to being compared to other cards. The higher the resolution the closer the 5870 and GTX 570 are, with this being only a 10% lead for the GTX 570 at 1920. However the SLI/CF cards are in a similar situation, and only pull ahead of the GTX 570 by around 10%.
As for the GTX 470, the GTX 570 is once again 25% faster at 1920, a reasonable outcome for this shader-heavy game.
Ubisoft’s 2008 aerial action game is one of the less demanding games in our benchmark suite, particularly for the latest generation of cards. However it’s fairly unique in that it’s one of the few flying games of any kind that comes with a proper benchmark.
HAWX is a game NVIDIA traditionally does well at, and the GTX 570 is no exception to the rule. In fact in one of the better showings for the card, as it consistently beats the GTX 480 by 5%, a feat we believe can be attributed to GF110’s Z-cull enhancements. This also puts the GTX 470 in a particularly precarious place, since the GTX 570 is now over 30% faster, ever so slightly exceeding the maximum theoretical shader performance improvement.
Meanwhile a strong NVIDIA showing also means the Radeons aren’t nearly as threatening. The GTX 570 is 42% faster than the 5870 here, and while the SLI/CF cards top it, it’s by a much tamer 16%.
The other new game in our benchmark suite is Civilization 5, the latest incarnation in Firaxis Games’ series of turn-based strategy games. Civ 5 gives us an interesting look at things that not even RTSes can match, with a much weaker focus on shading in the game world, and a much greater focus on creating the geometry needed to bring such a world to life. In doing so it uses a slew of DirectX 11 technologies, including tessellation for said geometry and compute shaders for on-the-fly texture decompression.
Civ 5 is a very interesting game for us. It’s one of the few games we’ve ever tested where performance increases with resolution, something we’re content to attribute to its unique use of DX11 features and what may be a better match between tessellated triangles and 16 pixel quads at higher resolutions. The result is that while it’s CPU limited most of the time, there’s very little reason to not crank up the resolution to your monitor’s native resolution.
Because we’re CPU limited several things are going on. Until we hit 2560 the 470, 480, 570, and 580 are all effectively tied with each other, with the 470 finally falling behind in the end. As a result at 2560 it appears that a GTX 570 is all that’s necessary to take full advantage of the game. Meanwhile NVIDIA has always done better than AMD here in single-GPU scenarios, so the Radeon 5870 is nearly 20% behind the GTX 570. On the flip side the 6850CF shoots well ahead of the GTX 570, hitting what appears to be the cap for AMD cards at around 51fps.
Battlefield: Bad Company 2
The latest game in the Battlefield series - Bad Company 2 – remains as one of the cornerstone DX11 games in our benchmark suite. As BC2 doesn’t have a built-in benchmark or recording mode, here we take a FRAPS run of the jeep chase in the first act, which as an on-rails portion of the game provides very consistent results and a spectacle of explosions, trees, and more.
Bad Company 2 is another game where the GTX 570 takes a notable lead over the GTX 480, and is clearly exceeding the theoretical clockspeed advantages of the GTX 570 and drifting in to architectural optimizations. Unfortunately this advantage for the GTX 570 plays out best at lower resolutions, resulting in the gap disappearing by the time we hit 2560. Even at 1920 the advantage is barely worth mentioning, meaning we’re once more at parity with the GTX 480.
Perhaps it would have been better for NVIDIA if that advantage had held, because it means the Radeon 5870 gets uncomfortably close at higher resolutions. Negligible at 2560, the GTX 570 advantage is only 10% at 1920. AMD’s strong CF scaling also puts NVIDIA in a tough position here, as the 6850 CF is a whopping 32% faster than the GTX 570 when it comes to Bad Company 2.
NVIDIA does manage to turn the tables with our Waterfall benchmark however, which serves as a proxy for minimum framerates. The GTX 570 is still tied with the GTX 480 here, but it’s also now at parity with the 6850CF and over 50% faster than the 5870, easily demonstrating that if you’re worried more about minimums than averages in Bad Company 2 that the 5870 and GTX 570 aren’t nearly as close as they were at first glance. Extra RAM would probably be of great benefit to AMD here.
STALKER: Call of Pripyat
The third game in the STALKER series continues to build on GSC Game World’s X-Ray Engine by adding DX11 support, tessellation, and more. This also makes it another one of the highly demanding games in our benchmark suite.
STALKER is another game where the GTX 570 and GTX 480 spend some quality time together in our charts, quickly running in to each other once we’re at 1920 at 2560. Similarly at these resolutions the GTX 570 enjoys a roughly 26% lead over the GTX 470, and even the Radeon 5870 isn’t much of a threat. As expected the SLI and CF configurations do pull ahead, but in the case of the GTX 460 SLI this is only a minor lead at best, while the 6850 CF enjoys its usual 20% lead.
