Official Nvidia GTX 970 Discussion Thread

[JustCallMeVlad]

If the 980 Is a V8 the 970 is a twin turbo v6 and can we please stfu about the memory already. Go buy a 290x if you're that upset. Good Riddance

[PCNOOB123]

What gamers tear at 1080p? Cuz mine is here. BTW CPU is 4690k

[Whorax]

Oh please stop whining people. People including myself loved the benchmarks when it was released, now that we know VRAM slows after 3.5GB does not make a difference just cause we know it 1 month later, the 970 now and the 970 at release have the same problem but both perform the same and perform well, at most Nvidia should give a free game to ppl who purchased it.

 

You people are butt hurt.

A free game doesn't make up for lying directly to consumers for more sales. It comes nowhere close to making it up and I'd feel absolutely disrespected if Nvidia actually were to think a free game would make up for what they did.

[mr moose]

A free game doesn't make up for lying directly to consumers for more sales. It comes nowhere close to making it up and I'd feel absolutely disrespected if Nvidia actually were to think a free game would make up for what they did.

 

Only if you can prove it was intentional to get more sales.  Plus people are asking for it, nvidia haven't offered anything, so if they do now it will be because that's what people asked for.

[JustCallMeVlad]

Since the issue has surfaced retailers are only reporting a 5% return of the 970's that they sold.

[Whorax]

Only if you can prove it was intentional to get more sales.  Plus people are asking for it, nvidia haven't offered anything, so if they do now it will be because that's what people asked for.

Huge companies simply do not "miscommunicate".

[Whorax]

-snip-

LOL. I see you fanboying it up everywhere. Nice seeing you here as well, pal.

[s3ns3]

Since the issue has surfaced retailers are only reporting a 5% return of the 970's that they sold.

below 5%. 1-2%

[zappian]

below 5%. 1-2%

 

Nvidia will be like.

 

"Sweet lying to consumers is so nice it didnt even hurt our bottom - line , lets do it again but worse okay?"

[AngryGoldfish]

Sure but that's just binning right? Isn't that how it basically works for all (or at least most) processors?

 

Although I dislike the idea of binning, it is the way the industry works and is not the issue. The issue is misrepresentation and poor design concept. The 970's VRAM issue is a poor design choice, in my opinion. Any enthusiast who wanted a stellar (and affordable) setup for a 1440p monitor with two 970's in SLI are going to experience a VRAM bottleneck, when before we would always hit the GPU limit before we'd hit the VRAM limit.

 

 

Since the issue has surfaced retailers are only reporting a 5% return of the 970's that they sold.

 

Well, Gigabyte are straight up refusing to allow dealers to accept returns, so that would equate for a large market share. Also, the issue isn't affecting everyone as most are still gaming on single 1080p monitors and aren't fussed about anti-aliasing or texture packs that look no better. It was never presumed that all users would be sending back their cards or that they should. Some people don't give a rats ass about the fine details in question here. They just like dem frames pa second for cheapzorz.

 

[R_Golden]

Huge companies simply do not "miscommunicate".

That logic... A big company is gonna miscommunicate a lot more than a smaller one.

[Whorax]

That logic... A big company is gonna miscommunicate a lot more than a smaller one.

Not to the degree that they did with the 970.

[RazrLeaf]

That logic... A big company is gonna miscommunicate a lot more than a smaller one.

So true. It's a lot harder for departments (i.e. engineering and marketing) to communicate than it is for 2 teams, or just 2 people. And the more people there are, the more the information is spread out.

[RazrLeaf]

Not to the degree that they did with the 970.

No one knew that disabling just one section of L2 cache was possible. It's not hard to imagine that the engineer sending the memo forgot to mention it, and the marketing team receiving the information wouldn't have even known to ask.

[zappian]

No one knew that disabling just one section of L2 cache was possible. It's not hard to imagine that the engineer sending the memo forgot to mention it, and the marketing team receiving the information wouldn't have even known to ask.

