Battlefield 4 Results
The GTX 970 memory issue led our readers to request testing of the cards paired up in SLI. Here you go.
At the end of my first Frame Rating evaluation of the GTX 970 after the discovery of the memory architecture issue, I proposed the idea that SLI testing would need to be done to come to a more concrete conclusion on the entire debate. It seems that our readers and the community at large agreed with us in this instance, repeatedly asking for those results in the comments of the story. After spending the better part of a full day running and re-running SLI results on a pair of GeForce GTX 970 and GTX 980 cards, we have the answers you're looking for.
Today's story is going to be short on details and long on data, so if you want the full back story on what is going on why we are taking a specific look at the GTX 970 in this capacity, read here:
- Part 1: NVIDIA issues initial statement
- Part 2: Full GTX 970 memory architecture disclosed
- Part 3: Frame Rating: GTX 970 vs GTX 980
- Part 4: Frame Rating: GTX 970 SLI vs GTX 980 SLI (what you are reading now)
Okay, are we good now? Let's dive into the first set of results in Battlefield 4.
Battlefield 4 Results
Just as I did with the first GTX 970 performance testing article, I tested Battlefield 4 at 3840×2160 (4K) and utilized the game's ability to linearly scale resolution to help me increase GPU memory allocation. In the game settings you can change that scaling option by a percentage: I went from 110% to 150% in 10% increments, increasing the load on the GPU with each step.
Memory allocation between the two SLI configurations was similar, but not as perfectly aligned with each other as we saw with our single GPU testing.
In a couple of cases, at 120% and 130% scaling, the GTX 970 cards in SLI are actually each using more memory than the GTX 980 cards. That difference is only ~100MB but that delta was not present at all in the single GPU testing.
Our performance data is being broken up into two sets: the GTX 980s in SLI running at all five of our scaling settings and, separately, the GTX 970s in SLI running on the same five scaling settings. Plotting 10 sets of data on a single graph proved to be a a bit too crowded, so we'll show the graphs successively to help you compare them more easily.
Unlike our first sets of results the SLI numbers are ALMOST in a playable state, making them much more real-world than before. The first thing I noticed when compiling this data was that the GTX 980 cards in SLI actually had a couple of more downward spikes in frame rate at 150% scaling than the GTX 970s did. I did confirm this was a regular pattern by re-running tests on both sets of hardware about six times, and the bright green line you see in the first graph above is actually one of the better results for the 980s.
It appears though that moving from 110% to 150% scaling results in the expected frame rate decreases in both configurations.
Average frame rates are where we expect them to be: the GTX 980 SLI is faster than the GTX 970 SLI by fairly regular margins.
| GTX 980 | GTX 970 | % Difference | |
|---|---|---|---|
| 1.10x Scaling | 47.3 FPS | 41.0 FPS | -15% |
| 1.20x Scaling | 41.1 FPS | 35.8 FPS | -15% |
| 1.30x Scaling | 35.4 FPS | 31.2 FPS | -13% |
| 1.40x Scaling | 31.0 FPS | 27.7 FPS | -12% |
| 1.50x Scaling | 27.7 FPS | 24.6 FPS | -13% |
The percentage differences here are actually more reliable than the single GPU results, which is a bit of a surprise to us. The GTX 970s in SLI range from 12-15% slower than the GTX 980s in SLI, but as we know from our years of GPU evaluation, that isn't the whole story.
Click to Enlarge
It should be painfully obvious that at 150% scaling the GTX 970s in SLI have a significant amount of frame time variance that is mostly limited with the GTX 980s in SLI. Even at 140% scaling, looking at the thinner gray line, you can see differences in the behavior of the frame time lines in the two graphs (you can click to enlarge them for a closer view).
Under closing thoughts,
Under closing thoughts, second paragraph.
Hmmm…..im still really on
Hmmm…..im still really on the fence between a pair of GTX 970s and another GTX 770…for 1440p.
*sigh*
Thankyou for the very informative article. Much appreciated.
I have 2×770 and a 1440p
I have 2×770 and a 1440p monitor. For some games, it’s mostly okay, but for non stuttering excellence, it’s just not enough. I will go for 2×970.
So basically you’re telling
So basically you’re telling us what we knew all along. The 970s perform worse than the 980s. Is there any chance we can see some frame time comparisons with the 290/290x in xfire? From a consumer standpoint, I think that’s the most important question that needs to be answered.
No need to compare AMD vram
No need to compare AMD vram vs Nvidia.. Since AMD does have much better VRAM @ 4/8GB 512bit (Which is why AMD needs more power then Nvidia). Nvidia just has the better GPU, an you can really notice this difference in 4k res benchmarks.
