Wednesday, February 4th 2015
Specs Don't Matter: TechPowerUp Poll on GTX 970 Controversy
In the thick of the GeForce GTX 970 memory controversy, last Thursday (29/01), TechPowerUp asked its readers on its front-page poll, if the developments of the week affected the way they looked at the card. The results are in, and our readers gave a big thumbs-up to the card, despite the controversy surrounding its specs.
In one week since the poll went up, and at the time of writing, 7,312 readers cast their votes. A majority of 61.4 percent (4,486 votes) says that the specs of the GTX 970 don't matter, as long as they're getting the kind of performance on tap, for its $329.99 price. A sizable minority of 21.2 percent (1,553 votes) are unhappy with NVIDIA, and said they won't buy the GTX 970, because NVIDIA lied about its specs. 9.3 percent had no plans to buy the GTX 970 to begin with. Interestingly, only 5.1 percent of the respondents are fence-sitters, and waiting for things to clear up. What's even more interesting is that the lowest number of respondents, at 3 percent (219 votes), said that they're returning their GTX 970 cards on grounds of false-marketing. The poll data can be accessed here.
In one week since the poll went up, and at the time of writing, 7,312 readers cast their votes. A majority of 61.4 percent (4,486 votes) says that the specs of the GTX 970 don't matter, as long as they're getting the kind of performance on tap, for its $329.99 price. A sizable minority of 21.2 percent (1,553 votes) are unhappy with NVIDIA, and said they won't buy the GTX 970, because NVIDIA lied about its specs. 9.3 percent had no plans to buy the GTX 970 to begin with. Interestingly, only 5.1 percent of the respondents are fence-sitters, and waiting for things to clear up. What's even more interesting is that the lowest number of respondents, at 3 percent (219 votes), said that they're returning their GTX 970 cards on grounds of false-marketing. The poll data can be accessed here.
143 Comments on Specs Don't Matter: TechPowerUp Poll on GTX 970 Controversy
As W1zzard said...
For Who mencioned HDD... I know (maybe you also do) that 1000GB is a unformatted space. For standard...
But did we Know that 4GB Nvidia VRAM is 3.5GB faster + 0.5GB slower, and 64ROPs means 56, and etc...??
That is a "little" diference...
If rumors are true then it should have significant improvements in performance probably beating a GTX 980 by a fair bit. I don't think it will have low temps since it's rumored to be a 300 watt card.
Yes, they support 4 new features included in DX12, but those are not the only new features of DX12.
When NVIDIA released the card and boasted about DX12 compatibility, DirectX development head from MS said that there are no final conformance tests available for hardware manufacturers.
Of those 4 features, Intel supports at least 1, possible 2, 3 or 4. In fact, one of those features is nothing more than DX-version of Intel PixelSync (Raster Ordered Views).
Of those 4 features, GCN supports at least 1 (Raster Ordered Views), 1 very very likely (Volume Tiled Resources) and possibly the other 2, too.
So, your point about what I bothered reading is what? That I don't care reading about AMD problems? If this is the case, don't create a false image by specifically choosing only one line from my post and ignoring/deleting/downgrading the rest of it.
If this is not the case, ignore this comment and tell me what you really meant. :)
at 4k in SLI, there would be virtually no performance difference between a true 4gb card and the 3.5gb the 970 has.
Which all of them are not seeing is grist to the mill to nVidia. Talking about the lower middle class !!!!!!!!!!!!!!! failed!!!!!!!!!!! it failure!!! cut out processor !!!!!!!!!!!!! it can not be measured with AMD R 9 290X.
The criterion would only be GTX780 !! Is Not 20 nm, but 27 nm a new generation of processors that should be sufficient for 4K gaming .wich does not for all the games and anyway overestimated by 120$. all together gave a good advertisement for Nvidia.
And hiding the fact that we will set you up with TITAN X again the same error and failure cut processor for much more 1000$ at list :shadedshu:
This isn't the real problem, though. As it has been said before, suddenly finding out that the card's memory configuration isn't what nVidia stated it was doesn't change the performance numbers, not in the slightest. The real problem is that nVidia LIED about it. There's NO WAY the decision-makers at nVidia didn't know about all this since before the card was launched. It's not like they let TSMC figure out the specs for the batches of chips that would end up in GTX 970 cards.
