NVIDIA GeForce GTX 670 Review Feat. EVGA: Bringing GK104 Down To $400
by Ryan Smith on May 10, 2012 9:00 AM ESTIn a typical high-end GPU launch we’ll see the process take place in phases over a couple of months if not longer. The new GPU will be launched in the form of one or two single-GPU cards, with additional cards coming to market in the following months and culminating in the launch of a dual-GPU behemoth. This is the typical process as it allows manufacturers and board partners time to increase production, stockpile chips, and work on custom designs.
But this year things aren’t so typical. GK104 wasn’t the typical high-end GPU from NVIDIA, and neither it seems is there anything typical about its launch.
NVIDIA has not been wasting any time in getting their complete GK104 based product lineup out the door. Just 6 weeks after the launch of the GeForce GTX 680, NVIDIA launched the GeForce GTX 690, their dual-GK104 monster. Now only a week after that NVIDIA is at it again, launching the GK104 based GeForce GTX 670 this morning.
Like its predecessors, GTX 670 will fill in the obligatory role as a cheaper, slower, and less power-hungry version of NVIDIA’s leading video card. This is a process that allows NVIDIA to not only put otherwise underperforming GPUs to use, but to satisfy buyers at lower price points at the same time. Throughout this entire process the trick to successfully launching any second-tier card is to try to balance performance, prices, and yields, and as we’ll see NVIDIA has managed to turn all of the knobs just right to launch a very strong product.
| GTX 680 | GTX 670 | GTX 580 | GTX 570 | |
| Stream Processors | 1536 | 1344 | 512 | 480 |
| Texture Units | 128 | 112 | 64 | 60 |
| ROPs | 32 | 32 | 48 | 40 |
| Core Clock | 1006MHz | 915MHz | 772MHz | 732MHz |
| Shader Clock | N/A | N/A | 1544MHz | 1464MHz |
| Boost Clock | 1058MHz | 980MHz | N/A | N/A |
| Memory Clock | 6.008GHz GDDR5 | 6.008GHz GDDR5 | 4.008GHz GDDR5 | 3.8GHz GDDR5 |
| Memory Bus Width | 256-bit | 256-bit | 384-bit | 320-bit |
| VRAM | 2GB | 2GB | 1.5GB | 1.25GB |
| FP64 | 1/24 FP32 | 1/24 FP32 | 1/8 FP32 | 1/8 FP32 |
| TDP | 195W | 170W | 244W | 219W |
| Transistor Count | 3.5B | 3.5B | 3B | 3B |
| Manufacturing Process | TSMC 28nm | TSMC 28nm | TSMC 40nm | TSMC 40nm |
| Launch Price | $499 | $399 | $499 | $349 |
Like GeForce GTX 680, GeForce GTX 670 is based on NVIDIA’s GK104 GPU. So we’re looking at the same Kepler design and the same Kepler features, just at a lower level of performance. As always the difference is that since this is a second-tier card, NVIDIA is achieving that by harvesting otherwise defective GPUs.
In a very unusual move for NVIDIA, for GTX 670 they’re disabling one of the eight SMXes on GK104 and lowering the core clock a bit, and that’s it. GTX 670 will ship with 7 active SMXes, all 32 of GK104’s ROPs, and all 4 GDDR5 memory controllers. Typically we’d see NVIDIA hit every aspect of the GPU at once in order to create a larger performance gap and to maximize the number of GPUs they can harvest – such as with the GTX 570 and its 15 SMs & 40 ROPs – but not in this case.
Meanwhile clockspeeds turn out to be equally interesting. Officially, both the base clock and the boost clock are a fair bit lower than GTX 680. GTX 670 will ship at 915MHz for the base clock and 980MHz for the boost clock, which is 91MHz (9%) and 78MHz (7%) lower than the GTX 680 respectively. However as we’ve seen with GTX 680 GK104 will spend most of its time boosting and not necessarily just at the official boost clock. Taken altogether, depending on the game and the specific GPU GTX 670 has the capability to boost within 40MHz or so of GTX 680, or about 3.5% of the clockspeed of its more powerful sibling.