Oh, sorry, I forgot. AMD
Oh, sorry, I forgot. AMD doesn’t have problems with stutter on dual card setups.
stutter is always going to be
stutter is always going to be a problem in games no matter what we do. So i don’t worry about it much.
AMD doesn’t have a problem? I
AMD doesn’t have a problem? I guess you forgot about how many years AMD had a CF problem yet complete ignored all the people complaining about it, before a tool that was developed by Nvidia of all people was released and revealed the extent of the problem with an AMD CF setup. So AMD isn’t exactly a clean slate on their side.
I think Jappetto was being
I think Jappetto was being sarcastic. 🙂
Right, when it works,
Right, when it works, Crossfire just runs like shit.
I completely agree.
This
I completely agree.
This tests and results are really in a vacuum!
At the light of the results above, I could very well state that my girlfriends’ Gigabyte GTX 770 4GB SLi setup runs the same games at same conditions better than a GTX 970 SLi setup because, the GTX 770 in question has REALLY 4GB VRAM and everybody would believe me, if they didn’t knew better!
We need AMD R9 290(X) frame-pacing results.
AMD is so confident about their 4GB VRAM, 512bit, I don’t see why not put them to the test, if only just to see R9 290(X) CF crush GTX 970 SLi at the same frame-pacing benchmarks
All realtime gameing
All realtime gameing benchmarks ive seen at high res have shown 970 sli and r9 290x to be within about 4 fps of eachother at 4k. Hardly measurable. and you can run 970 sli on a 600watt psu. you need minimum of 800 for r9 290x. 1000 if your useing amd cpu as well. Lets be realistic shall we.
Many users complaining many
Many users complaining many months with Skyrim mod, Shadow Mordor and other games about GTX 970 stutter behavior. PCPer asked again and again to test 970 with FCAT to prove it has bad design for new high texture games.
Big question is when PCPer bring Nvidia executives with live questions from users to answer for FRAUD and LIES about “same memory subsystem as GTX 980”, cache size, ROP readings from GPU-Z?
When will Nvidia PAY refund for telling lies to PCPer and to users?
The worst for Nvidia and it’s
The worst for Nvidia and it’s partners is still in front of them. It seems that Nvidia doesn’t understand that. They think that people will be persuaded that their cards are as good as they should be, no matter the false specs.
Well there are cases where people returned their cards. But most people think differently. Why take the card and return it today, when the only real options are 290X or to pay more for a 980? Why not keep it for a few more months? In a few months from now the new AMD cards will be out and possibly the new Nvidias too. Then at 350 there are going to be options that will be clearly faster that 970. The cards will still be under warranty and the fraud will still be fraud. Then I expect a second wave of people returning their cards, much bigger than the one we see today. That’s when Nvidia will realize that trying to twist reality by giving directions to the press to focus on the cards performance was a bad idea.
Yea sure, because nvidia will
Yea sure, because nvidia will allow you to return your card till forever. Either you are upset an return it now, or you aren’t upset enough and keep the card. Once you keep the card you are stuck with it, and rightly so. You can’t have the cake and eat it too.
P.s. Nvidia clearly made a mistake, perhaps with evil intention, probably not. this still is far from fraud. LOL. Now flame me if you must and call me an nvidia employee.
You’re under a
You’re under a misapprehension that it’s up to Nvidia to accept a refund or not. In Europe and Australia, retailers have to accept those returns no matter what. There’s no statute of limitations on returns for false advertising. I feel bad for them, but the ones that screwed them was Nvidia not the consumer.
There’s a storm brewing, and believe me it’ll hit landfall when AMD’s 3xx series comes out. I should know, because I’m waiting for this exact situation to return my 970.
Wrong. No one is forcing you
Wrong. No one is forcing you to watch technology news all the time and I bet Nvidia or the shops didn’t contacted all the customers to inform them about that “mistake” by email or phone. So someone can go at a shop in 3 months and say
“I just found out. This product is still under warranty and in perfect condition. I haven’t misused it in anyway. But specs are not as advertised. You lied to me. I want a refund”.
I tried making this argument
I tried making this argument to MSI and it went nowhere. They flat out refused to refund my card or trade up to a 980. Same with newegg. NVIDIA was the one that made the mistake here, and it’s up to them to make good on this.
I thought about a charge back
I thought about a charge back on my credit card… dunno if it would work or if I would be able to order from Newegg ever again… which I probably won’t as their return policy is crap compared to Amazon.
If you have the card in your
If you have the card in your possession and do that without even trying to send the card back. That would constitute Fraud and you could be looking at criminal charges for that not to mention what they could do to your credit rating.
You have a card that was sold
You have a card that was sold to you using fraudulent specs. Sue the retailer if they won’t give you your money back.