Let's face it, if the marketing department would have gotten the wrong spec-sheet from the engineering teams about the GTX 970 specs, considering that the whole of the internet has reported those numbers, they would have sent out a press release the next day (absolutely worst case scenario, the reviews are closely monitored to a level that would make the NSA jealous), saying that there had been a mix-up, along with a PDF with the real specs. That didn't happen. This means that nVidia's marketing department were either knowingly lying through their teeth, or someone in charge had ordered the engineering teams to feed them false information.
There is absolutely NO WAY the engineers didn't know the correct specs for the GTX 970 and even if there was a mix-up somewhere down the line, somebody would have noticed it before launch day. Launch day press events held for both AMD and nVidia graphics cards usually contain all the technical specs the reviewers would ever be interested in knowing. That's just about all the technical details you would normally read in one of W1zzard's reviews. Then there's a Q&A session, where the press gets to ask whatever they want to ask about the product. Now, unless the marketing staff has a technical background (they changed from the engineering teams to the marketing department somewhere down the line, usually), they are pretty much technically illiterate. In such a case, they wouldn't be able to properly mount a graphics card inside a PC and all they're ever able to do is quote from the presentation they've just gone through. In such a case, though, they ALWAYS have an engineer with them. Said engineer handles the technical part of the briefing and, of course, the technically-oriented questions from the press. Whichever the case, two things stand out clearly: 1) the engineering and marketing teams do more than just exchange a single botched-up PDF and 2) there's absolutely NO WAY this was an honest mistake, one of those things that routinely get overlooked, like, say, the plastic shroud isn't black, but a very dark shade of grey.
Personally, I only care about this because they most clearly lied about this willingly. I don't buy or recommend graphics cards because they're from nVidia or AMD, the only thing that actually matters to me is what said card offers for the money (performance, noise, overclocking, reliability, good drivers, etc.). I've had cards from both companies over the years and I've tested hundreds more. I don't care who's caught lying, I can always buy from the other guy after all. But I do care when both the press and (implicitly) the consumers are being lied to so blatantly. Such an event should never be treated as casual by either side because it sets the worst kind of precedent possible: it sends the liar a signal that it's ok to continue lying and that they can easily get away with it.So, in your opinion, the fact that you got to buy nVidia's third best card (for its generation) for a price that used to get you a high-end flagship not so long ago is reason enough to overlook the fact that your favorite company intentionally lied to its customers? Fanboys never, ever cease to amaze...
At launch
GTX 480 $500
GTX 580 $500
GTX 780 $650
GTX 780Ti $700
My opinion is that if fugitive war criminals whose part-time hobby was spreading Ebola through orphanages decided to sell graphics cards at 30% of MSRP, people would get crushed in the stampede to buy them.
As many have pointed out, those of us who actually take vram usage in to account when purchasing a card have certainly got grounds to stand on here. We expected to be able to properly use 4GB of video memory and we're missing ~12%.
True, that doesn't make the benchmarks and reviews any less factual. The issue persisted then. Still, consumers who purchased that card with the intent of it performing well down the line when vram usage is higher will be disappointed.
As for price, you just need to go a little bit before the GTX 480. Namely, the GTX 285 had a 400$ price tag attached to it at launch and it was the single-GPU flagship of its generation.
Most people don't realise that higher memory bandwidth offer little difference given a set GPU performance, while increasing the GPU performance also increases the need for memory bandwidth. This means a higher performing GPU is more bottlenecked by memory bandwidth.
This is why GTX 970 is a more "balanced" GPU than GTX 980, and is also why GTX 980 in theory should be ~32% faster but in reality is only <15% faster.
And wtf? You've been a member 9+ years and posted 13 times? That must be a record you hermit - post more :D
Want a famous example? Radeon HD 3870 (320 VLIW5 shaders) versus Radeon HD 4870 (800 of the same VLIW5 shaders). The 4870 was about 56% more powerful, even though the shader count alone would indicate a massive jump in performance. And memory bandwidth had absolutely nothing to do with it.
@the54thvoid : Heh, long story, that part of the nickname actually has to do with the initials of my name. Not that hard to guess that I was rather young (although not 13) when I chose the nickname. And yeah, I've been around a really long time. I just rarely post. I enjoy the site and the community (gotta love those slug-fests between fanboys, eh?), I'm just rarely tempted to get involved.