• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

970's having performance issues using 4GB Vram - Nvidia investigating

Status
Not open for further replies.
whyscotty said:
If by that you mean future proof, no such thing.
Click to expand...

You said "futureproof", I distinctly said "long term investment". This could mean 3 or more years in the world of graphics cards. Probably unusual on this forum but probably not unusual elsewhere.

I notice you managed to avoid answering the question about whether it would be fair if this problem shortened the perceived usable life of the card by 6 months.
 
rolypolyman said:
Does anyone actually believe the ' miscommunication ' story. This doesn't affect me at all but I have read through the thread and find it incredulous that anyone can buy into what Nvidia are saying, then of course having discovered their 'miscommunication' who at Nvidia decided not to tell anyone, was bound to get out. Only the extremely loyal Nvidia customers are still defending them, its not about 'does it affect you' or 'its only 4-6%' for me its about integrity etc, something sadly lacking in business. Be all for me is Nvidia should have come clean on this but decided it would affect sales of the 970 so didn't, now they have been caught they must take the consequences. :(
Click to expand...

PayneUK said:
This is genuinely what I am most annoyed about; I have almost exclusively owned NVidia GPU’s and haven’t even looked at AMD’s offerings for a few years, I am biased towards the green side but that does not give them a free pass.

Seeking a refund has nothing to do with anything other than its one of the few ways I have to tell companies it’s NOT OK, I will probably pick up a NVidia product again in the future, they normally make better stuff, but like a naughty child you have to tell them what they’ve done wrong before you give them another chance.
Click to expand...

This is exactly how I feel. I'm inclined to believe Nvidia when they say they made a mistake in the first place, but I find it hard to believe that no one spotted the mistake in 4 months. It seems much more likely that they realised their mistake and decided to keep quiet about it. I'm sure they'd have been quick to point out a mistake in their favour.

While I still think a 970 would satisfy my immediate needs, I'm inclined to cancel my pre-order for an OcUK reference 970 and wait for the AMD R390. I have been tempted to change my order to a 980, but that would mean rewarding Nvidia for their lack of integrity.
 
Stas823 said:
I honestly wouldn't have bought the 970's if I knew that they were less than 4gb VRAM. Nvidia need to do something about this its their fault not benchmarks/software.
Click to expand...

likewise, i don't get much funds to spend on PC parts and when i do i like to make sure they are no issues. if i had saw this before i wouldn't have brought a 970
 
Gregster said:
As for those who play at 4K on a 970, turn the settings down or get a bigger VRAM card and get 3 of them if you want to max settings out. 4K requires 3 Titans/ 8 GB 290Xs to have max settings in a few AAA titles.

Please stop saying that 0.5GB is unusable. It clearly is usable but just runs a touch slower than the other 7/8ths of the memory (4-6% slower).
Click to expand...

1. Why should I have to turn down texture settings on games at 4k just because Nvidia lied about the cards memory pool layout?

2. Please stop saying the 0.5Gb is 4-6% slower, It's actually 1/7th the speed of the other pool which is much larger then 6%

3. GTX970 SLI an max out 99% of all games at 4k, it's just a shame that the secondary VRAM pool makes certain games unplayable when they should be.

One of the deciding factors on me spending on ver£600 on a pair of GTX970's was 4Gb of VRAM and 64 ROP's.

The 4Gb VRAM cab be argued but my card doesn't have 64 ROP's and if Nvidia would been honest and released the true specifications in the first place I wouldn't of bought them.
 
Gregster said:
I have asked several times for people to post what issues they are getting in games and post a vid showing or even a screenie of the VRAM used but none :(
Click to expand...

I've posted the issues I'm having a few times now, if you can talk me through what to do I can post my videos up to highlight the issues, if nothing else it will show people what's happening and at the very worst hopefully I'm doing something stupid which contributes to the issues and someone can give me advice on fixing it if it's a problem I've created somehow (yes I'm still holding out hope :( ).

Also I contacted Overclockers a few minutes ago (I ended up calling 5 times as each time I got through to RMA someone put the phone down and at one point in the middle of me explaining my problem), when I called them on Friday explaining my issue they were happy to give me an rma number when I called back as I didn't have my order number, spoke to someone today and explained the same thing and as expected they are saying there's no problem and the cards are working as intended and no RMA will be issued.
I can also confirm they are talking to Nvidia at the moment but the guy did say there is no time scale for this issue being resolved and they have no time scale as to when they will hear back from Nvidia.

I don't think anything is going to come out of this, a lot of noise is being made now but the company will just let the dust blow over and after that it will be business as usual.
For me I'll continue to use these cards but I do feel like I've peed away £600 on something that I bought based on the fact I was told it's pretty close to a 980 but in reality it's not as the newer games that come out that will use the full 4gb will stutter for me considering the ones I'm playing now are currently doing that.
Ah well when the 390 comes out I'll have a good look at that and make my decision on whether to keep the cards or sell them to move back to the red team.
 