As for the memory subsystem, like the ROPs they have not been touched at all. GTX 670 will ship at the same 6.008GHz memory clockspeed of GTX 680 with the same 256-bit memory bus, giving it the same 192GB/sec of memory bandwidth. This is particularly interesting as NVIDIA has always turned down their memory clocks in the past, and typically taken out a memory controller/ROP combination in the past. Given that GK104 is an xx4 GPU rather than a full successor to GF110 and its 48 ROPs, it would seem that NVIDIA is concerned about their ROP and memory performance and will not sacrifice performance there for GTX 670.
Taken altogether, this means at base clocks GTX 670 has 100% of the memory bandwidth, 91% of the ROP performance, and 80% of the shader performance of GTX 680. This puts GTX 670’s specs notably closer to GTX 680 than GTX 570 was to GTX 580, or GTX 470 before it. In order words the GTX 670 won’t trail the GTX 680 by as much as the GTX 570 trailed the GTX 580 – or conversely the GTX 680 won’t have quite the same lead as the GTX 580 did.
As for power consumption, the gap between the two is going to be about the same as we saw between the GTX 580 and GTX 570. The official TDP of the GT 670 is 170W, 25W lower than the GTX 680. Unofficially, NVIDIA’s GPU Boost power target for GTX 670 is 141W, 29W lower than the GTX 680. Thus like the GTX 680 the GTX 670 has the lowest TDP for a part of its class that we’ve seen out of NVIDIA in quite some time.
Moving on, unlike the GTX 680 launch NVIDIA is letting their partners customize right off the bat. GTX 670 will launch with a mix of reference, semi-custom, and fully custom designs with a range of coolers, clockspeeds, and prices. There are a number of cards to cover over the coming weeks, but today we’ll be looking at EVGA’s GeForce GTX 670 Superclocked alongside our reference GTX 670.
As we’ve typically seen in the past, custom cards tend to appear when GPU manufacturers and their board partners feel more comfortable about GPU availability and this launch is no different. The GTX 670 launch is being helped by the fact that NVIDIA has had an additional 7 weeks to collect suitable GPUs compared to the GTX 680 launch, on top of the fact that these are harvested GPUs. With that said NVIDIA is still in the same situation they were in last week with the launch of the GTX 690: they already can’t keep GK104 in stock.
_thumb.jpg)
_(G57-V100950)_thumb.jpg)
Due to binning GTX 670 isn’t drawn from GTX 680 inventory, so it’s not a matter of these parts coming out of the same pool, but realistically we don’t expect NVIDIA to be able to keep GTX 670 in stock any better than they can GTX 680. The best case scenario is that GTX 680 supplies improve as some demand shifts down to the GTX 670. In other words Auto-Notify is going to continue to be the best way to get a GTX 600 series card.
Finally, let’s talk pricing. If you were expecting GTX 570 pricing for GTX 670 you’re going to come away disappointed. Because NVIDIA is designing GTX 670 to perform closer to GTX 680 than with past video cards they’re also setting the prices higher. GTX 670 will have an MSRP of $399 ($50 higher than GTX 570 at launch), with custom cards going for higher yet. This should dampen demand some, but we don’t expect it will be enough.
Given its $399 MSRP, the GTX 670 will primarily be competing with the $399 Radeon HD 7950. However from a performance perspective the $479 7970 will also be close competition depending on the game at hand. AMD’s Three For Free promo has finally gone live, so they’re countering NVIDIA in part based on the inclusion of Deus Ex, Nexuiz, and DiRT Showdown with most 7900 series cards.
Below that we have AMD’s Radeon HD 7870 at $350, while the GTX 570 will be NVIDIA’s next card down at around $299. The fact that NVIDIA is even bothering to mention the GTX 570 is an interesting move, since it means they expect it to remain as part of their product stack for some time yet.