Codemasters’ 2009 off-road racing game continues its reign as the token racer in our benchmark suite. As the first DX11 racer, DiRT 2 makes pretty thorough use of the DX11’s tessellation abilities, not to mention still being the best looking racer we have ever seen.
DiRT 2 is another game where the GTX 570 enjoys an advantage over the GTX 480 beyond a simple difference in core clocks. Once we move up to less CPU-limited resolutions we’re looking at a 4-6% performance advantage, giving the GTX 570 a small but noticeable lead. This lead also gives it a larger than normal advantage over the GTX 470, pushing the gap to over 35%.
As for the competition, this is another title where the Radeon 5870 can’t close the gap; in fact it does a bit worse than the GTX 470. The SLI/CF cards do better here, particularly the NVIDIA cards with their better scaling under this game.
Mass Effect 2
Electronic Arts’ space-faring RPG is our Unreal Engine 3 game. While it doesn’t have a built in benchmark, it does let us force anti-aliasing through driver control panels, giving us a better idea of UE3’s performance at higher quality settings. Since we can’t use a recording/benchmark in ME2, we use FRAPS to record a short run.
Coming from our past games, Mass Effect 2 throws us a bit of a curveball as it’s the only other game where the GTX 570 has any kind of remarkable disadvantage compared to the GTX 480; albeit tiny, the GTX 570 trails the GTX 480 by around 4% here. This is also one of the smaller advantages for the GTX 570 over the GTX 470, showing a smaller 17% gain. From what we’re seeing with the GTX 500 series, it looks to be the game in our suite most likely to be memory bandwidth bottlenecked.
With that memory bottleneck the GTX 570 doesn’t have a chance to run too far ahead of its AMD competition. At 1920 the performance advantage over the 5870 is only 13%, and of course the SLI/CF cards do quite well here by over 35%.
Finally among our benchmark suite we have Wolfenstein, the most recent game to be released using the id Software Tech 4 engine. All things considered it’s not a very graphically intensive game, but at this point it’s the most recent OpenGL title available. It’s more than likely the entire OpenGL landscape will be thrown upside-down once id releases Rage next year.
At lower resolutions Wolfenstein is CPU limited, and even 1920 isn’t too far off from this with our latest crop of cards. With this in mind, the GTX 570 finds itself on par with the GTX 480, while against the GTX 470 it enjoys a 25% lead.
Conversely AMD’s strong OpenGL performance undermines what little lead NVIDIA would have. At 1920 the GTX 570’s lead over the 5870 is only 7%, though this is largely an academic difference since even the 5870 gets more than 60fps at 2560.
Compute & Normalized Numbers
Moving on from our look at gaming performance, we have our customary look at compute performance, bundled with a look at theoretical tessellation performance. Unlike our gaming benchmarks where NVIDIA’s architectural enhancements could have an impact, everything here should be dictated by the core clock and SMs, with the GTX 570’s slight core clock advantage over the GTX 480 defining most of these tests.
Our first compute benchmark comes from Civilization V, which uses DirectCompute to decompress textures on the fly. Civ V includes a sub-benchmark that exclusively tests the speed of their texture decompression algorithm by repeatedly decompressing the textures required for one of the game’s leader scenes.
The core clock advantage for the GTX 570 here is 4.5%; in practice it leads to a difference of less than 2% for Civilization V’s texture decompression test. Even the lead over the GTX 470 is a bit less than usual, at 23%. Nor should the lack of a competitive placement from an AMD product be a surprise, as NVIDIA’s cards consistently do well at this test, lending credit to the idea that it’s a compute application better suited for NVIDIA’s scalar processor design.
Our second GPU compute benchmark is SmallLuxGPU, the GPU ray tracing branch of the open source LuxRender renderer. While it’s still in beta, SmallLuxGPU recently hit a milestone by implementing a complete ray tracing engine in OpenCL, allowing them to fully offload the process to the GPU. It’s this ray tracing engine we’re testing.
SmallLuxGPU is rather straightforward in its requirements: compute and lots of it. The GTX 570’s core clock advantage over the GTX 480 drives a fairly straightforward 4% performance improvement, roughly in line with the theoretical maximum. The reduction in memory bandwidth and L2 cache does not seem to impact SmallLuxGPU. Meanwhile the advantage over the GTX 470 doesn’t quite reach its theoretical maximum, but the GTX 570 is still 27% faster.
However as was the case with the GTX 580, all of the NVIDIA cards fall to AMD’s faster cards here; the GTX 570 is only between the 6850 and 6870 in performance, thanks to AMD’s compute-heavy VLIW5 design that SmallLuxGPU excels at. The situation is quite bad for the GTX 570 as a result, with the top card being the Radeon 5870, which the GTX 570 underperforms by 27%.
