The Real Nvidia GeForce GTX 970 Specifications

We’ve remained quiet on this issue until now because we wanted to be absolutely clear about the situation before weighing in. I love to use automotive metaphors, so I’m going to apply one here to describe this situation because I think it helps put the whole mess in perspective:

You’re a muscle-car buff and you decide to test drive the new 2015 Dodge Charger Hellcat. The car is advertised as a supercharged 8-cylinder, 6.2 Liter Hemi engine with 24 valves that produces 707 horsepower at 6,000 RPM. It’s one of the most powerful cars you can buy for the dollar, achieving 0-60 MPH in under three seconds and a quarter-mile in under 12 seconds. You take it for a test drive, you fall in love with the car, and you buy it. In the months to follow, you remain quite pleased with your purchase and the performance the car provides.

It later comes out that Dodge made a mistake on its marketing materials: the engine has 16 valves, not 24. It still produces 707 horsepower at 6,000 RPM though, and it still offers the same amazing road performance that it did the day you bought it. It’s still one of the fastest cars you could purchase for the dollar. But you can no longer say you own a 24-valve V8.

It’s upsetting. But does it make the Charger Hellcat a worse car that it was before you found out? Practically, no, as performance is unchanged. But it leaves a bad taste in your mouth. For Dodge, it would be a PR nightmare. Some Charger Hellcat owners would feel lied to, despite the fact that their car is every bit as fast as they expected it to be in the first place.

This is essentially the kind of misrepresentation scenario that Nvidia is dealing with. Introducing the actual GeForce GTX 970 specifications:


GeForce
GTX 980
GeForce GTX 970
(actual)
GeForce GTX 970
(originally reported)
GPU
GM204 (Maxwell)
GM204 (Maxwell)GM204 (Maxwell)
Process
28nm
28nm28nm
Shader Units
2048
1664
1664
Texture Units
128
104
104
ROPs
64
56
64
L2 Cache
2 MB
1.75 MB
2 MB
Core CLock
1126 MHz
1050 MHz
1050 MHz
Memory Clock
1750 MHz GDDR5
1750 MHz GDDR51750 MHz GDDR5
Memory Bus
256-bit
256-bit256-bit
Memory Bandwidth
224 GB/s
196 GB/s (3.5 GB)
28 GB/s (512MB)

224 GB/s
Memory Capacity
4GB
4GB
4GB
Max. TDP
165 Watts
145 Watts
145 Watts
Aux. Power
Connectors
2x Six-pin PCIe
2x Six-pin PCIe2x Six-pin PCIe

This issue reared its head when some users noted that in certain cases the GeForce GTX 970 reported 3.5GB of graphics memory despite being sold as a 4GB card. It turns out that this is a symptom of memory segmentation, as the card splits the 4GB into a 3.5GB high-priority segment and a 512MB low-priority segment. Yes, the card has access to a full 4GB of RAM. It is accessed differently and with different theoretical bandwidth, but it's all there and available. It remains to be seen if this technique causes any notable performance detriment in a real-world scenario, but from what we've seen so far, it does not.

But why is there an odd split in memory resources in the first place? It turns out, this is a clue that hints at the real issue: Nvidia’s incorrect reporting of the GeForce GTX 970’s technical specifications and GM204 GPU resources.

Note that one of the four ROP partitions is not fully enabled, but partially disabled. That partially disabled partition is responsible for the strange 3.5GB/512MB memory split, as memory controller resources are linked through them. With one-eighth of the ROP partitions disabled, one-eighth of the memory (512MB) must be accessed in a special way, through the working half of the partition. In this way, all 4 GB of RAM is usable, although segmented. As a result the 3.5 GB portion can be accessed with 196 GB/s of bandwidth, while the 512 MB portion has 28 GB/s of bandwidth available. There are probably some particular situations that will expose this weakness, but it is surprisingly difficult to create this scenario. We will keep an eye out for it and be able to call it out now that we know the actual specifications, but none of this new information invalidates the benchmark results we've already collected.

