NVIDIA Stuns Investors With Record Earnings

@04dcarraher said:

Again it has hitching issues are on all systems with gpus with 4gb when using ultra settings in certain areas. and again not all benchs posted are using overclocked 980ti's.....

they cant fix the underlying issue of the games allocating 4gb+, your shadow of mordor example is based on youtube videos and no real benchmark, and even after Crimson driver new games are seeing the same issue.....including ROTR

Fail. Absolutely worthless benchmarks as there was done using the older drivers. As PCPer pointed with the new drivers is significantly better:

"After some comments from readers that I should try the updated AMD driver released with Rise of the Tomb Raider fixes (16.1.1), I started to boot up the game again this morning to find the patch basically changes all of our previous results. (Nice!)" This is pretty much in line with what HardOCP found.

Here are the new benches when compared to the old one as the Frame Times is practically gone:

As for Shadow of Mordor. It's common knowledge with Fury X owners that stuttering issues has been resolved on updated drivers long time ago.

Keep on claiming that Crimson fixes everything...... Your still ignoring the inherit issue of Fury. Even with The game update with limiting Fury's buffer to below 4gb there are still spikes, not as prominent as before .

"At 4K the results look very similar - performance for both the GTX 980 Ti and the R9 Nano are unchanged from the day one version to the February 5th update. The only significant change is the loss of the hitches in the R9 Nano results, bringing the 95th percentile frame time down to 1.5ms from ~4ms."

Crossfire the issue is very prominent......

Why shouldn't both cards be compared at their fullest potential?

Just because the Fury X does not overclock well, doesn't mean that it should only be compared to the 980ti at stock speeds. If it was the comparison would be false, especially when you consider the majority of 980Ti's available are overclocked to one degree or another.

@04dcarraher said:

Keep on claiming that Crimson fixes everything...... Your still ignoring the inherit issue of Fury. Even with The game update with limiting Fury's buffer to below 4gb there are still spikes, not as prominent as before .

"At 4K the results look very similar - performance for both the GTX 980 Ti and the R9 Nano are unchanged from the day one version to the February 5th update. The only significant change is the loss of the hitches in the R9 Nano results, bringing the 95th percentile frame time down to 1.5ms from ~4ms."

Crossfire the issue is very prominent......

That was the point I was making not noticeable as per anyone who has used the latest drivers who has played the game on the Fury series not the crap old a$$ benchmarks you were using. As for Crossfire the newer 16.2 drivers was supposed bring additional improvements to RoTR. I hope to see better improvements as AMD recently hired Scott Wasson from techreport who started last month at AMD and who was one of the pioneers of doing frametime analysis on their reviews. Raja Koduri the new head of the Radeon Techologies Group is clearly focused on improving the driver side of things as per the Crimson launch.

I would love to see more custom board partners with better ASIC and better voltage control release of custom Fury X's but unfortunately AMD didn't allow that like that they did with R9 290X's. Also, there is the matter of price difference, you should be getting better performance if you are spending $70 - $120 more. You really need to use Trixx to get voltage unlock and to overclock the HBM (remember HBM was locked down by AMD engineers as you couldn't overclock it). Now with trixx you can get a descent 15% - 20% increase:

You clearly missed the point, showing that Crimson drivers have not fixed the inherit issue with Fury, even with latest driver at tests. Suggesting using old ass drivers? really? lol, both brands were using latest drivers at test vs that skewed example with hardocp using latest AMD driver but not Nvidia.( they had three days before test to get driver). Even with new beta drivers issue is there but much better then it was before.

Problem is that AMD can not fix Fury's issue of not having enough HBM with games wanting to allocate 4gb+. All AMD can do, is fiddle with minimum framerates with games allocating 4gb+ to make the frame jumps less noticeable. Or get dev's to set specific vram limits for Fury. ROTR is one of those examples seeing Fury allocating around 70mb less than any other 4gb gpu staying below the threshold.

Now regarding Fury overclocking...... 15-20% is rare . Vast majority of users cant even get ~10% stable overclock.

That 3,5Gb marked at 4Gb vram hurt them tho

The card still has 4gb of vram ,

Normally companies disable the defective parts of their product and sells them as lower-end units. In the past this meant reducing memory bandwidth and sometimes memory capacity. However the engineers got a middle ground allowing more to be done, by tweaking the architecture so that the disabled component/s could be helped by the other pool of memory. This meant dividing the available memory into a full speed 3.5gb section and a slower .5gb section.

If this was explained before and at release there would have been a lot less of an issue.

