The GTX 970 vs. R9 390 thread to (theoretically) end them all

[ivan134]

2 minutes ago, Majestic said:

*on a 4,5ghz 6700K

*with a reference 970

 

On an i5 with both overclocked cards, the 390 will not leave the 970 in the dust.

In what universe does a CPU play a role in a GPU bound scenario?

[PlayStation 2]

Just now, ivan134 said:

In what universe does a CPU play a role in a GPU bound scenario?

Far Cry 4 isn't exactly only GPU bound.

[ivan134]

Just now, Dan Castellaneta said:

Far Cry 4 isn't exactly only GPU bound.

So does this mean that pclab.pl is FOS like we already knew? That site has been discredited numerous times and yet people are praising one of Nvidia's resident shills for posting their benchmarks here.

[Majestic]

7 minutes ago, ivan134 said:

In what universe does a CPU play a role in a GPU bound scenario?

The fact that Far Cry 4 results have the same delta's as shadow of mordor. Far Cry 4 being extremely IPC heavy, SOM CPU unintensive.

Means the CPU is really fast and doesn't suffer from the driver overhead.

 

 

Look at the Far Cry 4 results and shadow of mordor. 

[Majestic]

9 minutes ago, ivan134 said:

So does this mean that pclab.pl is FOS like we already knew? That site has been discredited numerous times and yet people are praising one of Nvidia's resident shills for posting their benchmarks here.

No, not unless you also discredit Digitalfoundry. In which case, you are the shill.

 

And to be clear, I have yet heard any of you people make a solid argument for why their results make no sense. Only that you don't like the results.

 

EDIT; And a cheap adhominem to boot, really desperate man.

[ivan134]

11 minutes ago, Majestic said:

The fact that Far Cry 4 results have the same delta's as Mad Max. Far Cry 4 being extremely IPC heavy, Mad Max CPU unintensive.

Means the CPU is really fast and doesn't suffer from the driver overhead.

 

 

Look at the Far Cry 4 results and shadow of mordor. 

IPC plays no role in the video you just linked for FC4. All Skylake cores have the same IPC. The performance is going up with increased cores and threads. 

 

You also literally just made shit up about deltas between FC4 and Mad Max.

 

The Shadow of Mordor portion also shows the game doesn't care what CPU you have as the gains from going from the i3 to the 6600k or 6700k are very marginal compared to the CPU intensive games. So yes, in a GPU bound scenario, 390 does leave the 970 in the dust and trades blows with a 980.

[Majestic]

9 minutes ago, ivan134 said:

IPC plays no role in the video you just linked for FC4. All Skylake cores have the same IPC. The performance is going up with increased cores and threads. 

 

You also literally just made shit up about deltas between FC4 and Mad Max.

 

 

Between a game that shows 73% improvement between i3 and i7 versus a game that showed none. And you're telling me that that wouldn't affect scores on a slower CPU? When the i3-6100 on that OC is faster than a 2500K for example? When also armed with the DigitalFoundry video that was just released, showing another example of the driver overhead in tomb raider.

 

Yeah, no. I'm fairly certain you're wrong here.

 

Also, still no argument why the PCLAB.pl results are inherently fraudulent or make no sense.

[App4that]

1 minute ago, Majestic said:

 

[ivan134]

7 minutes ago, Majestic said:

-snip

No, you don't know what IPC is. All cores of the same architecture have the same IPC. The simplest explanation is that it's the measure of the strenghth of a single core. A Skylake i3 core has the same IPC as a Skylake i7 core. Same with Haswell i3 and Haswell i7, etc. This video you just linked actually shows that IPC plays a part since there is improved performance going from an Haswell i7 to a Skylake i7. So while you are right, the 1st video did not support what you were saying. That 73% performance increase was from adding more cores and threads.

Also, Shadow of Mordor having the same delta means absolutely nothing as you can see in both of your videos. There's no performance increase going for a Haswell i7 to a Skylake i7 like there was in FC4. So yes, it's completely made up nonsense. Your own evidence proves that. You will also need to provide evidence for that for Mad Max. I also like how you conveniently ignored Ryse.

Give me a while, I'll try and find the other thread where you were assblasted for using pclabs.pl.

 

[Majestic]

8 minutes ago, ivan134 said:

No, you don't know what IPC is. All cores of the same architecture have the same IPC. The simplest explanation is that it's the measure of the strenghth of a single core. A Skylake i3 core has the same IPC as a Skylake i7 core. Same with Haswell i3 and Haswell i7, etc. This video you just linked actually shows that IPC plays a part since there is improved performance going from an Haswell i7 to a Skylake i7. So while you are right, the 1st video did not support what you were saying.

 

That 73% performance increase was from adding more cores and threads.

 

Also, Shadow of Mordor having the same delta means absolutely nothing as you can see in both of your videos. There's no performance increase going for a Haswell i7 to a Skylake i7 like there was in FC4. So yes, it's completely made up nonsense. Your own evidence proves that. You will also need to provide evidence for that for Mad Max. I also like how you conveniently ignored Ryse.