The Running Man said:
You didn't read the reports correctly. If all 4gb is filled and is all being accessed the whole memory bus goes down and effective memory band with drops massively.
Click to expand...

No, I read PCPers report (twice in fact) and they clearly state that the memory speed drops 4-6% when the whole memory is being read.

Accessing that 500MB of memory on its own is slower. Accessing that 500MB as part of the 4GB total slows things down by 4-6%, at least according to NVIDIA. So now the difficult question: did NVIDIA lie to us?

At the very least, the company did not fully disclose the missing L2 and ROP partition on the GTX 970, even if it was due to miscommunication internally. The question “should the GTX 970 be called a 3.5GB card?” is more of a philosophical debate. There is 4GB of physical memory on the card and you can definitely access all 4GB of when the game and operating system determine it is necessary. But 1/8th of that memory can only be accessed in a slower manner than the other 7/8th, even if that 1/8th is 4x faster than system memory over PCI Express. NVIDIA claims that the architecture is working exactly as intended and that with competent OS heuristics the performance difference should be negligible in real-world gaming scenarios.
Click to expand...
 
As little as I undestand of this issue. Isn't one of the possible effects in addition of stuttering be flickering or missing textures,when too much data needs to be loaded from slower memory partition?
 
Gregster said:
No, I read PCPers report (twice in fact) and they clearly state that the memory speed drops 4-6% when the whole memory is being read.
Click to expand...

Tou didn't read correctly. Or have misinterpreted ithe info. From anand:

GTX 970 can read the 3.5GB segment at 196GB/sec (7GHz * 7 ports * 32-bits), or it can read the 512MB segment at 28GB/sec, but it cannot read from both at once; it is a true XOR situation. Furthermore because the 512MB segment cannot be read at the same time as the 3.5GB segment, reading this segment blocks accessing the 3.5GB segment for that cycle, further reducing the effective memory bandwidth of the card. In concept, the larger the percentage of the time the crossbar is reading the 512MB segment, the lower the effective read memory bandwidth would be from the 3.5GB segment.
 
Gregster said:
Please stop saying that 0.5GB is unusable. It clearly is usable but just runs a touch slower than the other 7/8ths of the memory (4-6% slower).
Click to expand...

That's not correct. According to the PC Perspective article:

"Let's be blunt here: access to the 0.5GB of memory, on its own and in a vacuum, would occur at 1/7th of the speed of the 3.5GB pool of memory".

"But the net result for gaming scenarios is much less dramatic than that, so why is that the case? It comes down to the way that memory is allocated by the operating system for applications and games".

"A GTX 970 without this memory pool division would run 4-6% faster than the GTX 970s selling today in high memory utilization scenarios".

The 0.5GB runs a lot slower (7 times slower) than the other 3.5GB, but the effect on overall system performance is believed to be of the order of 4-6%.
 
dewi said:
This is exactly how I feel. I'm inclined to believe Nvidia when they say they made a mistake in the first place, but I find it hard to believe that no one spotted the mistake in 4 months.
Click to expand...

This is how I feel about it. A previous job of mine was for a huge global company wherein each and every communication, either internal or external would go through a distribution and review process before final release. However, even with this checking and double checking mistakes occasionally would slip through. Nothing like as game changing as this one by nvidia, but mistakes nevertheless did happen.

However, I'm of the opinion that once discovered nvidia made a high level decision to keep the mistake hidden and hope things were kept under wraps. For them there has evidently been more potential upside (I.e. higher profit) to ignoring the issue and dealing with the fallout, than owning up in the first place and slowing the sales of their new card.
 
The stupid thing is Nvidia didn't HAVE to disable a L2/ROP partition to reduce the core count of the die.

They could have left all the L2 cache and ROP's in tact and just disabled 3 SM blocks.

They there no such thins as it's miss communication between Nvidia and press, they knew from the start.

They all read reviews on there own products and anyone with half a brain would have noticed that petty much every review mentioned 64 ROP's and no one internally at Nvidia bought that up in a meeting?

"Hey guys I was reading a few reviews of our GTX970 last night and they mentioned the card has 64 ROP's, should we not get the correct information sent out?"
 
Last edited:
Spacedeck said:
Has anybody who has tried to RMA a 970 actually had success in full refund yet? If not, what was the outcome?
Click to expand...

Well with OCUK I doubt anyone has been given a yes or no yet, the admission from NVidia only came out later on yesterday so their probably still deciding on what the official line will be.

I’ve made a request but I don’t expect an answer today, I don’t want retailers to lose money because of this, I want NVidia to lose money, so I’m happy to wait while they talk it out.
 
Status
Not open for further replies.
Back
Top Bottom