Overclock.net banner
1 - 14 of 14 Posts

·
Premium Member
Joined
·
16,904 Posts
Discussion Starter · #1 ·
BASED ON THIS ARTICLE: http://pclab.pl/art58123-3.html

I was bored (too much free time over the summer) so I figured why not make a small thread about this. If I'm lucky someone might even find it useful even though the data is nothing special.

Let's get started:

So I was reading the devil's canyon reviews when I noticed that the pclab.pl one had some really good data when it comes to game performance on different CPUs using both a 290X and a 780Ti. We don't really ever see this done by review sites because it's time consuming to test with two GPUs instead of one. So usually what we see is that for CPU reviews these sites pick one GPU and just run their whole game test suite with that GPU.

But since pclab.pl decided to test with a 290X as well as a 780Ti I decided that it would be cool to find out if the recent Nvidia CPU overhead improvements (from 337.50 drivers onwards) had actually done anything. And since one of the main things the drivers were supposed to do was better multithreading what better way to find out than to compare the 4690K and the FX 8350? (not beating a dead horse at all now am I?
tongue.gif
).

So I basically just rearranged the data with google spreadsheets making it easier to view and work with.

Here are the graphs I made for the individual games:

















Now from those you can already somewhat see that the 290X is falling behind the 780Ti more when using the FX8350 instead of the 4690K.

But let's make it even easier to see. Let's compare the average difference between the 780Ti and the 290X first on the 4690K platform and then on the FX8350 platform.

4690K platform:



As we can see the the 780Ti is around 6% faster than the 290X on average.
The difference makes sense and sounds just about right.

But what happens once we move to a more CPU bottlenecked platform? Before the 337.50 drivers the assumption would have been that the 780Ti and the 290X would be more evenly matched as the CPU is limiting the GPUs from reaching their maximum potential. However if the 337.50 driver set did its job we should see the difference stay the same or grow.

FX 8350 platform:



The difference between the cards did indeed change.
The 780Ti is now around 15% faster than the 290X

Conclusion:

The newer Nvidia drivers do indeed help with CPU overhead. High core count CPUs that benefit from highly multithreaded software are better off with Nvidia GPUs in situations that are CPU bound (in DX11 applications).
 

·
Premium Member
Joined
·
20,625 Posts
BF4 MP64 DX11 1080 Ultra 4MSAA with a 290 at stock . . .



why am i getting better numbers than a 290X? maybe a different map? maybe driver? 14.6 Beta here.
 

·
Registered
Joined
·
1,095 Posts
Different map would make a big difference. NVidia GPU and AMD CPU pairing was the thing to do for many years, very nostalgic combination
thumb.gif
 

·
Premium Member
Joined
·
16,904 Posts
Discussion Starter · #4 ·
Quote:
Originally Posted by rdr09 View Post

BF4 MP64 DX11 1080 Ultra 4MSAA with a 290 at stock . . .



why am i getting better numbers than a 290X? maybe a different map? maybe driver? 14.6 Beta here.
We have no clue where they did their testing so it's impossible to compare our own BF4 results.

You can get 200fps staring at a wall or 30 by standing in the middle of 20 explosions... Doesn't mean that either result is incorrect.
tongue.gif
 

·
Premium Member
Joined
·
20,625 Posts
Quote:
Originally Posted by Alatar View Post

We have no clue where they did their testing so it's impossible to compare our own BF4 results.

You can get 200fps staring at a wall or 30 by standing in the middle of 20 explosions... Doesn't mean that either result is incorrect.
tongue.gif
true. mine was operation locker but even with HT off i killed that 290X. i must have missed it . . . did they mention what amd driver they used?
 

·
Premium Member
Joined
·
16,904 Posts
Discussion Starter · #6 ·
Quote:
Originally Posted by rdr09 View Post

true. mine was operation locker but even with HT off i killed that 290X. i must have missed it . . . did they mention what amd driver they used?
Can't see that anywhere but I'd just assume the latest ones since it's new testing anyway.
 

·
Registered
Joined
·
419 Posts
i just sold my asus crosshair v formula-z and fx 8320 @ 4.8GHz (h100i) for an asus maximus vii gene and 4770K
my motherboard comes tomorrow so i cant wait to see what it can do with the h100i and if its to hot i will delid it.

anyhow its like this for me.
R9 290/290x vs gtx 780 (but in cfx and sli configs)

i really dont even care about the Maximum FPS what i care about is the minimum.

maximum fps is way over rated.

you can be out in the middle of nowhere on some map with 3000fps but when it counts and there is action going on is where i care.

BF4 specifically multiplayer
 

·
Registered
Joined
·
1,095 Posts
Quote:
Originally Posted by tdbone1 View Post

i just sold my asus crosshair v formula-z and fx 8320 @ 4.8GHz (h100i) for an asus maximus vii gene and 4770K
my motherboard comes tomorrow so i cant wait to see what it can do with the h100i and if its to hot i will delid it.

anyhow its like this for me.
R9 290/290x vs gtx 780 (but in cfx and sli configs)

i really dont even care about the Maximum FPS what i care about is the minimum.

maximum fps is way over rated.

you can be out in the middle of nowhere on some map with 3000fps but when it counts and there is action going on is where i care.