Not even in EU. Unless you
Not even in EU. Unless you use 2 weeks (or however long it is in particular country as this can vary between member countries) period there is no chance with warranty as there is no defect and majority of user facing materials (unless listed explicitly by shop or vendor) didn’t have affected data. (L2/ROP count) and thus there is no right to warranty return. It’s up to good will on side of shop or vendor.
Therefore there is even no requirement to contact all buyers. Nothing relevant has changed or affected hardware… (no defect like Phenoms TLB bug or Transaction memory bug or say improper solder on GPUs)
You do have 8 defective ROPs
You do have 8 defective ROPs compared with the original specs, but anyway, you might be right. We will see.
at least in sweden you can
at least in sweden you can “reklamera” an electronic product for up to 3 years if it for example fails before these 3 years because some capacitors fail, then you can point out that it was faulty from the beginning, the first 6 months is up to the reseller to prove it was you who broke it and the last 2.5 years it`s up to you to prove the problem was there to begin with and capacitors are an easy example to get a refund or an equal product, so the fact that nvidia lied about the specs and the memory then you can (reklamera) the card since it goes under the consumerlaws here that it is a defective produkt since you did not get exactly what you paid for(false advertising), the reseller here has 3 options, replace it with something that has the right specs
(not really possible in this case since all 970s suffers from this problem)
2. give a refund,
or 3 repair it (again not really an option),
though i hear many has gotten the option to uppgrade to a gtx 980 at a discount:) and while i wouldnt want to give nvidia more money atm i have tried out the new gtx 980 G1 ganing edition and it overcloked lika a beast 1570 core clocke and 7940 mhz, it was an extremley nice card but i was to curius about the new memory system om 380 i decided to wait and sent my gtx 980 back sin i want more powah!,
so right now im waiting for 380x/90x and nvidias GM 200, though the gm 200 will have to perform at least 30 % for me to even start looking at it since im a bit comprohasive about giving nvidia my money atm:)
Someone orders a pizza and it
Someone orders a pizza and it comes with only 7 slices instead of the 8 as advertised. But 7 slices are more than enough for most people to satisfy their hunger, so we should forget that the pizzeria lied, and only remember that “it offers a tremendous level of performance, some amazing capabilities courtesy of the Maxwell GPU and runs incredibly efficient at the same time.”
The review is excellent as usually and really informative. But the press keeps using the card’s performance to cover up the real problem here. A company lied. If we where talking about that other poor company, no performance results would have been enough to cover the scandal. We all saw that when Hawaii come out and almost no one from the press was paying attention to the performance of the card. All where looking at the gpu speed and pointing fingers at AMD.
Welcome to Nvidia AD
Welcome to Nvidia AD supported reviews sites. They don’t care about AMD at all really.
All the websites get ad
All the websites get ad revenue from all the makers of PC/laptop/mobile parts, it’s the whole market that needs to be looked at, by The FTC, the companies have much control, even more so than just advertising, they have the review samples, in addition. Some web pages are nothing more than generators of sponsored content, and certain writers do nothing other than write poorly reworded content from the marketing copy, straight from the marketing departments of the sponsors that provide them. When a writer reviews a product, really reviews a product, look for much comparison, and contrast, with that product’s competition, be it GPU, CPU, or whatever.
A shining example of the control big companies have over the online “journalists”, is the recent reviews of Intel’s latest server SKU offerings, there where many enthusiast web sites mentioning Intel’s server SKU, but there was no comparison of the Xeon with the SPARC, or the Xeon with the Power8, and very little to no server benchmarks where used to compare the competing products. Intel gets a lot of “mind share” by getting these friendly blurbs out there, its easy to tell, because the article(Ad Really) makes no mention of the competition.
This type of sponsored content, that is not listed as such, is what the FTC needs to be on the lookout for, that and the tech writer hacks, who do little more than repackage the marketing copy, and word it so that the readers think they are getting an objective description. If you are reading a “review” look for the necessary comparison and contrast with the competing SKU/s, among other more complete information, the single item reviews are OK if links are always be provided to a review of the competing product, preferably from more than one source. Fore sure these sites review products, and have a business relationship with the makers of the products that they review, and it’s this one hand washes the other relationship that will force the readers to always have to read as many different articles on the product as possible, that and waiting on the newest products, until there is enough user feedback on the blogs, to be able to clearly see what the product really can do.
Well I made a bet and won.
Well I made a bet and won. The bet was that PCPER, Techreport, TechPpowerUp and maybe even AnandTech will eventually test the GTX 970 SLI when it starts to get too obvious (Nvidia BIAS) but will exclude the R9 290X/290/295X, so the GTX 970 SLI don’t look as bad as they really are.
So predictable its not even funny anymore.