I agree but why did wait 6 months to say something?

 

Why did they let it blow up in their faces?

 

Someone probably knew there was an issue with the memory bandwidth after launch at nvidia , they made the product.

[RazrLeaf]

I agree but why did wait 6 months to say something?

 

Why did they let it blow up in their faces?

 

Someone probably knew there was an issue with the memory bandwidth after launch at nvidia , they made the product.

Because no one realized it was possible. Marketing didn't know that they could even make that mistake. And the mistake was only in the detailed spec sheets that got sent to the press.

 

It's hard to pick up on a mistake when you don't realize it's a mistake.  Based on the reaction and how Nvidia dealt with it, it was only engineering that knew the L2 could be disabled. And who knows if everyone in engineering knew. If the engineer that looked over marketing's initial release didn't know, then X_X

[othertomperson]

Sure but that's just binning right? Isn't that how it basically works for all (or at least most) processors?

 

Well yeah, they're both GM 204.

 

The GTX 770 came with a whole gigabyte less vram than its partner chip, the GTX 780.

 

 

I agree but why did wait 6 months to say something?

 

Why did they let it blow up in their faces?

 

Someone probably knew there was an issue with the memory bandwidth after launch at nvidia , they made the product.

 

Because it isn't an issue that's even noticeable when gaming. It hasn't been proven to cause stuttering, it isn't associated with any particular framerate drop and all 4GB is usable in almost all games that require it.

[AngryGoldfish]

Because no one realized it was possible. Marketing didn't know that they could even make that mistake. And the mistake was only in the detailed spec sheets that got sent to the press.

 

It's hard to pick up on a mistake when you don't realize it's a mistake.  Based on the reaction and how Nvidia dealt with it, it was only engineering that knew the L2 could be disabled. And who knows if everyone in engineering knew. If the engineer that looked over marketing's initial release didn't know, then X_X

 

This is a good point, and one many already know, but for me personally it does not explain such a blunder. Not only that, but why should the community feel like they were lied to and given little choice but to complain and moan on the forums when nVidia could have communicated better and offered a solution? A mistake should come before a great big sense of humility. Nvidia seem to be showing none of that.

[Lewellyn]

So true. It's a lot harder for departments (i.e. engineering and marketing) to communicate than it is for 2 teams, or just 2 people. And the more people there are, the more the information is spread out.

 

If Nvidia actually thought this was a "miscommunication" why is this still the specification?

 

 

Their argument is that the 970 is working "as intended" and "as advertised." If they actually cared about letting the consumer know what they were purchasing gimped VRAM they would have officially changed this crap by now, but they haven't. The only people to bring this up are the users in forums or other media, and even then if you bring this up you are met with retaliation.

[zappian]

-sniperino-

 

I find it hard to believe they didnt notice the insane GB/S discrepancy .

Thats very unlikely.

 

But thats just my opinion.

 

After a lot of testing , memory testing QA thats really hard.

[RazrLeaf]

If Nvidia actually thought this was a "miscommunication" why is this still the specification?

 

Their argument is that the 970 is working "as intended" and "as advertised." If they actually cared about letting the consumer know what they were purchasing gimped VRAM they would have officially changed this crap by now, but they haven't. The only people to bring this up are the users in forums or other media, and even then if you bring this up you are met with retaliation.

The spec isn't completely wrong. There's still 4GB of VRAM. The bus is still (physically) 256-bit wide. They should probably change the bandwidth spec though...

 

This is a good point, and one many already know, but for me personally it does not explain such a blunder. Not only that, but why should the community feel like they were lied to and given little choice but to complain and moan on the forums when nVidia could have communicated better and offered a solution? A mistake should come before a great big sense of humility. Nvidia seem to be showing none of that.