Our final compute benchmark is a Folding @ Home benchmark. Given NVIDIA’s focus on compute for Fermi and in particular GF110 and GF100, cards such as the GTX 580 can be particularly interesting for distributed computing enthusiasts, who are usually looking for the fastest card in the coolest package.
Once more the performance advantage for the GTX 570 matches its core clock advantage. If not for the fact that a DC project like F@H is trivial to scale to multi-GPU configurations, the GTX 570 would likely be the sweet spot for price, performance, and power/noise.
Finally, to take another look at GTX 570’s performance, we have the return of our normalized data view that we first saw with our look at the GTX 580. Unlike the GTX 580 which had similar memory/ROP abilities as the GTX 480 but more SMs, the GTX 570 contains the same number of SMs with fewer ROPs and a narrower memory bus. As such while a normalized dataset for the GTX 580 shows the advantage of the GF110’s architectural enhancements and the highter SM count, the normalized dataset for the GTX 570 shows the architectural enhancements alongside the impact of lost memory bandwidth, ROPs, and L2 cache.
The results certainly paint an interesting picture. Just about everything is ultimately affected by the lack of memory bandwidth, L2 cache, and ROPs; if the GTX 570 didn’t normally have its core clock advantage, it would generally lose to the GTX 480 by small amounts. The standouts here include STALKER, Mass Effect 2, and BattleForge which are all clearly among the most memory-hobbled titles.
On the other hand we have DIRT 2 and HAWX, both of which show a 4% improvement even though our normalized GTX 570 is worse compared to a GTX 480 in every way except architectural advantages. Clearly these were some of the games NVIDIA had in mind when they were tweaking GF110.
Power, Temperature, & Noise
Last but not least as always is our look at the power consumption, temperatures, and acoustics of the GTX 570. While NVIDIA’s optimizations to these attributes were largely focused on bringing the GTX 480’s immediate successor to more reasonable levels, the GTX 570 certainly stands to gain some benefits too.
Starting with VIDs, we once again only have 1 card so there’s not too much data we can draw. Our GTX 570 sample has a lower VID than our GTX 580, but as we know NVIDIA is using a range of VIDs for each card. From what we’ve seen with the GTX 470 we’d expect the average GTX 570 VID to be higher than the average GTX 580 VID, as NVIDIA is likely using ASICs with damaged ROPs/SMs for the GTX 570, along with ASICs that wouldn’t make the cut at a suitable voltage with all functional units turned on. As a result the full power advantage of lower clockspeeds and fewer functional units would not always be realized here.
|GeForce GTX 500 Series Voltages|
|Ref 580 Load||Asus 580 Load||Ref 570 Load||Ref 570 Idle|
With fewer functional units than the GTX 580 and a less leaky manufacturing process than the GTX 400 series, the GTX 570 is the first GF1x0 card to bring our test rig under 170W at idle. NVIDIA’s larger chips still do worse than AMD’s smaller chips here though, which is driven home by the point that the 6850 CF is only drawing 1 more watt at idle. Nevertheless this is a 4W improvement over the GTX 470 and a 21W improvement over the GTX 480; the former in particular showcases the process improvements as there are more functional units and yet power consumption has dropped.
Under load we once again see NVIDIA’s design and TSMC’s process improvements in action. Even though the GTX 570 is over 20% faster than the GTX 470 on Crysis, power consumption has dropped by 5W. The comparison to the GTX 480 is even more remarkable at a 60W reduction in power consumption for the same level of performance, a number well in excess of the power savings from removing 2 GDDR5 memory chips. It’s quite remarkable how a bunch of minor changes can add up to so much.
On the flipside however we have the newly reinstated FurMark, which we can once again use thanks to W1zzard of TechPowerUp’s successful disabling of NVIDIA’s power throttling features on the GTX 500 series. We’ve rebenched the GTX 580 and GTX 580 SLI along with the GTX 570, and now have fully comparable numbers once more.
It’s quite interesting that while the GTX 570 does so well under Crysis, it does worse here than the GTX 470. The official TDP difference is in the GTX 470’s favor, but not to this degree. I put more stock in the Crysis numbers than the FurMark numbers, but it’s worth noting that the GTX 570 can at times be worse than the GTX 470. However also take note of the fact that the GTX 570 is still beating the GTX 480 by 33W, and we’ve already established the two have similar performance.
Of course the stand-outs here are AMD’s cards, which benefit from AMD’s small, more power efficient chip designs. The Radeon HD 5870 draws 50W+ less than the GTX 570 in all situations, and even the 6850 CF wavers between being slightly worse and slightly better than the GTX 570. With the architecture and process improvements the GTX 570’s power consumption is in line with historically similar cards, but NVIDIA still cannot top AMD on power efficiency.
Up next are idle temperatures. With nearly identical GPUs on an identical platform it should come as little surprise that the GTX 570 performs just like a GTX 580 here: 37C. The GTX 570 joins the ranks among some of our coolest idling high-end cards, and manages to edge out the 5870 while clobbering the SLI and CF cards.