You did use old a$$ drivers to make your point that Shadow of Mordor has stuttering issues which was LAUGHABLE as anyone who has been following driver updates would know that issue was resolved LONG TIME AGO. And since you mentioned YouTube I did a quick search which pretty much confirms what Fury X owners were saying on owner's thread with respect to Shadow of Mordor running at 4K maxed out with no stuttering. As for the RoTR you tried to pull a fast one by using benchmarks with the older drivers where in the same article they stated the older benchmarks were invalid as the newer drivers "completely changes" things according to PcPer which now show significant improvements in Frame Pacing which now shows that the stuttering issues has been resolved with the newer drivers on RoTR which pretty much was confirmed by HardOCP in their benchmarks too.

As for 4GB memory allocation issue, I have said time and time again that they have assigned 2 engineers to do memory allocation of games that push 4GB (this is not something new) and that is what they have been doing. So, that is a good thing. You want AMD to keep their words.

As for the overclock, yes you are right about using non-voltage overclock and as a standard it's around 10% but you really need to use Trixx to get voltage unlock and to unlock the HBM so you can overclock it to get the best out of the Fury/X which from forum members and Youtube videos I have seen it's around 15% - 20%.

16.1 and 16.1.1 were the latest drivers used when each of those tests were done for the game.And again 16.1.1 still has slight frame time issue as shown by Pcper. Using hardocp is funny since they are using older drivers with Nvidia(while newer ones were out) but using the hotfix 16.1.1 drivers for AMD. That test voids directly comparing the two fairly.

Ignoring fact that AMD can not fully fix Fury's issue. Dont blindly believe that those two engineers can solve the issue.....

All they can do is soften the blow. Its because how a game will allocate vram, most games use as much is available VRAM for their own use with the intent of ensuring they have enough VRAM for future use, and otherwise caching as many resources as possible for better performance.

"To be clear here, there’s little AMD can to do reduce VRAM consumption, but what they can do is better manage what resources are placed in VRAM and what resources are paged out to system RAM. Even this optimization can’t completely resolve the 4GB issue, but it can help up to a point. So long as game isn’t actively trying to use all 4GB of resources at once, then intelligent paging can help ensure that only the resources that are actively in use reside in VRAM and therefore are immediately available to the GPU when requested."

Like I have been saying as long as games manage memory smartly keeping usage below the 4gb threshold Fury has no issue. As time goes on the Fury with only in regard to its 4gb will be its Achilles heel as time goes on. 980ti is the the better gpu over FuryX while Fury is the better gpu over GTX 980.

I wasn't talking about nVidia drivers. I was referring to your nosense about Rise of the Tomb Raider having stuttering issues when you initially posted with the old drivers which has since been fixed and has gotten now better with 16.2. Nothing wrong with Anandtech stated. I completely agree, the did intelligently paged to system RAM with no issues. Not with the so called latency you stated would have happened if they used System RAM, another BS you made up, as HardOCP showed with no issues with using System RAM because they were using a Core i7 6700K processors with 16GB Dual Channel DDR4 RAM which is plenty fast for paging. I have the same specs as the HardOCP test bench, so I am not really worried.

With respect to the 980 Ti vs Fury X I would agree, it would come down to the price. It's now $50 - $60 cheaper than the 980 Ti and that's why I brought it, I see no reason to pay the extra $50 - $60 for an extra 2GB and to get the same performance not to mention it runs cooler and quieter which was icing on the cake (not to mention it looks pretty slick with the Radeon logo lighting up and with GPU Red LED tachometer showing GPU load which is pretty cool if you have a see through case). And the memory primarily would affect at 4K which lets be honest is not practical for even the 980 Ti, Fury X or the Titan X. I plan to game at 1440P which is plenty fine with 4GB as that is the "sweet spot" for the Fury X. You really need to wait till next gen on the new node to get 4K 60 FPS.

What you seem not to get is that there is a problem with 4gb HBM with Fury when there is not enough buffer. Its on a game by game driver by driver fix that they have to adjust and fix to soften the blow of the inherit issue continuously .... And even with later drivers with the workarounds there is still higher frametimes vs other gpus. Also you cant ignore the point of the games actively allocating all 4gb or more. lol if you think that a gpu having to constantly dump data to make more room for new data does not get hampered by the sudden drop in bandwidth having to access slower memory ie DDR3/DDR4...... Why do think there is massive frame pacing with fury until a fix/workaround for vram allocation is issued by game patch or driver......

If there is a only a $50-$60 difference in price....... 980ti is still the better gpu, its overall faster, has more memory, and you have more options. But your wrong with the memory only affecting 4k, it effects games with 1080p/1440p and any other resolution in between when game assets are being allocated using 4gb+. As time goes on Fury will suffer more than 980ti with sub 4k resolutions. Fury is a good gpu but AMD waited too long to release, 1st gen HBM vram limits and its costs hurt the gpu. If they used 8-12gb 512bit GDDR5 for Fury it would have been better. Until HBM matured. Fury was a test bed for HBM.