 

Give me a while, I'll try and find the other thread where you were assblasted for using pclabs.pl.

 

It shows CPU scaling, so that means a GPU with a large driver overhead should be more adversely affected. The results show almost identical delta's, meaning the i7-6700K @ 4,5ghz is not an issue for these GPU's. Which is also supported by the PCLAB.pl results. the 390 wins with a CPU that fast. However as the DigitalFoundry video and PCLAB.pl i5 results BOTH showed, on a locked i5-6500 or i5-2500K OC it does. So the fact the 390 beats the 970 in that scenario is because of two caveats. One, the 970 is probably running reference clocks (so not even factory OC), and two the CPU is really fast meaning no driver overhead issues. Put the cards against each other in a more viable scenario (both factory OC and an i5) and the difference isn't anywhere near this level. Meaning you're cherrypicking.

 

It was supposed to highlight CPU scaling vs. no CPU scaling and it not being reflected in the results from TechPowerUp. You're debating semantics to avoid debating the real argument.

 

I know, that's the problem with YOUR results. 

 

What would more subjective/anecdotical and fanatic character assassinations prove. I'm not asking for more of the same, I'm asking for a fucking rational, technical argument. 

[don_svetlio]

You could add Bridgless CF - it's proven to work better than SLI bridges.

[ivan134]

Instead of finding the thread, I'll do the next best thing. According to pclab, overclocked i7s gets rid of the CPU overhead for AMD cards. These charts are from a review of the 380x using a 6700k OC'ed to 4.7 GHz. According to pclab, not only is a 960 faster than a 380 even with overpowered CPU, it's faster than a 380x. What a trustworthy site.

[ivan134]

This site has been discredited thousands of times because of their outlier benchmarks, and yet people are still using it here to spread fud.

[App4that]

2 minutes ago, don_svetlio said:

You could add Bridgless CF - it's proven to work better than SLI bridges.

Yes, it works awesome. The best part was when it offered no performance increase in games.

[Majestic]

1 minute ago, ivan134 said:

Instead of finding the thread, I'll do the next best thing. According to pclab, overclocked i7s gets rid of the CPU overhead for AMD cards. These charts are from a review of the 380x using a 6700k OC'ed to 4.7 GHz. According to pclab, not only is a 960 faster than a 380 even with overpowered CPU, it's faster than a 380x. What a trustworthy site.

 

Relatively, not in absolute terms.

 

 

What a trustworthy guy you are. Never cherrypicking.

[don_svetlio]

Just now, App4that said:

Yes, it works awesome. The best part was when it offered no performance increase in games.

Your personal incompetence does not make a superior technology inferior.

[ivan134]

8 minutes ago, Majestic said:

Relatively, not in absolute terms.

 

 

What a trustworthy guy you are. Never cherrypicking.

Yea, you're right. Every other outlets' benchmark is wrong, but they're right. PClab is the only valid reviewer to you, but apparently I'm the one who is cherry picking.

 

[App4that]

2 minutes ago, don_svetlio said:

Your personal incompetence does not make a superior technology inferior.

How exactly does one mess up plugging in two cards, choosing the crossfire profile in Crimson, and starting a game? Was there a secret handshake I missed?

 

 

[don_svetlio]

Just now, App4that said:

How example does one mess up plugging in two cards, choosing the crossfire profile in Crimson, and starting a game? Was there a secret handshake I missed?

 

 

Yes, using 2 cards with the same PCB rather than 1 with a custom PCB and 1 with a reference PCB - half the users on LTT told you that was the issue, you proceeded to ignore them. case closed

[App4that]

Just now, don_svetlio said:

Yes, using 2 cards with the same PCB rather than 1 with a custom PCB and 1 with a reference PCB - half the users on LTT told you that was the issue, you proceeded to ignore them. case closed

So your argument is that you have to have two of the same PCBs along with having the same GPU to crossfire, but somehow that makes it better than SLI with no such requirement? All you need with SLI is the same GPU and VRAM quantity.

[don_svetlio]

Just now, App4that said:

So your argument is that you have to have two of the same PCBs along with having the same GPU to crossfire, but somehow that makes it better than SLI with no such requirement? All you need with SLI is the same GPU and VRAM quantity.

then why do 680s and 770s reject each other? You need the same PCB for obvious reasons

[Prysin]

31 minutes ago, Majestic said:

 

 

Between a game that shows 73% improvement between i3 and i7 versus a game that showed none. And you're telling me that that wouldn't affect scores on a slower CPU? When the i3-6100 on that OC is faster than a 2500K for example? When also armed with the DigitalFoundry video that was just released, showing another example of the driver overhead in tomb raider.

 

Yeah, no. I'm fairly certain you're wrong here.

 

Also, still no argument why the PCLAB.pl results are inherently fraudulent or make no sense.

are you trying to argue that TWO GAMES MADE ON SEPARATE ENGINES, does not behave or load the game differently, and thus is apples to apples?

