Support NeoGAF

[Unknown Soldier]

Nvidia fanboy logic. Facts don't matter.

GTX980 BARELY beating a previous generation AMD card.

Also, I own cards from both teams. I just tell it like it is. For now, I definitely looks like my next card will be an AMD. Of course I don't know what the future would hold.

"I own card from both teams" is the videocard equivalent of "I have black friends" lol

AMD's "current" generation is rebadged last generation with the exception of Fury so your rhetoric is hilarious

The 980 was never slower than the R9 290/390 so I don't even get what you're arguing here

 

[Herne]

I can't say I'm surprised.

 

[fred]

Of course those thinking about getting ANY GPU for DX12 benefits may want to remember that without a half decent CPU those benefits might not be much to write home about.

 

[Grassy]

for 980ti owners here, what are you planning to do? See what happens? Trade up now?

I just bought one and I love it. Best gaming product ive ever owned but I'm so confused right now.

I'm going to keep my cards and wait for Nvidia to mature their drivers for DX12. I'm not going to carry on like a pork chop because Nvidia and AMD cards are close in a DX12 benchmark for a game currently in Alpha.
Basically, the benchmarks are comparing a mature AMD driver to an immature Nvidia driver.

You have to understand that Async Compute is a feature that is intended to increase performance. The problem is that Nvidia GPUs scale negatively with Async Compute: Using this features reduces the performance of Nvidia GPUs. A 2013 AMD GPU being on par with Nvidia's 2015 top-tier GPU is a complete desaster. What Nvidia wants to do now is to optimize their drivers in order to prevent their GPUs to scale negatively with Async Compute. It's not about getting more performance out of Async Compute. It's all about to prevent negative scaling. This is not just a driver issue. It is an issue that is due to different architectual approaches. All Nvidia can do now is to do damage control. If they can't get rid of the negative scaling, Nvidia GPUs will be in deep trouble for the rest of this generation since GCN is designed for Async Compute.

The problem is that you're speculating to the nth degree and spreading FUD. The developer themselves said that they thought Async Compute was fully implemented in the Nvidia drivers, when it wasn't.
The ball's in Nvidia's court, let's wait and see how they respond before making melodramatic claims about their doom and gloom.

 

[Renekton]

http://www.neogaf.com/forum/showthread.php?t=1107677

Nvidia just wants to protect the environment by preventing us from furmarking our GPU!

 

[fertygo]

Ashes of Singularity would be more relevant at this fanboy warzone than the game itself.

 

[Renekton]

Ashes of Singularity would be more relevant at this fanboy warzone than the game itself.

No one buys Crysis 3 for the excellent gameplay.

 

[Amey]

Anybody has 980ti vs FuryX in Ashes benchmark ?

 

[Self_Destructive]

Anybody has 980ti vs FuryX in Ashes benchmark ?

They're pretty much dead even... even with Nvidia's deficiencies.

http://www.extremetech.com/gaming/2...-singularity-amd-and-nvidia-go-head-to-head/2


But now the excuse is that AoTS is only moderately Async compute based... and games that are heavily compute based will perform better on AMD.

 

[alexandros]

O
The problem was, that the aforementioned group of users wasn't able to prove me wrong with arguments. So they decided to do pretty much the same as you're doing now: They attacked and discredited me whenever possible which led to dozens of derailed threads.

That is your interpretation. The impression I got from you, based on your fanatical devotion to Mark Cerny and your utter lack of proof to justify your claims, was that you are Sony's misterxmedia. You seem to be knowledgeable and I would love to have a serious discussion with you on technical stuff, but it is difficult when you are so hostile towards PC gamers.

 

[W!CK!D]

...you're speculating...

Sure?

 

[Bastardo]

That is your interpretation. The impression I got from you, based on your fanatical devotion to Mark Cerny and your utter lack of proof to justify your claims, was that you are Sony's misterxmedia. (...)

I didn't read all of his posts, but just this one, where he explains basically HSA:

A CPU is a processor that consists of a small number of big processing cores.
(... lots of explanation...)
The only problem is, that PC gamers don't have a unified system architecture. The developers of multiplatform engines have to consider that fact. 1st-party console devs can fully utilize the architecture, though.