I honestly think we're beating a dead horse, and crucifying a company that is made up of humans for a human mistake. Their VP of Engineering had a meeting with the press on a Sunday night to explain the memory architecture. I think that's doing their best to resolve this with "humility." And unfortunately, there is no realistic solution. I work for a company that once made a mistake. People died. The only solution was to say sorry and implement programs to make sure that it never happened again.

 

Nvidia communicated. But there is no solution that they can offer. Mass refunds? That's up to their AIBs and dealers. Mass upgrades to 980s? Same. Re-enabling the disabled L2 caches? If they could, they still wouldn't be able to fix the ones that had a faulty L2 and necessitated the disability. They'd get some at best. If we still want Nvidia to exist, we have to accept their apology and move on.  Many of the solutions we seek would destroy the company, either by ruining their finances or destroying their relationships with AIB partners and retailers. And I'm willing to bet that people on a forum like this looked at at least one review, and witnessed the real world performance of the card, which influenced our decision to purchase more than the spec sheet. 

 

I own a 970. I'm bummed that I'll have to treat it more like a 3.5GB card. But Nvidia is made of people that make mistakes. Sometimes the chain of mistakes is seen by the public. The only thing I can do is forgive them, and see what product they have to offer when it's time for me to buy again.

[Lewellyn]

The spec isn't completely wrong. There's still 4GB of VRAM. The bus is still (physically) 256-bit wide. They should probably change the bandwidth spec though...

 

I honestly think we're beating a dead horse, and crucifying a company that is made up of humans for a human mistake. Their VP of Engineering had a meeting with the press on a Sunday night to explain the memory architecture. I think that's doing their best to resolve this with "humility." And unfortunately, there is no realistic solution. I work for a company that once made a mistake. People died. The only solution was to say sorry and implement programs to make sure that it never happened again.

 

Nvidia communicated. But there is no solution that they can offer. Mass refunds? That's up to their AIBs and dealers. Mass upgrades to 980s? Same. Re-enabling the disabled L2 caches? If they could, they still wouldn't be able to fix the ones that had a faulty L2 and necessitated the disability. They'd get some at best.

 

Sorry to disagree, but I think "It still has 4GB of VRAM" is very disingenuous of the tech industry. I feel like the main reason that this is allowable by certain people lies solely on the quantity of the product that was affected. If it was 2GB (half) then surely the same defensive arguments would bear less weight.

[R_Golden]

Not to the degree that they did with the 970.

It's not like Nvidia is ten guys at a table... They are a huge company, they have different branches for every task. Engineering, Marketing, Design, etc... The 970 HAS 4GB of VRAM, therefore the marketing team marketed it as such, they aren't the engineers and the 970 was built in a very complex way. The only issue is the issue of why didn't Nvidia tell the consumer when/if they noticed but because we are not Nvidia there is no way for us to know whether they intentionally kept it a secret for longer, didn't know or whatever else could have been the case. Until there is some proof there is no point in arguing about it in my honest opinion and if you think Nvidia lied that's fine but don't pass it on as fact until you can prove it.

[RazrLeaf]

Sorry to disagree, but I think "It still has 4GB of VRAM" is very disingenuous of the tech industry. I feel like the main reason that this is allowable by certain people lies solely on the quantity of the product that was affected. If it was 2GB (half) then surely the same defensive arguments would bear less weight.

If they called it 3.5GB, then a lot of people would have be confused since there are 8x512MB on the card. It's not 4GB of VRAM in the same way it's been used in the past. But it's not quite 3.5GB either, even though it behaves like a 3.5GB almost all the time. Maybe they should change it to "3.5GB + .5GB?"

 

You have a fitting avatar btw. I like.

[othertomperson]

If they called it 3.5GB, then a lot of people would have be confused since there are 8x512MB on the card. It's not 4GB of VRAM in the same way it's been used in the past. But it's not quite 3.5GB either, even though it behaves like a 3.5GB almost all the time.

 

I, for one, would be baffled if they told me it had 3.5GB since I can use 3.7GB in Skyrim without issue at all.