Gamers Petition for GeForce GTX 970 Refund Over Error in Specs

[[Coi] Mr.CLuBBeR]

Internal miscommunication at Nvidia led to confusion over the GTX 970's specs

Sometimes the tech world can be like a geek version of a soap opera, and this is one of those times. The main characters in this case are Nvidia and the GeForce GTX 970. If you're looking for a quick summary of events, it's this: Gamers noticed a slowdown in performance when games tried to access more than 3.5GB of memory on the GTX 970. This in turn led to Nvidia explaining a new memory architecture in the GTX 970, along with clarification of specs that were different than originally reported. In light of all this, there's a petition floating around demanding a refund for anyone who purchased a GTX 970, but to really understand what's going on, a deeper explanation is necessary.
This all began a week ago when users on various forums began investigation a memory issue with the GTX 970. At a glance, it seemed that the card was only using 3.5GB of its 4GB of GDDR5 memory. Upon closer look, it was discovered that a serious performance drop could occur when accessing that final .5GB of VRAM, which isn't an issue on the GTX 980.
To clarify what was happening, Nvidia issued the following statement:
"The GeForce GTX 970 is equipped with 4GB of dedicated graphics memory. However the 970 has a different configuration of SMs than the 980, and fewer crossbar resources to the memory system," Nvidia said. "To optimally manage memory traffic in this configuration, we segment graphics memory into a 3.5GB section and a 0.5GB section. The GPU has higher priority access to the 3.5GB section. When a game needs less than 3.5GB of video memory per draw command then it will only access the first partition, and 3rd party applications that measure memory usage will report 3.5GB of memory in use on GTX 970, but may report more for GTX 980 if there is more memory used by other commands. When a game requires more than 3.5GB of memory then we use both segments.
"We understand there have been some questions about how the GTX 970 will perform when it accesses the 0.5GB memory segment. The best way to test that is to look at game performance. Compare a GTX 980 to a 970 on a game that uses less than 3.5GB. Then turn up the settings so the game needs more than 3.5GB and compare 980 and 970 performance again."
Nvidia Senior VP of GPU Engineering, Jonah Alben, spoke with PC Perspective and broke things down even further with a quite a few technical details. He also offered a helpful diagram, seen below.

As you can see in the graph, there are 13 enabled SMMs, each with 128 CUDA cores for a total of 1,664. There are also three that are grayed out -- they've been disabled from the full GM204 found on the GTX 980. But what's really important is the memory system, which is connected to the SMMs through a crossbar interface.
"That interface has 8 total ports to connect to collections of L2 cache and memory controllers, all of which are utilized in a GTX 980. With a GTX 970 though, only 7 of those ports are enabled, taking one of the combination L2 cache / ROP units along with it. However, the 32-bit memory controller segment remains," PC Perspective writes.
There are a couple of takeaways there. First is the GTX 970 has less ROPs and L2 cache than the GTX 980 even though it was reported otherwise. Why? Nvidia blames the gaffe on an error in the reviewer's guide, which is usually a PDF (or actual paper) containing detailed info on a product prior to its launch that manufacturers send out to reviewers, and a misunderstanding between the engineering team and the technical PR team on how the architecture actually functioned.
Bottom line is, the GTX 970 has 56 ROPs and 1,792KB of L2 cache instead of 64 ROPs and 2,048KB of L2 cache like the GTX 980.
That's actually not as big of a deal as it sounds, as the SMMs are the true bottleneck, not the ROPs.
"A quick note about the GTX 980 here: it uses a 1KB memory access stride to walk across the memory bus from left to right, able to hit all 4GB in this capacity," PC Perspective writes. "But the GTX 970 and its altered design has to do things differently. If you walked across the memory interface in the exact same way, over the same 4GB capacity, the 7th crossbar port would tend to always get twice as many requests as the other port (because it has two memories attached). In the short term that could be ok due to queuing in the memory path. But in the long term if the 7th port is fully busy, and is getting twice as many requests as the other port, then the other six must be only half busy, to match with the 2:1 ratio. So the overall bandwidth would be roughly half of peak. This would cause dramatic underutilization and would prevent optimal performance and efficiency for the GPU."
There are a LOT more details to digest, and rather than continue to quote bits and pieces, we suggest you read PC Perspective's detailed report. If after doing so you come to the conclusion that it's much ado about nothing, great, there's nothing more to see here. However, if you fall on the other side of the fence and feel duped, you can check out and sign the petition at Change.org.
Our take? It's an unfortunate situation Nvidia created for itself, and gamers have a right to be angry over the misreported specs. At the same time, it appears that the impact on real-world performance is negligible, at least for now -- this could be a bigger issue as higher resolution game play becomes more common. Even still, it remains a great card for the price.
Follow Paul on Google+, Twitter, and Facebook


More...