What he is stating there does not require actual proof besides knowledge about the architectures. The only things he is stating is:

About PS4

1. There is no latency to copy data from GPU to CPU memory, because it's one and the same
2. The PS4 can also work on the datasets with part of its gpu cores asynchronically.

These are facts, we know that the architecture is designed to do that. He also goes on to judge a little bit here and there and I leave that out now, because that's his opinion.

Async Compute allows point 2 of my list above. It makes it a bit easier for developers to code for the architecture. On the PS4 you can always decide: Do I want to calculate something on this dataset on CPU or GPU? On the PC you need to decide first: Is it worth it to invest the latency to copy the data first and then work on it on the architecture better suited for it?

An example I always do is Lara's hair in the first TR reboot: This is generated on GPU. Now imagine you actually wanted to do collision detection for the hair on CPU. You would need to copy over the large dataset of hair strains to CPU and then work on it. PS4 avoids that copy. In case of only ASYNC compute without a unified architecture you would still need to do this collision detection on GPU, but could at least do it, while doing something else.

 

[laxu]

Was this here yet? Nvidia will fully implement async compute via driver support. Basically it will be a hardware/software solution whereas it's largely hardware on AMD. This probably has pros and cons, remains to be seen how well it works.

http://www.guru3d.com/news-story/nvidia-will-fully-implement-async-compute-via-driver-support.html

 

[alexandros]

He also goes on to judge a little bit here and there and I leave that out now, because that's his opinion.

Then it should be clearly marked as such. I highly appreciate the technical insight, I don't appreciate drawing conclusions based on as of yet unproven technologies. Let's see what the actual benefits are in the real world, not in theory. Because in theory the console gpus were supposed to perform twice as much as the equivalent PC gpus and that is very clearly not the case.

 

[Bastardo]

Because in theory the console gpus were supposed to perform twice as much as the equivalent PC gpus and that is very clearly not the case.

It's not as clear cut as you think. The unified architecture allows for possibilities you don't have on PC. I could design an algorithm, which runs a lot faster on PS4 than a PC with 3 of the newest Titans, although the PS4 could never beat the Titans in rendering performance:

Example:
Imagine your CPU has an instruction set for encrypting data in hardware and you also wrote a very fast parallel algorithm to generate the data on GPU.

On your PC this algorithm looks like this:
GPU: Generate Data, 2ms
Send Data to CPU, 5ms <- no matter the GPU or CPU, you will not get this step significantly faster
CPU: Encrypt Data, 2ms

On PS4 it looks like this:
GPU: Generate Data (slower): 4ms
CPU: Encrypt Data (slower): 4ms

Even though the PS4 in my thought experiment is twice as slow as the PC, it executes the algorithm 1ms faster, because it eliminates the intermediate step. The only thing I want to clear up here is: Yes, the PS4 is globally slower and will execute any PC optimized algorithm with the same speed as an equivalent specced PC, but it also allows to execute algorithms very fast, which would have horrendous runtimes on PC.

Disclaimer: I have absolutely no proof for the following statement, it's just a theory.
One possibility explaining Arkham Knight performing so badly in the beginning on PC is that many effects used the GPU for computing on Console without accounting for the fact that this induces latency on PC and doesn't on console.

 

[Kezen]

One possibility explaining Arkham Knight performing so badly in the beginning on PC is that many effects used the GPU for computing on Console without accounting for the fact that this induces latency on PC and doesn't on console.

Well they must have figured out a way then, because it runs flawlessly on my PC after the recent performance patch.

 

[KKRT00]

Even though the PS4 in my thought experiment is twice as slow as the PC, it executes the algorithm 1ms faster, because it eliminates the intermediate step. The only thing I want to clear up here is: Yes, the PS4 is globally slower and will execute any PC optimized algorithm with the same speed as an equivalent specced PC, but it also allows to execute algorithms very fast, which would have horrendous runtimes on PC.

Theoretically You are right, practically there is not a single 'type' of gaming related task that would put PC in disadvantage when utilizing compute.

Batman AK is just bad programming.

 

[icecold1983]

Theoretically You are right, practically there is not a single 'type' of gaming related task that would put PC in disadvantage when utilizing compute.

