Should you be compensated for the GTX 970 issues and spec changes?

[Paul98]

Post 636 from the other thread on the subject.

Don't look at absolute performance so much as what is being isolated. Notice the difference in BF4 going from 160% to 165% resolution scaling? The minimums shouldn't drop like that in my opinion, if the card had a true 4GB of usable memory, not from such a small bump.

But at any rate, I'd really like to see a good tech site dig into this issue and do some similar testing.

That is exactly the problem you will run into, the normal fps won't really change but you will get spikes that will ruin your experience. Especially if you are used to smooth gaming.

 

[Pneumothorax]

Do you have a link to that? Not that I would ever ask newegg to eat the $50 or so they'd lose selling my 970 as an open box, as newegg has been really good to me since 2001. Just interested to see. I can't imagine how many return requests they had.

Here you go!

 

[Firestorm007]

I voted yes. Sorry, I don't really care what the apologists say. You buy something that's advertised as such, you expect it to be exactly that. Some of the responses are beyond comprehension. I guess some like being ripped off. I certainly don't.

 

[flexy]

Funny thing:

If this was about a car advertised as having, say, 8 cylinders and it turns out it doesn't - it would only have 6 or 7 it would be a HUGE scandal and certainly tarnish a companies reputation on end.

But it's not a car (or any other mainstream product) we're talking about here, it's something for nerds who are usually willing to spend good money on "silly" upgrades anyway. (In other words: We're a rather stupid/silly audience to begin with which AS IT LOOKS can be fooled unlike others. We spend like €350 for hardware which makes noises (coil whine etc.) and just accept it..and now we're even accepting that technical specs of a product we bought "magically" changed from one day to the other. And it needed nerd-enthusiasts with a self-written tool to find this and have the maker admit...

NOT COOL, not cool at all.

IMHO, what NV should do is lower the price of their 980.

PS: I also find it ridiculous how some people think that a "free game" given by NV or something would solve things. We're not talking about a game or other crap, we're talking about hardware that costs several hundreds of dollars. We're talking about people who spent those dollars based on specs which now turned out not true.

*

As for me: I am not a rich guy who buys "the newest" every 6 months.(I did 10 years ago but back then graphics card were on avg. 50% cheaper...and in addition didn't exhibit flaws. A high-end card 10 years ago cost what the "cut versions" cost today.

So..when I spend money I think about the future, it should last at least a while. It's not so much that the "crippled" 970 is now suddenly becoming slower, it is not. But it now has a much shorter shelf-life since when it's already hitting limits NOW.....what in 3 months, what in 6 months? GTA V, mods etc...I don't want to buy this card and then run into memory- and therefore performance problems two months down the road because it only has 3.5GB and not 4GB like it says on the box...

 

[dn7309]

some one posted this over there at Nvidia forum, really sums up how real 970 owners over there are feeling.

If I see this picture one more time the next person will get an infraction over it. This is not 4chan.

-Rvenger

 

[dn7309]

Also I find it insane that people say "The GTX 970 benches just as well as it did before so I don't care!"
You seriously think Nvidia is that backwards of a company to release a GPU with a flaw like the GTX 970 has, and not ensure it works on the games that people will be using to benchmark it?

It's tomorrows games you have to worry about. It does great in 2014 benchmarks, but the year is 2015. What happens when we see a new host of games pushing the borders this year?

And I know people keep saying "4K isn't coming anyway anytime soon!"
I'm sure the R9 380x/390x and GTX 980Ti have something to say about that. Not to mention next year.

Ya, maybe you guys may only plan on keeping your GTX 970 for 1 year and don't care, but there are many keeping it for 3-4 years and don't think it's fair that they won't have Nvidia's driver team optimizing for an odd 3.5+.5GB configuration in 2018 on Triple A releases.

Sometimes I wonder if companies (even AMD) make their card specifically run very well on current popular benchmarks just to sell their cards, instead of making cards that actually stick to their claim of "running all of the today's and tomorrow's (NOT) games at blazing speed".