Back when 970 was new they
Back when 970 was new they compared them to 290’s and did SLI testing. So go search for them first before acting like a dumbass with no damn clue about anything.
The petition to Nvidia for a
The petition to Nvidia for a refund hasn’t even reached 10,000 yet.
Given that Nvidia has sold well over a 1 million of these cards I think it’s safe to assume that the vast majority of people don’t care and will likely never run into the issues requiring over 3.5GB of ram or more ROPs.
Nope, most people/kids still
Nope, most people/kids still don`t know
As story reports, card works
As story reports, card works fine til you start pushing really high resolution with really high graphic settings. Generally settings in question fps gets to point anyway that is pretty low and unplayable anyway.
Not if you’re using using SLI
Not if you’re using using SLI @4K. I actually have a 3-Way setup.
My thought exactly. I visited
My thought exactly. I visited that petition site a couple of times and the number of participants is completely negligible in a relative sense.
A couple of years ago, games
A couple of years ago, games would fit on a single dvd, now 15-16GB seems a common affair, so let me ask you a question, are you ready to invest that much of money in a GTX 970, knowing fully well that it is not future proof? Plus, had nvidia limited the card to only 3.5GBs and removed the slower memory altogther then it would have been a much wiser move. But that’s not what they did, and in a year or two when you see games with higher requirements(say crisis 4 or farcry 5), 4K becoming more common, you’ll rue your decision to buy this blasted card, cause everytime you exceed the 3.5GB limit you are gonna stagger !
What a lame statement. I have
What a lame statement. I have an AMD card and the video driver continues to crash. AMD sucks big time.
The Pizza has 8 slices, its
The Pizza has 8 slices, its just that the crust at the edges is thicker and takes longer to chew then what you expected, just like the 970 has 4 meg, its just that 1/8 of it is slower then expected
The pizza has 8 slices, just
The pizza has 8 slices, just one of them’s no good for eating, it’s really chewy and maybe tastes odd. You bought it instead of a pizza from a different place which is a bit more hit and miss but would’ve definitely given you the 8 slices you ordered, charged you a bit less, and thrown in some onion rings and a coke.
Some people might be pissed off. Others with less of an appetite mightn’t care, they’d’ve only put the 8th slice in the fridge anyway.
In this case it’s more like
In this case it’s more like the pizza comes with the advertised 8 pieces but the 8th piece has nothing on it.
Haha, I like analogy, well I
Haha, I like analogy, well I think its off but I like pizza… problem with the analogy is you didn’t get 7/8ths of a GPU. Better would be you ordered a pepperoni pizza that advertised 8 pieces of pepperoni on each slice and upon getting it you counting and maybe it only had 7 pieces per slice, or maybe it had 8 but one of the 8th was unusually small and it was a bit debatable whether or not it is fair to count it as a piece of pepperoni at all. Even though it may technically be a piece of pepperoni it doesn’t really add much to the pizza being so tiny. So you’re a bit annoyed but even when eating the pizza you’re having a hard time noticing the difference. Maybe you go make a fuss about it on principal but for anyone who just wants some good pizza nothing has really changed.
I think you are looking at
I think you are looking at this in a wrong light. I believe most people buy video cards based on performance reviews rather than on what the manufacturer states. I bought a card that produces X average frame rates in a given title and Y minimum frames in a certain title, for instance.
So I don’t care if the pizza was cut in 7 or 8 slices, it is still the same amount of pizza.
Obviously, Nvidia publicised wrong information about their product, which is in its own a crime against its customers and should be pursued.
The reason many people don’t really care about this is that the cards perform the way they knew it to perform before they bought it.
How many days did it take to
How many days did it take to see 4GB 760GTX after launch?
Answer: 1 day
960GTX? still awaiting.
Where is our the 970gtx exchange process?
I haven’t bought ATI for the past 15 years because i felt they screwed me over with their customer support.
Nvidia is playing with fire.
Guys, this is fantastic data,
Guys, this is fantastic data, but it would be so much better if you use more contrasting colors than slight variations of blue and green in future articles. I can hardly tell which line is which. It would be much easier to eyeball whats going on if the 970s were shades of red, the 980s shades of blue, etc.
As far as the meat of the article – I still have questions. So the conclusion here is basically that it doesn’t matter. I guess what I’m wondering, is how is that even possible? There’s no question that the last segment has less bandwidth. What if it were 1GB in the slower partition? 2GB? It can’t be that bandwidth doesn’t matter at all, so how are they basically breaking the rules like this?