Batman AK is just bad programming.

not true, there are various things you can do when programming at an extremely low level on the ps4 that just arent currently possible on pc. far cry 4s hraa for example.

 

[GhostTrick]

I can't believe some people would gladly take only one game benchmark comparison as the truth for all games.

It's not the first time a game or engine performs better on one brand than the other.
Take Ryse for exemple, which performed a lot better on AMD hardware, especially with lower end cards like HD7870 outperforming GTX680 or even R9 290x vastly outperforming GTX Titan or GTX 980:

While another exemple, Metal Gear Solid V, with GTX680 outperforming R9 290:

Basically, depending of the game or engine, you'll likely get different performances... and sometimes the same kind of performances in the same league of GPUs.

 

[Dictator93]

I can't believe some people would gladly take only one game benchmark comparison as the truth for all games.

It's not the first time a game or engine performs better on one brand than the other.
Take Ryse for exemple, which performed a lot better on AMD hardware, especially with lower end cards like HD7870 outperforming GTX680 or even R9 290x vastly outperforming GTX Titan or GTX 980:

Those were pre-ryse patch that increased the performance on NV. But still, that game deifnitely ran relatively better on AMD configs more often than not.

 

[GhostTrick]

Disclaimer: I have absolutely no proof for the following statement, it's just a theory.
One possibility explaining Arkham Knight performing so badly in the beginning on PC is that many effects used the GPU for computing on Console without accounting for the fact that this induces latency on PC and doesn't on console.

Or maybe because it was a shit port with very little QA, at the point it had ot be removed from sales.
I mean, it's the only exemple you can account for. There's a lot of games running the same effects on PC and on console. No, I'll say it better, a lot of games where you have better effects on PC than on console, effects running heavily on GPU computing such as particles and such.

Those were pre-ryse patch that increased the performance on NV. But still, that game deifnitely ran relatively better on AMD configs more often than not.

Ah, maybe then ! But my point was that one benchmark is never painting the full picture. And if we had to take only into account these Ryse benchmarks pre patch, you could claim that a lower end AMD GPU performs a higher end Nvidia one... same for Ground Zeroes (Phantom Pain having a better AMD optimisation), in which you could claim even a GTX680 outperforms the higher end R9 290.

 

[Quotient]

Even with all the arm-chair analysis in this thread I still plan on buying a Nvidia GPU - they perform much better for DX11 games. In 12 months time I'll re-evaluate the situation once we have more DX12 cards on the market and Nvidia has launched Pascal.

 

[GhostTrick]

Even with all the arm-chair analysis in this thread I still plan on buying a Nvidia GPU - they perform much better for DX11 games. In 12 months time I'll re-evaluate the situation once we have more DX12 cards on the market and Nvidia has launched Pascal.

What you plan to buys all comes down to your preferences. Performance per watt for some people, driver and software for some people, performance per dollar for others. On the DX12 point though, it'll be the same as DX11: Some games or engine will prefer one GPU architecture than the other.

 

[KKRT00]

not true, there are various things you can do when programming at an extremely low level on the ps4 that just arent currently possible on pc. far cry 4s hraa for example.

Thats not due to cpu->gpu 'issue' with not unified architecture, its just DX11 API limitation.

 

[Costia]

What kind of parallelism is this about? GPU core/workgroup level or cycle level on a specific core?
Meaning is this about :
1)
Utilizing unused GPU cores while another process/kernel is running on a few of them.
For example if you have 100 tasks and 70 cores. once the first 70 are done, without async you would have to wait for the remaining 30 to finish before running anything else, vs with a-sync, where you can use the idling 40 cores while the remaining 30 are still running.
or
2)
Utilizing unused cycles on a "busy" core - something like Intel's hyper threading. where 2 threads can run on a single physical core and each thread can use the other thread's stall cycles.
?

 

[dogen]

What kind of parallelism is this about? GPU core/workgroup level or cycle level on a specific core?
Meaning is this about :
1)
Utilizing unused GPU cores while another process/kernel is running on a few of them.
For example if you have 100 tasks and 70 cores. once the first 70 are done, without async you would have to wait for the remaining 30 to finish before running anything else, vs with a-sync, where you can use the idling 40 cores while the remaining 30 are still running.
or
2)
Utilizing unused cycles on a "busy" core - something like Intel's hyper threading. where 2 threads can run on a single physical core and each thread can use the other thread's stall cycles.
?

https://forum.beyond3d.com/posts/1835425/

 

[icecold1983]

Thats not due to cpu->gpu 'issue' with not unified architecture, its just DX11 API limitation.

pci bus latency still prevents a lot of things that would be possible on consoles.

