GTX 970 flaw

Status
Not open for further replies.
GTX 970 Reviewer's Guide: Specifications
970RG_575px.png


Official site
tWLGq7p.png


Where does it say up to 64 ROP units, up to 2048K L2 cache, up to 256-bit memory interface or up to 224 GB/s memory bandwidth?

I was talking about your bad analogy about ssds.
 
The larger portion is faster. The problem is they've advertised the card as having 224 GB/s 4 GB memory, when in fact it has a 196 GB/s 3.5 GB section and a 28 GB/s 0.5 GB one. It can only read or write to one of them at the same time, not to the whole 4 GB. If the game reads from the whole memory range, total combined bandwidth goes somewhere in between 28 and 196 and stutters appear because of the huge speed discrepancy between different memory sections.

Oh I wasn't talking about the 970 when I was talking about which portion was larger, I was talking about my SSD example. Some confusion both ways lol.


I always see in the specs that it says up to. They do not claim you will always get certain speeds.
I was talking about your bad analogy about ssds.

Some reading issues. Wasn't his analogy, that was mine. Second of all the only way it's really "bad" is I guess it's true that SSD advertisements always have the words "up to". If you take that part out I think it represents the extremes of this type of thinking fairly well. You can bet your butt that if Nvidia is allowed to get away with this misrepresentation, we'll only see more of this crap in the future.
 
For those saying nvidia didn't falsely state the specs or provide them through channels to advertise to consumers:

CiclosSanos said:
First sorry about my english.

Well, we know that Nvidia and the manufacturers are doing everything possible to hide the false advertising. I found something that can be useful now to report it.

Manufacturers and Nvidia are telling us that there is no information about Rops and L2 Cache on their websites.

NVIDIA.

Well, if you visit this page: http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-970/reviews You can see the official reviews approved by Nvidia.

We click on the review of TomsHardware for example : http://www.tomshardware.com/reviews/nvidia-geforce-gtx-980-970-maxwell,3941.html
As we see in the specifications we can see about Rops. It says our graphic card has 64 ROPS.

If we see another APPROVED review by Nvidia like this: http://techreport.com/review/27067/nvidia-geforce-gtx-980-and-970-graphics-cards-reviewed we can see they talks about the GM204 chip (970 & 980) and it has 64 ROPS and 2048 L2 Cache.

And another http://www.pcgamer.com/gigabyte-geforce-gtx-970-g1-gaming-review/ they talk about L2 cache and ROPS.

And finally http://hothardware.com/Reviews/NVIDIA-GeForce-GTX-980--970-Maxwell-GPU-Reviews/ another table comparing both cards. Both supposedly have 64 ROPS and 2048 L2 Cache.


MANUFACTURERS/RETAILERS


The same, go to their websites and you will find tens of reviews talking about the Rops, the L2 cache, bandwidth....


Remember the reviews are supposedly officially APPROVED. Many buyers read these reviews before buying.

Now you have proof of false advertising. Go to the pages before they delete it, save it to pdf and send to Nvidia/Manufacturer/Retailer.
Source: https://forums.geforce.com/default/...formation-about-false-advertising-must-read-/

Add in the admittedly-known-by-them design issue/problem with the segmented VRAM sections and it's very clear-cut that they falsely advertised these cards.
 
Wow....Nvidia I cannot believe the shit you are saying now....

This reminds me when the Iphone 4 drop calls bullshit, that Steve Jobs said there isnt an issue, just hold it differently.

I mean really?....I knew Nvidia was arrogant but this takes the cake.

Too bad I cannot get a refund on newegg :(
 
Did some more testing using Shadow of Mordor, SLI enabled and disabled, sub-3500mb and over 3500mb VRAM consumption. Frametimes stay within normal variation/acceptable consistence when in single-card mode regardless of VRAM, but going to over 3500mb in SLI causes wild and rampant stutters/hitches with vastly fluctuating frametimes to match.

mhOevfQ.png


Long story short, I have to agree for single-card that while it is a big false advertisement and spec change it may not have a giant practical impact (at least from what I can see so far... unless you may want to go dual-card later for example), but in SLI it is a very real and major issue.

This is really good data. Thank you for taking your time to test this out. Kudos to you :)
 
to be honest at first i didnt plan to return my two cards

its really about real world benchmarks not just specs and truth be told i dont really have alternatives to my sli 970 setup price wise (i have a small room and 290x runs to hot)

but Nvidia response to this or lack of it is deplorable it made me rethink rather fast
 
Have you not seen multiple real world benchmarks? 970 SLI is especially bad with high RAM usage.
 
I bet Jen-Hsun put his foot down and said "tough fucking shit, deal with it" to those that are upset about the false specifications and are wanting a refund or a fix.
 
The new video from AMD is retarded... they should focus on saving their $ in useful stuff!
 
LOL I was waiting for this.