Also, what happens with games like shadow of mordor that use the VRAM as a buffer for textures, etc? In some sense by inflating VRAM with a huge frame buffer you’ve created a scenario that doesn’t reflect reality, especially in the single card tests. No one is running BF4 at 6K on a 970. No one. What I’m most curious about is what happens 1-2 years from now when we’re dealing with games that run in 1080p and still easily soak up 4GB of VRAM with epic amounts of textures, shaders, etc… If this is based on an algorithm figuring out what to keep in fast memory and what to keep in slow memory, what happens when NVIDIA stops supporting that algorithm for new games?
The assumption is that in a
The assumption is that in a few years, if games really take 4gb of vram on 1080p or some other resolution, the performance will be comparable to what shown by these tests. What will actually happen, nobody knows.
Even now games handle their memory differently from each other and will keep doing that in the future. Since nvidia supports their products quite a long time with new drivers, I don’t see why they should stop evolving the algorithm, if there truly is some. The hypothesis was that the OS handles the matter.
In that case I think it’s
In that case I think it’s worth investigating what happens with the games that people have reported issues with. I think this article clearly shows that the memory segmentation can handle huge framebuffers, and can deal fairly effectively with large texture…in a linear games that probably never refer back to cached textures. The games I’m most concerned with are open world games where this is a much larger degree of random access. In my recent experience with far cry 4, watch dogs and AC IV, there was a persistent, unavoidable stutter. Maybe that’s just ubisoft…but maybe it was my 970 too. I’m still very suspicious that this memory segmentation is causing ill effects in certain games.
I bought my 970 on launch day so I’m way outside of the retail return window. Had I known it had a bizarre memory setup I might have sprung for the 980, so I’m very, very frustrated with NVIDIA’s lack of response on this. Whether or not it’s a major issue, this wasn’t something that should have been withheld from buyers, and common sense tells me you can’t cut bandwidth and not suffer in some way. I stick with a single card because I’m completely turned off by the SLI microstutter – so I very much feel like I’ve been directly cheated by this misinformation if I’m stuck with a card that’s prone to stuttering, even in rare situations.
In 2 years most AAA games
In 2 years most AAA games will require 8GB so 3.5 or 4GB makes no difference.
In 2 weeks you wont need any
In 2 weeks you wont need any card! Think Grid.
Thank you for the excellent
Thank you for the excellent review Ryan. I feel very well informed about the limitations of the 970GTX and do not think that I will foolishly put them in a configuration in which they would fail.
My only disappointment was for COD at 1440p, I could care less about the game but I’m wondering if it isn’t badly optimized as well. I enjoy playing BF4 and the functionality and features of the Frostbite engine are great, then I see this review for COD performance and think “what could possibly be taking this many resources?”
Thanks again Ryan 😀
Different game
Different game engine.
[Forbes] ‘Call Of Duty: Advanced Warfare’ Will Feature Brand-New Engine
http://www.forbes.com/sites/davidthier/2014/05/06/call-of-duty-advanced-warfare-will-feature-brand-new-engine/
Its not foolish in way it was
Its not foolish in way it was done in such that they had to push settings of game to such extent to show where that fail limit is. Most people don’t run BF4 at settings which pretty much Ultra settings @ 6k rez. It just shows that issue is not as big as a lot of people have claimed it as and most games that buy say single 970 for 1080p gaming won’t see an issue. 970SLI, yea they are more likely to see a problem but still takes a bit and you least as of games now have to make it fail.
The market for used gtx 970
The market for used gtx 970 gpu’s remains very strong on eBay, so the “loss” seems to be in the 20% range including selling fees. The more difficult dilemma remains whether to spend $550 for a gtx 980. If you are currently experiencing issues, it may justify trading up, otherwise save your powder for newer technology down the road.
Very informative article
Very informative article Ryan, thanks. Still on a GTX 780 here. I’m thinking of getting another for SLI while they’re still available.
The alternative is to wait for a GM200 behemoth card like the rumored Titan X/II.
Some users reporter
Some users reporter “stuttering” ( More tha usual ) in Multi Gpu Setups with GTX 970.
I’ve read the entire review and there’s “Nothing” related to stuttering in GTX 970 Multi GPU Config.
My question is:
GTX 980 SLI, presents stutter ?
GTX 970 SLI, presents stutter ?
I don’t know if some users just reported Stutter in SLI setups because they got angy or whatever …
GTX 970 Will problably have its cost reduced due to the memory issue, so, I’m considering take another one.
Does anyone can help ? This question is not clear in the review …
Above both cards show stutter
Above both cards show stutter as they try to push over 4GB, but that is expected. The 980 actually has more severe stutters as it pushes over 4GB. The 970 has more of them.
Under 4GB there are no stutters, but they have the usual frame variance issues. The 970 has more variance, as cheaper cards tend to have. Less variance then AMD, I’ll add.
Perhaps some imagined it, perhaps some are sock puppets, as a AMD sales pitch seems to follow. A lot of people got the 970 and a lot of people could be new to SLI. You need more CPU power for it as you need to double up the draw calls and double up transfers. It’s very easy to start to have driver issues.