In addition, an eighth of the L2 cache is not used. As a result, there are 56 functional ROPs in the GeForce GTX 970, and the chip has access to 1.75 MB of L2 cache. This is less than the 64 ROPs and 2 MB of L2 that Nvidia originally indicated in its press materials.

How Did This Happen?

The company claims this is the result of a misunderstanding on the part of the technical marketing team, which was not aware of the partially-enabled ROP cluster, and that the issue was not identified internally until this month when people began digging into the memory issue.

It’s times like these that conspiracy theories begin to fly. In this case it’s understandable: how does a highly technical team of GPU experts, one that works directly for the chipmaker, make such a huge mistake?

I’ve never been a conspiracy theorist: I believe human greed and stupidity are simple explanations for all that’s wrong in the world, and that we don’t need to concoct a shadowy Illuminati organization to take the blame. I also don’t see a lot of motive for Nvidia to open themselves up to this kind of PR nightmare in the first place. There is absolutely no advantage or benefit to have done so. Occam's razor points to a screw-up, plain and simple.

It does surprise me that one of Nvidia's engineers never spotted this major snafu after reading a few launch reviews. But if you were that guy, would you even want to report that a huge product launch suffered from flubbed specs? If you were made aware of this inconsistency, would you offer the information to the world after a successful launch was behind you, or keep it to yourself and hope that the press would never find out?

Does It Matter?

I’m not omniscient. I don’t know if Nvidia knew about this and chose to keep it close to its chest, or if the company found out with the rest of us last Friday. But I have no good reason to believe the company is lying. To tell the truth, from a practical standpoint, I’m not sure if it matters.

That’s not to say I think it’s OK to be fed misinformation about GPU specifications. This is a subject I am patently passionate about. I think it’s important for the technical press to have the right information at its disposal when analyzing hardware and deciphering the repercussions of what goes on under the hood. Nvidia needs to work hard to make sure this kind of mistake never happens again.

But from a purely practical standpoint, this doesn’t really change anything for the end user. The GeForce GTX 970 remains one of the best graphics card buys on the market. It performs the same way it did at launch — which is really good. As such, we will continue to recommend it until there is a better-performing option for the price.

We can empathize with buyers who feel betrayed, though. Nvidia definitely has some mind-share to earn back. But to us the price/performance ratio trumps everything else, and that is no different today than it has been since the GeForce GTX 970 was released.

Create a new thread in the US News comments forum about this subject
This thread is closed for comments
134 comments
Comment from the forums
    Your comment
    Top Comments
  • ern88
    It's false advertising plain and simple!!!
    48
  • tpi2007
    Don, you forgot to not only mention but also correct in the table that the card's memory bandwidth is not 224 GB/s, but rather 196 GB/s for the 3.5 GB segment and a max. of 28 GB/s for the second segment. And it's likely less than 28 GB for the second segment as Nvidia has said the smaller resources available to access it make it slower. Oh, and the fact that you can't use both segments at the same time.

    So, also using a car analogy, it's like saying your car can do 196 mph in fifth gear and 28 in first gear, therefore your car's top speed is 224 mph. Everybody knows you can't do this math, but somehow some sites are letting Nvidia have a free pass on this item.

    And yes, it does matter, because it changes the value proposition when making a buying decision. People thought they were getting an awesome deal because the card only had less CUDA cores, but the whole rest intact to help it out in the future. Now it performs well, but what about in one year with new games using more VRAM coming out each month ?

    Also, don't forget the importance of PR: what were the real effects of announcing a card with the same number of ROPs as AMDs counterparts that had been available for a year ? The message: this card is well prepared for high resolution rendering. And the full complement of L2 to help ease the concerns about a 256-bit memory bus and apparently low memory bandwidth (compared to AMD's offerings). Remember that part of the PR was emphasizing Delta Colour compression, which practically nobody had heard about until then, but that was in its third iteration already. This time they gave it a highlight, which indicates they wanted to transmit a message. Saying that the card has the full complement of L2 cache, the same as the GTX 980, would also fit that message.