God I hope AMD will be around for a very long time...for all our sake.

Odd that they will drop an F bomb but then use the word "butt" in the rant...

Either way I am hoping Kyle and crew chime in with less pitchfork & torch and more facts.
 
SO does the 980 have this issue with memory as well? Some posters mentioned something similar which I find hard to believe.
 
I think people are taking themselves a little too seriously here.

The GTX 970 meets the performance that reviewers and benchmarks produced, and we all purchased it based on the benchmarks rather than the stats.

If we purchases based on specs then nobody would have bought Athlons back in the day when Intel chips were clocked 30% higher... it wasn't the spec that mattered, it was the benchmark performance that put the Athlons in the lead.

The specs are not wrong and they're not dishonest in any way. The hardware is as specified and the controller behavior isn't advertised.
 
If we purchases based on specs then nobody would have bought Athlons back in the day when Intel chips were clocked 30% higher... it wasn't the spec that mattered, it was the benchmark performance that put the Athlons in the lead.

Not a really valid comparison, we all knew IPC mattered over clock speed, but their is no replacement for VRAM. The specifications were wrong which is false advertising. The card also stutters at 3.6GB of VRAM even if the GPU isn't being fully utilized.

Wcq4OBo.gif

1gjdAX7.gif
 
I think its funny all the hate right now. The card has the ram, the card can use the ram. I realize some people bought this in the hopes to run 4K, but did you really think a card that is priced at 320 bucks would run 4K without problems? There are going to be trade offs. If you can afford a 4K monitor, why cheap out and get a mid range card. Last I heard, the GTXx80's are always the flagship card. I have a 970 and love it. We all knew this wasnt the full sized maxwell anyway, its a mid range chip, get over it people. It is however, the fastest card you can get for the money when under 4K resolutions. If the VRAM limitation gets helped out with a driver update, cool, if not, give us an option to ignore the last .5GB and carry on with our lives people. The card is good, the performance hasn't changed since release. Nvidia has segmented memory on cards like this since the GTX500 days.

Also, I laughed pretty hard at the Hilter video. It is pretty funny and sums up how most people are reacting toward this.
 
Last edited:
Not a really valid comparison, we all knew IPC mattered over clock speed, but their is no replacement for VRAM. The specifications were wrong which is false advertising. The card also stutters at 3.6GB of VRAM even if the GPU isn't being fully utilized.

On top of that, the clock speed of the Athlon's weren't hidden. An Athlon 2600+ specifications didn't say it ran at 2.6GHz.

The GTX 970 meets the performance that reviewers and benchmarks produced, and we all purchased it based on the benchmarks rather than the stats.

No, I just bought one from a forum member and it wasn't JUST because of the benchmarks. Benchmarks didn't hurt, but I wanted it because it had 4GB VRAM and I'm looking forward to getting myself into 4k as well as more demanding games. I'm well aware it's not top notch like the 980 and that it's got some sacrifices that could hurt me... But one of them was not a blow to the VRAM.

I think its funny all the hate right now. The card has the ram, the card can use the ram. I realize some people bought this in the hopes to run 4K, but did you really think a card that is priced at 320 bucks would run 4K without problems? There are going to be trade offs. If you can afford a 4K monitor, why cheap out and get a mid range card. Last I heard, the GTXx80's are always the flagship card. I have a 970 and love it. We all knew this wasnt the full sized maxwell anyway, its a mid range chip, get over it people. It is however, the fastest card you can get for the money when under 4K resolutions. If the VRAM limitation gets helped out with a driver update, cool, if not, give us an option to ignore the last .5GB and carry on with our lives people. The card is good, the performance hasn't changed since release. Nvidia has segmented memory on cards like this since the GTX500 days.

Also, I laughed pretty hard at the Hilter video. It is pretty funny and sums up how most people are reacting toward this.

Personally I'm looking forward to the card. But why not inform buyers of the true specs? I'm also one of those people that believes nVidia would have still sold a ton 970's with 3.5GB of VRAM. So why not just come clean? I think that's the biggest issue everyone has.
 
Last edited:
I think its funny all the hate right now. The card has the ram, the card can use the ram
Disingenuous and you know that full well.
The Card has 4GB of ram but it cannot use it without a severe performance penalty. Oh yeah it also doesn't have the advertised bus speed, Rops, or cache that we told you about months ago. No one could have guessed that this card would get micro stutter horribly at vram settings of 4GB with under 100% GPU utilization. The Drivers even attempt to not use that 3.5GB of video ram.
 
I see people excusing Nvidia focusing purely on the RAM split , but as part of this , aren't there several other specs that are incorrect dealing with the memory bandwidth measurements also ?
 
Too many pages to browse in 2+ threads; are [H] going to give this even the slightest mention or are they taking their time working on an article running deep tests :confused:
 
Status
Not open for further replies.
Back
Top