The 980 Ti is a brilliantly crafted card. No doubt about that.

It can't however do asynchronous computation very well, which is the future. But again, no games really use asynchronous computation, so buying a 980 Ti is probably the safest bet at this point for someone who is looking to build a new PC.

But we're approaching a turning point. Nvidia needs to come up with a solution to improve their asynchronous computing capabilities BEFORE they start to become adapted in mainstream games, which, make no mistake, they will.

That's the one I want in my next build.

If you're not in a rush, might as well wait for pascal

Yeah, I was thinking about it. I was going to build a new PC this year, but I may push till next year.

@Heil68:

Early next year is the perfect time.

You will have Intel and AMD in the 14 nm space. Intel is already there and it looks like AMD's upcoming Zen line won't be another Bulldozer (bulldozer is the chipset they've been stuck with since 2011 and it was terrible, really hurt them as a company).

You will have NVidia's Pascal line as well as AMD's Polaris architecture. We know nothing about Pascal, but if it's as good as their Maxwell architecture, then you'll have some difficult choices to make.

:)

I usually go back and forth with GPU's and right now I have a 7950, so next build will be Nvida with an I7.

@organic_machine:

Maxwell is able to do Async Compute, its not the same way as AMD does it. Maxwell's architecture has one graphics engine and one shader engine, capable of doing 1+31 queues very quickly. Tests shows that Maxwell can match GCN with 8 ACE units with 64 command queue, while using 128 command list it runs roughly the same as 64.

@04dcarraher: not many games use 4GB memory especially at 1080p and 1440p which is my target. And based on hardocp and other benchmarks shows no issues like you claimed it does when accessing system RAM which was my point, GDDR4 is plenty fast. It really is an issue at 4K which really isn't ideal gaming with 4K even then AMD is fixing with driver updates, really 1 - 2 games being affected which is hardly much which AMD fixed.

With respect to pricing vs the 980 Ti the 980 Ti would had to be a lot better instead of just reaching parity with the Fury X. Rather would pocket the extra $50 bucks and buy games. If it was something like 8GB HBM 2, with 2x the memory bandwidth with significantly better performance than the 980 Ti not similar performance then I would gladly spend the extra 50 - 60 bucks maybe even a 100.

Not that I actually looked into their earning, eps, or their actual financial statements, but they were acquiring a lot of companies in 2015

Basically, I'm sure someone with an accounting degree can give a longer explanation. I only have a finance but when a company acquires smaller company, they also take the debt too. When you take debt, you can do a lot of shady stuff that perfecto ally legal to skew with eos and earning. That's actually a very popular method and one that many companies do. I believe over 80% of top companies did it in 2014.

@Xtasy26:

With poorly memory managed games even at 1080p 4gb isnt enough which is sad, but at 1440p we are starting to see 4gb start to not being enough with max settings. As time goes on as game assets continue to get more detailed 4gb wont be enough for 1440p or higher.

Even with DDR4 your still going to see issues and continue to see same affects DDR3. Because the problem lies with PCI-e lane, PCI-E 3.0 at x16 only allowing upto 16GB/s which translates to only 16,000 MB/s which is 4000MB/s less than 1600mhz DDR3 in dual channel. Which means that even with DDR4 that is able to do 5-10 GB/s more than DDR3 in dual channel or in quad channel able to do 20-30 GB/s more than dual channel DDR3, the PCI bus will cut the bandwidth going to GPU. Which will lead you to the same affects.

Yeah if Fury had 8GB it would be a beast of a card and would be the better option but it didnt happen sadly.

@lamprey263: "Oh great, the Lex Luthor of video cards made a profit through their dirty tricks, I'm not surprised."

This

It depends on how intelligently they handle memory management as stated by AnandTech which so far they have done good job. I will look forward to seeing if they continue to do it going forward. So far, with the updated drivers they aren't running into issues.

With respect to 4GB, I have really seen that with respect to 1440P/1080P. Maybe as I stated one or two games, which is not much to go by. I will have to see how the this year and the next pans out. I could definitely see major issues at 4K which double the resolution at 1440P going forward in the next 1 - 2 years but then I am not going to be gaming at 4K.

Also, another reason choosing the Fury X over the 980 Ti is that I am seriously planning to upgrade to a 1440P FreeSync monitor if I go with G-Sync is going to cost me an extra $200 that's a WAY high too of a price to pay for essentially the same thing.

What companies did they actually acquire? They have very little debt with respect to the amount of money they make. I see 0 problems with them acquiring companies.

All this talk of 980 Ti's and Fury X's...and I'm just here with my little 960...