 

[Dictator93]

not true, there are various things you can do when programming at an extremely low level on the ps4 that just arent currently possible on pc. far cry 4s hraa for example.

CSAA stuff like that is totally possible on PC hardware, just an API limitation.

pci bus latency still prevents a lot of things that would be possible on consoles.

When those things start happening and are obvious, please let us know.

So far the whole CPU / GPU wall breaking possible due to nexgen console APU has not been done yet in any game to my knowledge.

 

[dr. apocalipsis]

pci bus latency still prevents a lot of things that would be possible on consoles.

PS4 or Xbox One have buses too.

 

[icecold1983]

CSAA stuff like that is totally possible on PC hardware, just an API limitation.


When those things start happening and are obvious, please let us know.

So far the whole CPU / GPU wall breaking possible due to nexgen console APU has not been done yet in any game to my knowledge.

i said would be possible, it remains to be seen if that happens this gen. should consoles start using gpu compute to do anything that actually affects gameplay, pc will probably have to run that code on the cpu.

PS4 or Xbox One have buses too.

data transfer over those busses is so much faster than transferring data back and forth over the pci bus.

 

[Costia]

https://forum.beyond3d.com/posts/1835425/

So its similar to hyperthreading.

According to this:
https://www.pgroup.com/lit/articles/insider/v2n1a5.htm
and http://docs.nvidia.com/cuda/maxwell-tuning-guide/#axzz3kz31f89B

NVIDIA's SMM's support up to 64 warps per SMM. So the issue is that it can't do that while switching from/to graphics to/from compute? That each SMM has to be exclusively dedicated to either graphics or compute at any given time?

And apparently this is a driver issue?
http://www.guru3d.com/news-story/nvidia-will-fully-implement-async-compute-via-driver-support.html

Edit: also, in the current state, can maxwell run a compute task on one SMM and a graphics task on another in DX12? (this would cap the maximum performance degradation to under-utilizing a single SMM)

 

[dr. apocalipsis]

data transfer over those busses is so much faster than transferring data back and forth over the pci bus.

PCIe3=15,754 GB/s of real world measurable performance.
Onion='Almost 20GB/s' says Cerny.

 

[mordecaii83]

PCIe3=15,754 GB/s of real world measurable performance.
Onion='Almost 20GB/s' says Cerny.

I think latency would be the bigger issue, I'm not sure how they compare on that though.

 

[nico1982]

PCIe3=15,754 GB/s of real world measurable performance.
Onion='Almost 20GB/s' says Cerny.

I think latency would be the bigger issue, I'm not sure how they compare on that though.

Apples and oranges. Comparing Onion to PCIe means totally missing the point.

PC:
CPU <-- DDR3 --> RAM <-- PCIe --> VRAM <-- DDR5 --> GPU

PS4:
CPU <-- Onion --> RAM <-- Garlic --> GPU

 

[icecold1983]

PCIe3=15,754 GB/s of real world measurable performance.
Onion='Almost 20GB/s' says Cerny.

i was referring to latency

 

[gatti-man]

I just want to know what realistic gains im looking at for the 290x. I run a 4.4ghz i7 quad core (overclocked) so I'm hoping for some nice gains. Before this I was looking at upgrading but maybe I can wait another 2 years.

 

[dr. apocalipsis]

Apples and oranges. Comparing Onion to PCIe means totally missing the point.

PC:
CPU <-- DDR3 --> RAM <-- PCIe --> VRAM <-- DDR5 --> GPU

PS4:
CPU <-- Onion --> RAM <-- Garlic --> GPU

Actually not. Including the RAM type to make it larger was funny though.

Hint: Ps4 GPU, as Xbox360 GPU, can pass data directly from CPU to GPU.

i was referring to latency

Numbers, please.

 

[icecold1983]

Actually not. Including the RAM type to make it larger was funny though.