 

[DaveSimmons]

Sometimes I wonder if companies (even AMD) make their card specifically run very well on current popular benchmarks just to sell their cards, instead of making cards that actually stick to their claim of "running all of the today's and tomorrow's (NOT) games at blazing speed".

Both companies have been caught cheating on benchmarks in the past, so yes they do.

Skipping work, lowering image quality, pre-generating results and baking them in to the drivers, writing code that only works for one game in its benchmarking mode, it's all been done.

 

[Pneumothorax]

And so it begins....

Vram Usage (too bad the 970 wasn't tested)

If Maxwell's memory compression is hugely reducing VRAM usage and making a big difference here vs. the 290x, I'm not seeing it.

My 970 SLI is almost as good as the 980 at 1080P

Ouch 20% drop from 1080P for the 4gb 980, 30% hit for the '3.5gb' 970 along with it's 780ti 3gb older brother.

 

[Teizo]

I voted yes. Sorry, I don't really care what the apologists say. You buy something that's advertised as such, you expect it to be exactly that. Some of the responses are beyond comprehension. I guess some like being ripped off. I certainly don't.

Just curious....Are 970s all of a sudden not performing the same? Have they all of a sudden lost 10% performance based on these findings?

It's just FUD. Has nothing to do with apologist. It has everything to do with being wise enough, and experienced enough, to see things for what they are: a manufactured PR ploy to try to blunt 970 sales because over the last couple of months the card has been selling like crazy.....to the tune of allot of happy people claiming how great, and fast, the card is.

As I mentioned before....I was a bit concerned before Nvidia released their statement on the architecture of the card and the performance numbers relative to the 980 when >3.5gb of memory was utilized. IE...there was no marked difference....so no big deal.

 

[cmdrdredd]

Both companies have been caught cheating on benchmarks in the past, so yes they do.

Skipping work, lowering image quality, pre-generating results and baking them in to the drivers, writing code that only works for one game in its benchmarking mode, it's all been done.

My fondest memory of that is Quake 3 when the LOD was adjusted way, way out in a spot you couldn't see easily. Details were load and blurred to increase performance in that game.

 

[cmdrdredd]

And so it begins....

Vram Usage (too bad the 970 wasn't tested)

My 970 SLI is almost as good as the 980 at 1080P

Ouch 20% drop from 1080P for the 4gb 980, 30% hit for the '3.5gb' 970 along with it's 780ti 3gb older brother.

That game is terribly unoptimized. It even uses one thread on the CPU and shows i3 CPUs beating some i5 and i7s. There have been reports of stuttering that cannot be eliminated. Poor example IMO. At least in it's current state.

Unfortunately, Dying Light doesn't allow for Texture Quality to be changed mid-game. This, combined with checkpoint-style respawn locations and a constantly-changing time of day make direct comparisons in interesting locations impossible. From testing, it appears that there is no difference in quality between the two settings, and that High may merely be storing more textures in memory on suitably equipped GPUs. For example, running around an area on Medium resulted in a modicum of texture pop-in and VRAM usage of around 2GB. Repeating the test on High resulted in zero pop-in and VRAM usage that topped out at 3.3GB, though during longer gameplay sessions usage of nearly 4GB has been observed.

http://www.geforce.com/whats-new/gui...exture-quality

 

[Pneumothorax]

That game is terribly unoptimized. It even uses one thread on the CPU and shows i3 CPUs beating some i5 and i7s. There have been reports of stuttering that cannot be eliminated. Poor example IMO. At least in it's current state.

Even with being unoptimized, why does the 970 go from 3% slower to 16% slower than the 980 when going from 1080p to 1600p? An unoptimized game is going to hit both just as hard right? If it was being held back by the CPU, the drop in performance should be almost the same between the two cards.