    Was it an honest communication mistake ? I don't know. But I do know that it does matter and it does affect the value proposition of the card in the long term (of a card's useful lifetime, that is).


    Edit:

    Don, I see that you have now corrected the table (it's good practice to note article changes, which you didn't as of this edit) to read:

    224 GB/s aggregate
    196 GB/s (3.5 GB)
    28 GB/s (512MB)

    But this is still incorrect. As Nvidia itself admitted, both segments can't be used at the same time, so you cannot therefore add the two bandwidth numbers. It's one OR the other at any given time. Anandtech (now your sister site) has an article saying exactly this. Saying "224 GB/s aggregate" is at the very least misleading.

    I think that at the end of all this misleading situation, reporters should be the first to be accurate.

    [Answer By Cleeve]
    Fair enough, I've removed the 'aggregate' spec.
    Keep in mind it's general practice to describe dual-GPU cards as having a 512-bit aggregate bus when each GPU really mirrors a 256-bit bus, so I considered it in the same vein. But honestly I've never liked that practice myself, so I'm quite OK with dumping it.

    As for your (and others) concern with the car analogy, I stand behind it. Real-world measured performance is the metric that will always matter most to me. In the case of a car that's the quarter mile, and in the case of a graphics card, it's frames-per-second. The benchmarks we measured in frames-per-second have not been changed by this revelation, or rendered any less accurate, so I'm going to have to agree to disagree with you as to the merit of this metaphor.

    As far as the merits of PR, I think you might be overestimating that vs. raw performance. Nvidia has released other asymetrical cards in the past and I never got the impression that the public boycotted them because of it. If they *were* avoided, it was because they were slower than the Radeon competition, plain and simple. It should come down to frames-per-second.

    But once again, everyone is free to disagree and their own opinion. I'm simply calling it as I believe it to be. I would probably feel different if the company had a history of lying about technical specifications, but I can't recall something similar in the last 20 years or so. I *can* recall them owning up to other, strange memory configurations with similar limitations, so it doesn't seem logical to assume they decided to blatantly lie this time around when they previously came clean.

    But who knows? Regardless of what any of us believe, by design or by accident, Nvidia has tremendous mindshare to earn back if it wants the public trust. This kind of mistake should be taken very seriously. If it ever happens again, I don't think anyone would believe that it's an accident.

    But to my mind, that doesn't affect the GTX 970's proven performance, nor make it any less desirable for the money. If you feel it does, more power to you. Your opinion is as valid as my own as long as you have valid reasons to justify it.
    40
  • DCNemesis
    I think that this author's article is ignoring the issue, and not appropriately focused on the facts regarding of this omission of information.

    As soon as you start gaming at 1440p or 4k, this segmented RAM issue *WILL* affect performance.

    Any game that uses large texture files (read Skyrim mods, etc) will start to approach and exceed the prioritized 3.5gb threshold. As soon as the VRAM usage exceeds that 3.5gb, the remaining 512mb is 1/8th the speed of the primary segment.

    This is a serious misrepresentation of the product on NVidia's part. This article is much less objective, impartial, and scrutinizing than I would expect from this website and its reviews and authors.

    The car analogy is inaccurate and the NVidia shill is inappropriate.

    I expected more from both NVidia and Tomshardware.

    Too bad.
    23
  • Other Comments
  • damric
    Maybe Nvidia will make this right by giving early GTX 970 adopters $10 off on a new Titan.
    1
  • ern88
    It's false advertising plain and simple!!!
    48
  • TeamColeINC
    I read that the last 500MB of VRAM actually runs significantly slower than the first 3.5GB, bringing the actual memory bandwidth just under 200GB/s (when all VRAM is being utilized)
    11