Update 5/11: NVIDIA said GTX 670 supply would be better than GTX 680 and it looks like they were right. As of this writing Newegg still has 5 of 7 models still in stock, which is far better than the GTX 680 and GTX 690 launches. We're glad to see that NVIDIA is finally able to keep a GTX 600 series card in stock, particularly a higher volume part like GTX 670.
| Spring 2012 GPU Pricing Comparison | |||||
| AMD | Price | NVIDIA | |||
| $999 | GeForce GTX 690 | ||||
| $499 | GeForce GTX 680 | ||||
| Radeon HD 7970 | $479 | ||||
| Radeon HD 7950 | $399 | GeForce GTX 670 | |||
| Radeon HD 7870 | $349 | ||||
| $299 | GeForce GTX 570 | ||||
| Radeon HD 7850 | $249 | ||||
| $199 | GeForce GTX 560 Ti | ||||
| $169 | GeForce GTX 560 | ||||
| Radeon HD 7770 | $139 | ||||
Facebook
Twitter
RSS
414 Comments
View All Comments
versesuvius - Friday, May 11, 2012 - link
Hey, what is your problem. I acknowledged general better frame rates of 680 over 7970 up front. And I am not talking about SLI setups either. It is single card performance which AMD delivers superior to NVIDIA over the years. Yes, NVIDIA has a way with video game engine makers. It has been like that for as long as anyone can remember.I said I am not going to insisting on NVIDIA skipping frames, but I am going to insist that you read the post correctly. Skipping frames is different from frame rate. A card can render as many frames as it likes and skip as many as it likes. A card that renders all frames is the honest one, while the card that skips is the frame rate can be the frame rate champ.
I am not a fanboy of anything. However, since you are so keen on numbers, AMD is the king of compute which is by far more important at this point in the evolution of the GPU than the few extra frames that NVIDIA can churn out. AMD should definitely keep a hefty price premium over NVIDIA cards for that reason alone. In other words it is up to NVIDIA how low it wants to go.
CeriseCogburn - Friday, May 11, 2012 - link
Go to Tech Report, they do a render time count on frames, which indicates according to theory, which cards offer "smooth" frame rates in which games.Your "feeling" might have something to do with that, although by your last post I'll go even more with amd fanboy for you now, than I had thought before.
If "smoothness" is your problem, and hence you "feel" nVidia is cheating, you perhaps should get the 301.24 driver and then use it's CP adaptive v-sync ( on non 6000 cards back to 8 series) or less so possibly the prefer full power setting. Beyond that use the EVGA precision frame rate target.
On the other hand, you think nVidia is cheating, so don't do any of the above and go with that thought. I'm sure no one has discovered nVidia skipping frames (roll eyes)
Furthermore, there is a new tech designed by the Israeli company called LucidLogix for z68 sandy bridge boards that allows one to use their discrete GPU card in conjunction with the onboard SB gpu core (HD3000) outputting the frames, and it uses an algorithm software and a certain cache/computing portion of the SB gpu to eliminate extra frames and parts of frames that would not be fully rendered and be cut off by the refresh rate - reducing or eliminating tearing for instance in frame rates higher than the monitors resolution -
It COUNTS frames it throws away - resulting in a higher on screen FPS score, smoother gameplay, and a higher frame rate overall, although not as high as it's counter claims. So THAT would be something akin to your suspicious amd fanboy complaint, but it works on amd and nvidia alike.
That would be very nice, BTW, if amd "kept" a hefty price premium over nVidia, but first they'd have to currently be priced higher - and since they are not because they suck in comparison, good luck with that. :)
If you need your dream fulfilled, go check the Cuda and Stream high end Commercial rendering cards and maybe you'll have a glow of red pride.
CeriseCogburn - Friday, May 11, 2012 - link
It's like your idiotic "feeling" that people who need high end GPGPU cards are going to buy the stupidly cheap and driverless gaming card by amd called the 7970 or the 7950.The COMPUTE CARDS cost THOUSANDS of dollars each.