 

[cmdrdredd]

Even with being unoptimized, why does the 970 go from 3% slower to 16% slower than the 980 when going from 1080p to 1600p? An unoptimized game is going to hit both just as hard right? If it was being held back by the CPU, the drop in performance should be almost the same between the two cards.

It could be that the 980 simply has more muscle to power through certain aspects. Dying Light is a game with issues right now.

 

[Gloomy]

source: http://www.pcgameshardware.de/Geforce-GTX-970-Grafikkarte-259503/Specials/zu-wenig-VRAM-1149056/

They downclocked a 980 until it matched the speed of a 970, then benched both. As expected, the 970 has serious frametimes spikes not present on the 980, despite both having the same performance (30FPS vs 31.8FPS)

They also have a 290X there, but no 290 so I don't know what to make of it... other than "Watch_Dogs runs like garbage on AMD GPUs" of course.

This is at 1080p 8xMSAA.

 

[Silverforce11]

Just shows how misleading "average fps" metrics can be.

 

[Gloomy]

What's interesting is it even happens when the 970 is OC against a downclocked 980, not at insane 8x MSAA, but 4x MSAA.

That's at 4k, lol. I'm not sure those results are worth talking about. 11FPS... :|

 

[Silverforce11]

That's at 4k, lol. I'm not sure those results are worth talking about. 11FPS... :|

Could be if someone rocks Quad SLI. ;p

 

[cmdrdredd]

That's at 4k, lol. I'm not sure those results are worth talking about. 11FPS... :|

I don't find 4k playable in most games. Not even with 2 cards. At least not according to the benches I have seen. I'm not happy to turn settings down to get to 4k. Rather do 1440p or something.

source: http://www.pcgameshardware.de/Geforce-GTX-970-Grafikkarte-259503/Specials/zu-wenig-VRAM-1149056/

They downclocked a 980 until it matched the speed of a 970, then benched both. As expected, the 970 has serious frametimes spikes not present on the 980, despite both having the same performance (30FPS vs 31.8FPS)

They also have a 290X there, but no 290 so I don't know what to make of it... other than "Watch_Dogs runs like garbage on AMD GPUs" of course.

What matters most to me is how it feels. It's kind of subjective but the graph says one thing but if it doesn't affect the gameplay at all and everything still feels good then I'd lean toward the no big deal side. For everyone that would be different.

 

[Riceninja]

SLI users really got screwed:

I benchmarked GTX 970's in SLI at 1440P and above 3.5gb. Here are my impressions.

 

[cmdrdredd]

SLI users really got screwed:

I benchmarked GTX 970's in SLI at 1440P and above 3.5gb. Here are my impressions.

I play at 1440p and don't have a problem. I am not forcing 8x aa though to push the issue.

 

[lehtv]

@Riceninja
Nice find... I would be pretty annoyed if I ran 1440p and SLI, as it's infinitely more likely to come across a situation where this is an issue, compared to running a single 970 on 1080p.

 

[happy medium]

If I'm reading the Newegg statement right, the card does not lose any more performance than a gtx980 above 3.5gb of memory. At least in line within a margin of error.

http://kb.newegg.com/FAQ/Article/1729

Am I missing something? Is this about Nvidia lying about specs or performance?
If a card with 3.5gb's +512mb of memory performs just like a 4gb card ,whats the problem?

 

[.vodka]

That is nvidia's official statement on the matter they're reproducing on their page.

 

[happy medium]

That is nvidia's official statement on the matter they're reproducing on their page.

Are the benchmarks on Neweggs site wrong?

AGAIN......
Am I missing something? Is this about Nvidia lying about specs or performance?
If a card with 3.5gb's +512mb of memory performs just like a 4gb card ,whats the problem?

 

[Gloomy]

What matters most to me is how it feels. It's kind of subjective but the graph says one thing but if it doesn't affect the gameplay at all and everything still feels good then I'd lean toward the no big deal side. For everyone that would be different.

Sorry, you are not spinning 100ms spikes where there shouldn't be any as subjective. Please go away :thumbsdown: