Nvidia GTX 970 | Nvidia Admits To Incorrect Specs & Memory Issues

nvidia-gtx-970-memory-spec-issuesA surprising turn of events from not only the top GPU manufacturer but probably one of the most reliable company’s in the hardware community. Nvidia always deliver a top of the line experience with every product from design to development and has created confidence with the consumer with all their highly successful and reliable products.

First it was mentioned that the GTX 970 could not utilize the full 4GB of its memory. Although this is kind-of true, it’s more that the GTX 970 appears to have a serious bug which would create issues with data allocation when of the card being utilized more and more. So if you start filling up memory and you a running on the last 500 MHz to 700 MHz is being accessed. So if you’re clocking 3.3GB to 3.5GB of usage with your GTX970, then the performance would drop for no reason. The problem does not just limit the card to the above values, but instead when the remaining memory is accessed, the card will manifest in stuttering, frame drops and other related performance issues.

One of Nvidias selling points for the GTX 970 & 980 is 4K gaming and having the necessary amount of VRAM available for such high demanding gaming. But if using the GTX 970 for 4K, you most likely will see these issues and will be apparent whilst running such high demanding games.

Nvidia have since commented on these issues, and it appears that the information appearing in initial reviews from Nvidia wasn’t accurate. The GTX 970 has both fewer rops and less level 2 cache than the GTX 980. Nvidia are protesting that it was a misunderstanding (between both the engineering team and the technical PR team that’s found its way into the reviewer’s guides). So, we’re left with 56 ROPS versus the advertised 64 ROPS and 1792 KB of L2 cache vs 2048.

Further, the GTX 970’s 4GB of memory was divided into two pools because of the reduced ROPs count: one 3.5GB section, and one 0.5GB section. This leads to a rather large performance difference in bandwidth, with the GPU not hitting the bandwidth speeds Nvidia promised on packaging.

It’s further confusing, because looking at the GPU-Z info, you’ll spot that the info reported by GPUZ is actually that of the review guides. This has lead to a Change.org campaign which reads as follows:

It was made official by Nvidia that they made a mistake in advertising. Whether or not they lied is irrelevant. Customers were misled by the factually incorrect specs listed on the packaging.

When using GPU-Z, the specs actually reflect this error which still incorrectly reads the card as having 64ROPs, ie. a cover-up.

It is believed that Nvidia should offer it’s customers a refund or at the very least acknowledgement and compensation, perhaps in the form of a “step-up” program, so the customer can choose what is best for them.

Keep checking here and we will keep you updated!