Windows isn’t a realtime OS, so you get stutters from driver interrupts. There are a few tools to test for that. DPC latency checker and LatencyMon. There could even be issues with Nvidia’s driver overhead. Those who were very insistent on having the cards, having issues and explaining it, did seem to have system stutter. That is the FPS wasn’t dropping, but the game was stuttering. I had some check and they did have high latency.
I had a bitch of a time with that on this build. The Asus software and drivers were a no go. I needed to use the default windows stuff.
AI suite II was the worst offender. I couldn’t even watch video or play audio with that installed.
This has been a growing issue with each new version of Windows. It’s getting harder not to stutter. No BSOD’s now, but PC’s can get turrets syndrome. This can put you into xperf hell as you dig through it.
This whole fiasco has really
This whole fiasco has really made me question the ethics of this site. I’ve used this site as a resource for years and never once questioned them. Now I realize that PCPER will defend Nvidia in order not to lose the close relationship they have with them. Pretty disappointed right now….
Honestly, and this is coming
Honestly, and this is coming from someone with a 970 that no longer wants it because of this….people are being totally unfair to PCPer with statements like this. I don’t think they’re defending NVIDIA or making excuses for them. They’re presenting the data they independently collected, and calling it as they see it. If there’s an issue here, its a small one and they’re being realistic about it. If there was a bigger deal to be made they’d have made it, just like they did with the 840 EVOs.
They clearly have a friendly relationship with NVIDIA, but they do with AMD as well. As they should. I don’t think it’s their job to have a default antagonistic stance towards the industry and make a mountain out of a molehill. The 970 is still a great card, what this comes down to is whether or not this should have been marketed as a 4gb card vs a 3.5gb card. I personally think the latter but there’s reasonable arguments to be made on either side. On the podcast they all agreed NVIDIA F-Ed up, they’re diving deep to see how bad they F-Ed up…what more can you expect from them?
I somewhat agree with you,
I somewhat agree with you, but that COD data looks absolutely atrocious. FPS is pretty good, in a playable range really, but those frametimes are just awful.
And? Hardly a negative thing to say about it. Again it is more excuses, not very impartial imo.
Did you not read the closing
Did you not read the closing comments? They’re being pretty plain about there being an issue and they’re directly attributing it to the VRAM. Even to the point where he’s saying that the data doesn’t even reflect the extent of the issues he was seeing in COD. That doesn’t sound like making excuses to me, if anything it’s the opposite. I mean the 970 is still a great card. It really is, I’m playing on one right now and it’s every bit worth the money I paid for it, despite the fact that I’m not happy about the VRAM either and never would have bought it if I knew.
Of course they didn’t read
Of course they didn’t read the closing comments Mark D. They hear what they want to hear.
Ryan could say that the GTX 970 was the worst card he had ever reviewed but the AMD Fanboys would still claim that he was being soft on Nvidia.
I can’t believe the writers
I can’t believe the writers of this is or their apologists. NVIDIA actually and intentionally fudged the card to make it slower so that they could sell 980s. Doesn’t anyone see this? Does no-one understand oligopoly power? You can do it because there’s nothing to compete. Simple. It’s about money. Same as Coke and Coke Light. Why is Coke Light not cheaper than Coke when the largest cost (sugar) is left out? Because their marketing department convinced the public that the lack of sugar or substitute for it was worth the additional price. Result – profits soar.
When companies renege on ethics and are defended because “the product is good enough anyway” we lose the point. NVIDIA acted unethically – finished. They purposely dragged the performance down and sold it as a higher spec’d article. Did they lie completely? No, there is indeed 4GB on board, so no lie. But no mention of the split in performance between the 3.5GB and 512KB until some clever guys figured out what was happening.
Then the excuses and the tech jargon from NVIDIA and buddy-buddy with PCPerspective and other sites – because they have the money essentially to buy votes.
The 970 is a great card. No doubt. But it and NVIDIA are tainted now. This is a reputational issue – can we trust NVIDIA going forward. One thing NVIDIA has learnt is that the community aren’t as stupid as thought. Scams will be found out and fingers pointed back and difficult questions asked.
I hope AMD makes a meal of this very stupid blunder NVIDIA made just to charge a higher price on the 980. If it hadn’t have purposely restricted the 970 you would have noticed no visible difference between the 2 and gamers would have chosen the cheaper option on a dollar for performance basis.
I dont understand what you’re
I dont understand what you’re insinuating.
NVIDIA disabled failed blocks of a full 980 chip to make a 970. That’s a normal and non-controversial practice across the chip industry, so that can’t be it.