With the same cooling apparatus as the GTX 580 and lower power consumption, we originally expected the GTX 570 to edge out the GTX 580 in all cases, but as it turns out it’s not so simple. When it comes to Crysist the 82C GTX 570 is 3C warmer than the GTX 580, bringing it in line with other cards like the GTX 285. More importantly however it’s 11-12C cooler than the GTX 480/470, driving home the importance of the card’s lower power consumption along with the vapor chamber cooler. Is does not end up being quite as competitive with the 5870 and SLI/CF cards however, as those other setups are between 0C-5C cooler.
As for FurMark, temperatures approach the mid-to-upper 80s, putting it in good company of most other high-end cards. The 5870 edges out the GTX 570 by 2C, while everything else is as warm or warmer. The GTX 480/470 in particular end up being 4C warmer, and as we’ll see a good bit louder.
It’s once we look at our noise data that it’s clear NVIDIA has been tinkering with the fan parameters of their design for the GTX 570, as the GTX 570 doesn’t quite match the GTX 580 here. When it comes to idle noise for example the GTX 570 manages to just about hit the noise floor of our rig, emitting 1.5dB less noise than the GTX 580/480/470. For the purposes of our testing, it’s effectively a silent card at idle in our test rig.
Under load, it’s finally apparent that NVIDIA has tweaked the GTX 570 to be quieter as opposed to cooler. For the slightly higher temperatures we saw earlier the GTX 570 is 2dB quieter than the GTX 580, 3.6dB quieter than the GTX 470, and 6.2dB quieter than the similarly performing GTX 480. For its performance level the GTX 570 effectively tops the charts here – the next quietest cards are the 5850 and 6870, a good pedigree to be compared to. On the other side of the chart we have the 5870 at 1.4dB louder, and the SLI/CF cards at anywhere between 1.2dB and 2.4dB louder.
If we took the conclusion from our GeForce GTX 580 article and replaced 580, 480, and 6870CF with 570, 470, and 6850CF respectively, our final thoughts would be almost identical. But then the GTX 580 and GTX 570 are almost identical too.
Whereas the GTX 580 took a two-tiered approach on raising the bar on GPU performance while simultaneously reducing power consumption, the GeForce GTX 570 takes a much more single-tracked approach. It is for all intents and purposes the new GTX 480, offering gaming performance virtually identical to the GTX 480 at a lower price, and with less power consumption along with lower temperatures and less noise. As a lower tier GF110 card the GTX 570 won’t wow the world with its performance, but like the GTX 580 it’s a solid step forward. In this case it’s a solid step towards bringing yesterday’s performance to the market at a lower price and with power/thermal/noise characteristics better suited for more systems. If nothing else, NVIDIA has translated the GTX 580’s excellent balance of performance and noise to a lower priced, lower performing tier.
Furthermore at $350 NVIDIA is the only game in town for single-GPU cards for the time being. Until an AMD competitor comes along NVIDIA has done a good job of filling the gap between the GTX 580 and GTX 470, an action very reminiscent of the GTX 470 and how it filled the gap between the Radeon HD 5870 and 5850 earlier this year. With no single card alternative on the market right now the only competition is the GeForce GTX 460 1GB SLI and the Radeon HD 6850 CF. The Radeon in particular should not be underestimated – it can trounce the GTX 570 almost at will – however it’s dogged by the fact that 6850 prices are running high right now, putting it at a $30+ price premium over the GTX 570. And of course both multi-GPU solutions face the usual caveats of uneven performance scaling, more noise, and a reliance on driver updates to unlock the 2nd GPU on new games. As with the GTX 580 we’d pick the simplicity of a single-GPU setup over the potential performance advantages of a multi-GPU setup, but this is as always a personal decision.
As a gap-filler the GTX 570 is largely what we expected the moment we saw the GTX 580 and we have no serious qualms with it. The one thing that does disappoint us is that NVIDIA is being conservative with the pricing: $350 is not aggressive pricing. The GTX 570 is fast enough to justify its position and the high-end card price premium, but at $100 over the GTX 470 and Radeon HD 5870 you’re paying a lot for that additional 20-25% in performance. Certainly we’re going to be happy campers if AMD’s next series of cards can put some pressure on NVIDIA here.
And finally, that brings us to AMD. AMD’s schedule calls for the Radeon HD 6900 series to be launched by the end of the year, and the year is quickly running out. There’s still too much uncertainty to advise holding off on any GTX 500 series purchases (particularly if you expect to have a card for Christmas), but if you’re not in a rush for a card it could be worth waiting a couple more weeks to see what AMD has up their sleeves. A holiday slugfest between AMD and NVIDIA and the resulting price drops are certainly at the top of our wish lists.