Nvidia CEO is Mocked for Explanation of GeForce GTX 970 Memory Issue

[[Coi] Mr.CLuBBeR]

Here come the memes

Nividia ticked a lot of people off when it came to light that its GeForce GTX 970 graphics card was suffering from performance issues when games tried to access onboard memory above 3.5GB. Turns out it's the result of an architectural design, one that doesn't exist on the GTX 980, and one that wasn't communicated to Nvidia's internal marketing team or externally to reviewers. There's been a lot of negativity surrounding the issue ever since, and in attempt to diffuse the situation, Nvidia CEO Jen-Hsun Huang has offered up an explanation of the GTX 970 memory issue.
Before we get into that, we suggest reading this, this, and this as primers to what's going on. If you're crunched for time, the Cliff Notes version is that the above scenario, along with the discovery that the GTX has less ROPs and L2 cache as advertised, has led to a class action lawsuit.
Seeing that the contempt is growing bigger, not smaller, Huang tried explaining away the issue as a "feature" that should have been bragged about.
"We invented a new memory architecture in Maxwell. This new capability was created so that reduced-configurations of Maxwell can have a larger framebuffer – i.e., so that GTX 970 is not limited to 3GB, and can have an additional 1GB," Huang stated in a blog post.
"GTX 970 is a 4GB card. However, the upper 512MB of the additional 1GB is segmented and has reduced bandwidth. This is a good design because we were able to add an additional 1GB for GTX 970 and our software engineers can keep less frequently used data in the 512MB segment," the CEO continued.
According to Huang, the expectation was that users would be "excited" about this, but were ultimately "disappointed that we didn't better describe the segmented nature of the architecture." He also admitted that the "new feature of Maxwell should have been clearly detailed from the beginning."
Perhaps so, though looking at the comments to his blog post makes me think this was a ticking time bomb no matter how you slice it. For those holding Nvidia's feet to the fire over this, the bottom line here is that the GTX 970 is gimped compared to the GTX 980, which doesn't have an issue accessing all 4GB of VRAM, and that they were misled, both by the impact this would have and by the advertised specs.
"Yes, a 'new feature,' a 'good design' not included on GTX 980 because [it] decreases performance," a reader commented. Another stated, "I will most likely never buy from Nvidia again, they care nothing about their customer. And blatantly lie to our faces."
Others took to posting memes and doctored videos, like this one:

It's hard to watch the above clip without busting a gut, though for Nvidia, this is no laughing matter. To Nvidia's credit, the performance issue seems to only crop up when gaming at high resolutions and shouldn't bother folks gaming at 1080p. And based on the benchmarks when the performance issue doesn't creep up, the bang-for-buck here is pretty high.
But in the end, Nvidia is finding out that none of that matters, as their fan base feels it's been lied to. It's going to take more than a blog post to win back their trust and/or make this go away.
Follow Paul on Google+, Twitter, and Facebook


More...