They developed a new process to salvage more of the chip than possible with previous methods…so that’s questionably good for users, depending on the impact and how it done. Whether or not it’s better to have a compromised 512mb partition, or just have a 3.5GB or 3GB card instead…that’s the debate. Those are your choices. There’s no physical way to make a 970 as it is with a full 4GB. It was either have a compromised 4GB or less memory, period. Either way everyone agrees they shouldn’t have hid this info.
But they did. They essentially inflated the specs of the 970, making it look closer to a 980 than it is. That’s not going to sell more 980s, it’ll sell more 970s.
So either you’re saying something that doesn’t make sense, or you’re not understanding what’s going on here.
Your Coke comparison is
Your Coke comparison is pretty garbage and takes away from your argument. Distribution is where the cost is, not the cost of raw materials.
Nvidia is tainted huh? AMD is
Nvidia is tainted huh? AMD is not so squeaky clean in their history of things either. If you look back over last 3-4 years, AMD’s history is pretty dirty. Prime example is their complete lack of trying to fix crossfire stuttering issue that people repeatedly complained about for years.
How so? My 329$ GTX 970 is
How so? My 329$ GTX 970 is 6.5 inches long, and can maintain a 1390 mhz oc all day at 75c with a quiet fan profile. At that point, I’m beating my friends much more expensive r9 290 and GTX 780 (bought just a few months ago). You would have to be an idiot to even try to hate in this thing. It’s not like Nvidia fabricated some giant lie or something (like amd did when it posted official clocks for the blower cooler 290x). The small team who typed up the technical specs for the reviewers mistyped 2 small details, and no one noticed BECAUSE IT BARELY MATTERED. The benchmarks are perfectly accurate, and they show this card absolutely justifying it’s cost all day. What’s your problem?
Just because some issues arise in situations that are best described as stupid (it was well known at launch from benchmarks that if you thought 4gb cards this powerful were gonna be enough for 4k gaming, you were gonna have a bad time). As a 970 owner who sees vram usage above 3gb in a couple games with smooth framerate, I personally believe having 2 of these (which should mandate twice the vram to cover twice the power) is about as useful as running 4gb of system ram in a mainframe, and you are an idiot (or just have money to spend) if you are running two 970s in your system in the first place. My 1440p performance is excellent, though I plan on upgrading again in a few months when I make the jump to 4k since there really isn’t a setup today aside from a couple 8gb 290x cards that can take the load.
ROPs and Cache were falsely
ROPs and Cache were falsely advertised five months. I doubt they found out about that when things starting hitting the fan. Selling 970 with false specs was better than correcting it when they were selling 1+ million of them. Maybe only 2% of the people that have the card knows about the problem. That was great for business. The 970 is the best card out there for the money (my opinion). I think people are questioning Nvidia’s ethics more than the card.
Oh shut up already. You had
Oh shut up already. You had your 15 mins and it’s over. Less than 10k petition sigs out of over 1 mil cards sold- that means people are happy with the performance as is, and couldn’t give a crap about .2 L2 cache and some goddam ROPS that are meaningless to everybody but weirdo geeks who want attention.
Nvidia will do fine. They will not be hurt by this one bit. When they release their new best ever card, the same idiots around on these forums for the past few weeks will be first in line to buy them. Let’s be real.
I’d say that PCper was
I’d say that PCper was tilting in the direction of you crazies. The data showed no issue with the VRAM. You are proved wrong over and over again. You response is to simple make up another story and claim that people are conspiring against you. Reality is reality.
Please seek the help of a qualified mental health professional. They should be able to put you on an effective program of treatment.
Typo on Closing Thoughts,
Typo on Closing Thoughts, “…differently. Yes the GTX 970 has fewer shader cores and runs at slightly lower clocks than the GTX 970, but…”
Nice work, keep it up, cheers.
Typo on Closing Thoughts,
Typo on Closing Thoughts, “…differently. Yes the GTX 970 has fewer shader cores and runs at slightly lower clocks than the GTX 970, but…”
Nice work, keep it up, cheers.
“The 970 is still a great
“The 970 is still a great card, what this comes down to is whether or not this should have been marketed as a 4gb card vs a 3.5gb card.”
You had me on your side until you said the above. 4GB on the 970 is not the same as 4GB on a 290/X. It just doesn’t work that way. If this was simply a matter of density, we ALL still would have bought the card because the reviews didn’t lie. What the reviews lacked were frame time tests. If they were used by more sites on all their titles, people would’ve been able to see how much difference the two memory pools affected performance a lot better.
“The 970 is still a great
“The 970 is still a great card, what this comes down to is whether or not this should have been marketed as a 4gb card vs a 3.5gb card.”