Because, in that case, Maxwell = Hawaii and Hawaii = Maxwell.

 

Incase you didnt know.

FC4 is made with Dunia 2 Engine, which is based on Dunia, which is a extremely heavily worked over version of Cry Engine 1

SOM is made with LithTech "Juptier EX" Engine, which was made by Monolith Productions and Microsoft..

 

Neither engine has ANY resemblance other then supporting DX11

 

You cannot compare games with different levels of load without knowing what load they exert on the hardware.

A simple numbers comparison between two games, are about as revealing as clock speed and core count is between processors. And we all know that just cuz an Skylake i5 has lower clocks and fewer cores, it is by far NOT worse then a FX 8350.

 

So, your whole argument, which is based on the most retarded comparison i have seen to date, falls apart.

 

Both of those games wins on AMD, because they utilizes more of what AMD excels at. Meaning, shaders. Not Asynchronous shading, but shaders in general. If you look at AMDs shader array vs Nvidias at a deep level, AMD is going to run circles around any current Nvidia GPU in terms of shaders. Likewise, come Geometry (IE: tesselation ), Maxwell and Kepler runs circles around any current AMD GPU.

It simply boils down to the strength of each GPU.

 

Memory bandwidth is also greater on AMD cards by default, because their memory bus is much wider, this helps their GPUs in memory resolution heavy tests, because whilst memory capacity (VRAM) may not be an issue (except SOM with HD packs), memory bandwidth IS. Also, whilst i have seen no proof for or against, may have something to do with how well the GPU in AMD cards handles increase of resolution vs how Nvidia GPUs does it. Dunno how the internal workings of AMD and Nvidia GPUs handles resolution change, but there is probably something to look into right there. Just need to be a frikking professor of advanced micro circuitry to understand what you are looking at.

 

 

[Majestic]

Just now, ivan134 said:

Yea, you're right. Every other outlets' benchmark is wrong, but they're right.

You're just going to continue using strawmen fallacies when you're caught slipping huh.

 

I'm pretty sure, in your head, you feel like you've yet again proven PCLAB.pl is fraudulent.

[App4that]

Just now, don_svetlio said:

then why do 680s and 770s reject each other? You need the same PCB for obvious reasons

So, now you're arguing that 3D Mark and their Firestrike benchmark is invalid?

[Majestic]

8 minutes ago, Prysin said:

are you trying to argue that TWO GAMES MADE ON SEPARATE ENGINES, does not behave or load the game differently, and thus is apples to apples?

Because, in that case, Maxwell = Hawaii and Hawaii = Maxwell.

 

You cannot compare games with different levels of load without knowing what load they exert on the hardware.

A simple numbers comparison between two games, are about as revealing as clock speed and core count is between processors. And we all know that just cuz an Skylake i5 has lower clocks and fewer cores, it is by far NOT worse then a FX 8350.

 

So, your whole argument, which is based on the most retarded comparison i have seen to date, falls apart.

 

Both of those games wins on AMD, because they utilizes more of what AMD excels at. Meaning, shaders. Not Asynchronous shading, but shaders in general. If you look at AMDs shader array vs Nvidias at a deep level, AMD is going to run circles around any current Nvidia GPU in terms of shaders. Likewise, come Geometry (IE: tesselation ), Maxwell and Kepler runs circles around any current AMD GPU.

It simply boils down to the strength of each GPU.

 

Memory bandwidth is also greater on AMD cards by default, because their memory bus is much wider, this helps their GPUs in memory resolution heavy tests, because whilst memory capacity (VRAM) may not be an issue (except SOM with HD packs), memory bandwidth IS. Also, whilst i have seen no proof for or against, may have something to do with how well the GPU in AMD cards handles increase of resolution vs how Nvidia GPUs does it. Dunno how the internal workings of AMD and Nvidia GPUs handles resolution change, but there is probably something to look into right there. Just need to be a frikking professor of advanced micro circuitry to understand what you are looking at.

 

 

No i'm saying that an engine which puts more stress on the CPU (significantly), has to effect the graphics hardware more which suffers from a driver overhead. Just as DigitalFoundry and PCLAB.pl found games where driver overhead is an issue, and ones in which it is not. The point was, because there was no significant difference between the delta between the 970 and the 390 in such vastly different engines, means the CPU wasn't a factor in both scenario's. Which means that the 6700K @ 4,5ghz obscures some of the more nuanced results you'd have running a lower-end CPU. Which was the whole point Both DF and PCLAB did the tests.

 

Are you saying the FX-8350 isnt worse than a skylake CPU in Far Cry 4? In that case, you're incredibly disingenuous and we're done here. 

Again, semantics. The point was that FC4 scaled heavily on CPU, SOM didn't. And yet the delta's are the same.

 

Yes, i know the 390 is faster hardwarewise. On a fast. Fucking. CPU.

 

On  a fast. c. p. u.