You people have turned into drooling idiots repeating dummy of world talking points.
I mean the absolute stupidity is really irritating.
HERE PLEASE SAVE US FROM FUTURE STUPIDITY !
http://www.newegg.com/Product/Product.aspx?Item=N8...
That's one, ONE ONLY, amd HD79xx in COMPUTE CARD form - it's Was: $2,149.99
Now: $2,139.99
Save: $10.00
plus shipping....of $7.87
OK, well whoop die doo - you won't be gaming with, and the cards here reviewed WON'T be COMPUTING LIKE IT.
versesuvius - Friday, May 11, 2012 - link
You have failed miserably. I can assure you of that. That is, by displaying fairly and squarely in a public forum that you do not know anything about GPU computing and when you are wised up to it, you go ahead and sample workstation graphic cards. NVIDIA has that class of cards, too. It is called Quadro. Both Firepro and Quadro brands where there long before GPU computing took off, and both brands where equally proportionally priced far higher above the mainstream gaming cards. Those cards use the same chips that mainstream cards too, with wider io capabilities and most important of all highly tuned drivers. And surprise, surprise, the compute prowess of a Firepro or Quadro card is the same as the gaming mainstream cards . It is in the control that those cards exercise on what gets out of the card, that sets them apart. So, there you go. Blow off all the air you can. It is not going to change anything. You have no idea about what you are talking about.CeriseCogburn - Friday, May 11, 2012 - link
You didn't say anything I didn't already say, you idiot.versesuvius - Friday, May 11, 2012 - link
You fail again, you miserable know nothing. Just because you talk too much, way way over your coupon, doesn't mean you have said anything of significance. What imbecile would cite Firepro as a computing GPU! Oh, you just did, and even fail knowing what you, your very self, said. GPU computing with Firepro or Quadro. That is stupid. How did you come up with that?CeriseCogburn - Friday, May 11, 2012 - link
" Anyway, 7970 is a superior product over 680 if only for the compute performance it offers. "The 680 whomped the 7970 winning 3 compute benches over your superior if only for compute amd loser card.
Congratulations on another great teaching lesson for all of us, boy you are really on top of things.
CeriseCogburn - Friday, May 11, 2012 - link
Firepro is a compute GPU you moron.versesuvius - Saturday, May 12, 2012 - link
You fail again, you know nothing imbecile.All modern GPUs come with compute capabilities. You fail again in reading and absorbing basic facts. You fail in reading and remembering what you have said on this very thread and yet expect to be taken seriously.
Read up thread once again. Firepro and Quadro use the same chips that are used in mainstream, consumer cards. It is the drivers and outside of the chip, i.e. on the board, where the differences emerge. 7970 is far, far superior to NVIDIA for HPC and compute heavy applications. 580 is a lot better than 680 when it comes to GPU computing. That is what Anandtech called " What is left behind" with the 680. 680 is more power efficient than 7970 by about 10% on load. On idle power efficiency is about the same, with 7970 even better. Yet 7970 is about twice faster than NVIDIA. That means having half the number of 7900 chips in an HPC application will achieve the same result at half the cost. And since 580 is discontinued, NVIDIA is in a lot of trouble. They will have to rely on 570 chips for any competition with 7900 chips. That is NVIDIA's crowning achievement in the retro computing department.
Frankly, HPC is not the domain of gamers, and I am not surprised that you are totally ignorant about it.
CeriseCogburn - Saturday, May 12, 2012 - link
Well you have no point just more amd fanboy rhetoric, and you've been wrong how many times in a row now ?Let's just add up your number of posts here and we'll give you a big break and count one "WRONG!" for each of your posts.
I'm (not) looking forward to you explaining the 7970 loser status on the benches compute page in the review, because you won't, you can't and amd lost, and you're sad and pissed and frankly you don't understand why, obviously, ever lessening is the outside chance you could actually bring yourself to admit it's loss.
Have a nice day fella.