Hint: Ps4 GPU, as Xbox360 GPU, can pass data directly from CPU to GPU.



Numbers, please.

i think i remember it being somewhere in the vicinity of 40 to 50 ms. it was some time ago tho so i could be remembering wrong.

 

[nico1982]

Actually not.

Actually not what? You replied to icecold, implying that, because the PCIe bandwidth is comparable to Onion "Almost 20GB/s", the advantage of an unified architecture is moot. Overlooking the fact that copying data over PCIe, even at "15,754 GB/s of real world measurable performance", is an extra cost

 

[Spladam]

I kept telling people you could not make direct comparisons of the consoles hardware to the 7XXX and R9's before the R9 285, as the additional ACE's made a difference, something the Tonga added to the R9 series with the 285, and why the 285 gets comparible performance at a lower bit-depth then the 280 and 280X. Everyone kept telling me the ACE's made no difference.

 

[Unknown Soldier]

I just want to know what realistic gains im looking at for the 290x. I run a 4.4ghz i7 quad core (overclocked) so I'm hoping for some nice gains. Before this I was looking at upgrading but maybe I can wait another 2 years.

The longer you wait to upgrade an AMD card, the more AMD dies. If you love AMD that much you should buy an AMD product.

The problem with AMD's "fans" is none of them are actually willing to buy AMD products otherwise AMD wouldn't be dying. Meanwhile Intel and Nvidia have figured out how to get people to buy their products on a regular basis. AMD has the worst "fans" around if you think about it.

Not my problem though I guess, this whole DX12 flamewar is funny to me because I bought a 980 Ti this year and next year if Pascal is amazing I'll just buy a Pascal, if not I'll wait until the following year. I don't care that Nvidia designs cards to be amazing now as opposed to whatever AMD does because I upgrade them so regularly, the money spent on a hobby I enjoy is worth it to me.

 

[gatti-man]

The longer you wait to upgrade an AMD card, the more AMD dies. If you love AMD that much you should buy an AMD product.

The problem with AMD's "fans" is none of them are actually willing to buy AMD products otherwise AMD wouldn't be dying. Meanwhile Intel and Nvidia have figured out how to get people to buy their products on a regular basis. AMD has the worst "fans" around if you think about it.

Not my problem though I guess, this whole DX12 flamewar is funny to me because I bought a 980 Ti this year and next year if Pascal is amazing I'll just buy a Pascal, if not I'll wait until the following year. I don't care that Nvidia designs cards to be amazing now as opposed to whatever AMD does because I upgrade them so regularly, the money spent on a hobby I enjoy is worth it to me.

Wat. Before my 290x I had dual 680gtx's and dual 8800gtx's as well as a 7970. Money wise im a fan of Nvidia but truth is I just go with whoever fits my needs at the time. Like a normal consumer.

 

[Beerman462]

The longer you wait to upgrade an AMD card, the more AMD dies. If you love AMD that much you should buy an AMD product.

The problem with AMD's "fans" is none of them are actually willing to buy AMD products otherwise AMD wouldn't be dying. Meanwhile Intel and Nvidia have figured out how to get people to buy their products on a regular basis. AMD has the worst "fans" around if you think about it.

Not my problem though I guess, this whole DX12 flamewar is funny to me because I bought a 980 Ti this year and next year if Pascal is amazing I'll just buy a Pascal, if not I'll wait until the following year. I don't care that Nvidia designs cards to be amazing now as opposed to whatever AMD does because I upgrade them so regularly, the money spent on a hobby I enjoy is worth it to me.

Not everyone that owns AMD cards is a an AMD fan. A lot of people me included just buy what fits their needs and budget. I guess I can see how someone who has a lot of disposable income and only considers the high end could be dismissive of AMD, but this post is embarrassing. I never bought an AMD card until my 7870XT. I bought it because it hit a sweet spot performance and budget wise. I considered the 660 and 660Ti, but the 660 was kind of weak and the 660Ti was kind of expensive. Now I have to figure out do I buy a 390 8GB in the next few months or wait till next year. I would get NVidia, but my next card will have more than 4GBs of ram and I'm not spending $600+ to get it.