You had me on your side until you said the above. 4GB on the 970 is not the same as 4GB on a 290/X. It just doesn’t work that way. If this was simply a matter of density, we ALL still would have bought the card because the reviews didn’t lie. What the reviews lacked were frame time tests. If they were used by more sites on all their titles, people would’ve been able to see how much difference the two memory pools affected performance a lot better.
Okay, well now I have to know
Okay, well now I have to know how running CoD (and maybe other titles) with various AA modes affects frame times at high resolutions. Hehehe.
test FAR CRY 4 e SHADOW OF
test FAR CRY 4 e SHADOW OF MORDOR, please.
Ok, the only problem i have
Ok, the only problem i have with this review is that only 2 games were looked at in the review. We need to be looking at “Next generation titles” like Shadow of Mordor, Dying Light, Metro last light or 2033 redux. Any of the games that are direct x 11 capable and tax your system already out of the box. Thats where you will see even more issues, due to them pushing your system to the limit.
I own 2x GTX 970 G1 gaming editions, and i have huge issues with the latest games with this memory issue. I have run many tests myself and besides the terribly optimised games, most new games are causing use issues in SLI on my system with stuttering and frame time variances, whereby it because extremely noticeable and irritating.
I am also running a Nvidia Surround setup of 4820 x 900. This compounds the problem and makes this issue more noticeable. My battlefield runs great with everything maxed out and resolution scale at 130%. But every other next gen title struggles when it hits the 3.5gig point or over.
Wold it be at all possible to do more tests on some other titles? Would love to put this to bed once and for all.
Thanks for all the hard work by the way! Been waiting for a good SLI review on this issue since it broke.
You’re running a surround
You’re running a surround monitor setup (resolution of 4820×900) with GTX 970 SLI and wondering why the latest and greatest games don’t run well?
Exactly what games and what settings?
I honestly feel like there are a ton of 970 owners who are now complaining because they don’t have 980 performance after hearing about this “issue”.
No, not at all. I had been
No, not at all. I had been experiencing these issues before it even broke about the vram issue. I kept wondering if it was just my SLI setup and poor profiles. I am running Shadow of Mordor with everything maxed. I used to have one gtx 970 and it would run the game flat out at around 30fps. But it felt like 30fps and was playable. I put in my second card and now it runs the game between 45 and 55fps, but it stutters to death and the game play feels like 15fps. Try to do a camera turn with my mouse 360 degrees and performance just tanks.
I do realize that Shadow of Mordor only just had an SLI profile added on this last update so i thought ok maybe its that.
So i tried farcy 4, it is a lot less apparent when it comes to stuttering but i do get random frame time variences and it once again feels way slower than the 50 – 60fps i am receiving in game. I am running everything flat out with smaa or fxaa.
I then got Dying light. Same thing happened when running it maxed out. MSI after burner say 3.5gigs of ram and there is stuttering to all hell. I thought the new patch would help and it did a bit but it still doesn’t feel like i am getting the fps fraps says i am getting. Also get random slow downs.
I am doing my best not to make this up in my mind, as i payed a lot of money here in South Africa to get these cards, and was almost broken when i heard about the Vram issue after having experienced these random problems before reading the full story.
Gtx 970 performance in all reviews are stellar! So i dont understand why any game that uses the “max” Vram of the card wouldn’t feel like Sli gtx 970s.
Good example would be my sli Gtx 670 OC windoforce 2gig cards i used to have. Sure i ran out of Vram in a second but the performance i got out of them was rock solid! that’s what made me go straight out and buy 2x gtx 970’s.
I don’t expect gtx 980 performance from these cards, because they are gtx 970’s with different clock speeds and now Vram, however i do expect them to run without random anomalies in games that makes your actual game play feel far worse than your fps reader is telling you it should be.
Dying light would be a
Dying light would be a terrible game to test on, it has some kinda bottleneck issue that it doesn’t utilize SLI properly so FPS barely is any better on 2 cards compared to 1. You can turn off most graphic options with minor fps impact but you mess with “draw distance slider” it is death to fps but yet draw distance is barely effected by it least for most game what you see.
Some of those games have been
Some of those games have been shown to have high Drawcalls. This is 2X of a problem for people with SLI/Xfire.
Any sort of a driver problem will tend to show up as stutter without the FPS dropping. The game animation starts to stutter. Agian SLI can compound this.
I had this problem and Fuck Asus and their shitty drivers and software. I’m still mad and I still have some issues.
Are you guys done with
Are you guys done with testing the 970 issue ?
I feel that conclusion that problems only arise at effectivly running 6K is a bit dishonest as Battlefield 4 nor CoD dont use that much VRAM anyway compared to some other games. Shadow of Mordor might suffer from it already in some dual monitor/4K monitor cases so would love to see testing with more demanding games on VRAM.