Also FuryX is out of stock on New Egg every time I check so someone has to be buying them.

 

[Spladam]

The longer you wait to upgrade an AMD card, the more AMD dies. If you love AMD that much you should buy an AMD product.

The problem with AMD's "fans" is none of them are actually willing to buy AMD products otherwise AMD wouldn't be dying. Meanwhile Intel and Nvidia have figured out how to get people to buy their products on a regular basis. AMD has the worst "fans" around if you think about it.

Not my problem though I guess, this whole DX12 flamewar is funny to me because I bought a 980 Ti this year and next year if Pascal is amazing I'll just buy a Pascal, if not I'll wait until the following year. I don't care that Nvidia designs cards to be amazing now as opposed to whatever AMD does because I upgrade them so regularly, the money spent on a hobby I enjoy is worth it to me.

That's just being a bad consumer. Anyone who buys GPU products because of a "fan" relationship with the manufacturer is just a sheep of the Marketing team of said manufacturer, and not very smart. Competition and sales are what make brands work to release better products, one should buy what fits their needs so that manufacturers make things that fit our needs.

 

[GhostTrick]

Not everyone that owns AMD cards is a an AMD fan. A lot of people me included just buy what fits their needs and budget. I guess I can see how someone who has a lot of disposable income and only considers the high end could be dismissive of AMD, but this post is embarrassing. I never bought an AMD card until my 7870XT. I bought it because it hit a sweet spot performance and budget. I considered the 660 and 660Ti, but the 660 was kind of weak and the 660Ti was kind of expensive. Now I have to figure out do I buy a 390 in the next few months or wait till next year. I would get NVidia, but my next card will have more than 4GBs of ram and I'm not spending $600 to get it.

Also FuryX is out of stock on New Egg every time I check so someone has to be buying them.

Exactly this. How is that even possible to be fan of a hardware brand ? The point with PC is that you have the choice of each of your parts... no need for fanboyism with that. It all comes down to your budget and preferences. I spent 240€ on my GPU, that was a R9 290. That was the best performance I could get for that price. If it was an Nvidia GPU, I would've bought it, regardless of brands or whatever.

 

[Lumberjackson]

I kept telling people you could not make direct comparisons of the consoles hardware to the 7XXX and R9's before the R9 285, as the additional ACE's made a difference, something the Tonga added to the R9 series with the 285, and why the 285 gets comparible performance at a lower bit-depth then the 280 and 280X. Everyone kept telling me the ACE's made no difference.

That's more likely due to the bandwidth-saving color compression they added to the 285, rather than the ACEs.

 

[Dictator93]

Exactly this. How is that even possible to be fan of a hardware brand ? The point with PC is that you have the choice of each of your parts... no need for fanboyism with that. It all comes down to your budget and preferences. I spent 240€ on my GPU, that was a R9 290. That was the best performance I could get for that price. If it was an Nvidia GPU, I would've bought it, regardless of brands or whatever.

Agreed, being the fan of a brand is silly. But I definitely am a fan of NV drivers and features.

 

[GhostTrick]

Agreed, being the fan of a brand is silly. But I definitely am a fan of NV drivers and features.

Well, thats part of the product appeal: The drivers stability and up to date but also the range of software which keeps improving.

 

[Beerman462]

Agreed, being the fan of a brand is silly. But I definitely am a fan of NV drivers and features.

I've never had an issue with AMD drivers, but I was pretty salty about the lack of down sampling. Now that they have it I almost never use it, because the performance hit on newer games is too much.

 

[Jedipottsy]

Exactly this. How is that even possible to be fan of a hardware brand ? The point with PC is that you have the choice of each of your parts... no need for fanboyism with that. It all comes down to your budget and preferences. I spent 240€ on my GPU, that was a R9 290. That was the best performance I could get for that price. If it was an Nvidia GPU, I would've bought it, regardless of brands or whatever.

After having tried AMD multiple times, i always regret it. Nvidia just works (added with the fact software i run only runs on nvidia. (CUDA, and OpenGL based apps).

Also checkout AMD linux performance....

I'm a massive nvidia fanboy. I've never regretted any nvidia purchase, its always worked and never given me any issue (barring game specific issues upon launch, that are usually fixed with patches/new drivers)