BF4 specifically multiplayer
I've had better experiences with SLI than Crossfire.

If I were buying one card I'd get a 290, otherwise I'd consider SLI 780 maybe.
 

·
Registered
Joined
·
419 Posts
Quote:
Originally Posted by Vario View Post

I've had better experiences with SLI than Crossfire.

If I were buying one card I'd get a 290, otherwise I'd consider SLI 780 maybe.
too bad for me
gtx 780 sli is way more $$ then R9 290 cfx

are you sure that with an i7 4770k that the min frame rates will be better
im really just trying to get rid of those fps spikes at certain points when there is alot of action

with the single gtx 770 i had (its in my specs with the system i did lots of tests on) that there was even spikes on that on 1080P high
i made videos with shadowplay @ 60fps that do show it.
im wondering if its the bandwith of the memory of the cards.

note i also seen the same problem on 16GB ram systems with cfx and sli and also the problem wasnt there on systems with 8GB on amd / intel cpu systems with cfx and sli.

its really weird
also i not talking about the spikes after i die

im not sure why people dont use that graph more but it sure would help out alot to use it
that is why dice gave it to us
 

·
Premium Member
Joined
·
5,508 Posts
Minimum framerate between8320/8350 and i7 SB-E/IVB/IVB-E/HW/HW-E, the latter is SIGNIFICANTLY better in every game I played with both in the (fairly short) time I had an 8350/CHV-FZ.
That's with a 3.5Ghz Intel and the AMD at 5Ghz, and once the former are OC'ed, the gap only widens.

It's definitely the best way to go, ESPECIALLY for 2+ cards!

Oh, and as you are like me in favoring a high minimum over high avg/max, there's another thing that helps...
Memory frequency and timings!
I would not go with less than DDR3-2400 9-11-11-30-1T for IVB/HW, compared to 1600 7-8-7 it made a fairly significant difference in a number ofgames (10pct or mmore), and made some difference in all games (at least 3pct). Dramatically reduced stuttering, too.
This is most noticeable in 2+ GPU systems.
 

·
Premium Member
Joined
·
8,764 Posts
Could this improved CPU overhead driver improvement be what accounts for the massive increase in performance when using a small pre-render when the game is CPU bound? From my tests:

GTX 460
pre-rendered frames set to 2 = 194 Minimum. 300 Average.
pre-rendered frames set to 1 = 188 Minimum. 304 Average.

HD7950
pre-rendered frames set to 2 = 194 Minimum. 300 Average.
pre-rendered frames set to 1 = 155 Minimum. 230 Average.
 

·
Registered
Joined
·
419 Posts
Quote:
Originally Posted by nleksan View Post

Minimum framerate between8320/8350 and i7 SB-E/IVB/IVB-E/HW/HW-E, the latter is SIGNIFICANTLY better in every game I played with both in the (fairly short) time I had an 8350/CHV-FZ.
That's with a 3.5Ghz Intel and the AMD at 5Ghz, and once the former are OC'ed, the gap only widens.

It's definitely the best way to go, ESPECIALLY for 2+ cards!

Oh, and as you are like me in favoring a high minimum over high avg/max, there's another thing that helps...
Memory frequency and timings!
I would not go with less than DDR3-2400 9-11-11-30-1T for IVB/HW, compared to 1600 7-8-7 it made a fairly significant difference in a number ofgames (10pct or mmore), and made some difference in all games (at least 3pct). Dramatically reduced stuttering, too.
This is most noticeable in 2+ GPU systems.
this info is exactly what i was wondering.
i just got rid of my complete system in my sig (against everyone saying there is no "bottlenecking" in the fx 8320 @ 4.8GHz with cfx or sli systems with gtx 770 and up or r9 280 and up)

maybe its not bottlenecking but i know there has to be something that is causing those spikes on i think amd system.

i did find a video of a guy with an fx 8350 OC (but i forget what it was) and he had 2 R9 290's or 290x's
his video lasted 30mins and was smooth as silk so i really just dont get it.

i have seen alot more intel sli systems that were smooth as silk but only like one or two amd systems.

im leaning on pairing the intel 4770k with gtx 780 sli but i sure wish it was R9 290 instead because they cheaper with almost the same performance (when you look at frame rates alone)
as we all know that dont tell the whole story.
when there is action or an intense scene weird things show up and thats the stuff i trying to stop.
good info on the higher speed ram.
 

·
Premium Member
Joined
·
20,625 Posts
Quote:
Originally Posted by tdbone1 View Post

this info is exactly what i was wondering.
i just got rid of my complete system in my sig (against everyone saying there is no "bottlenecking" in the fx 8320 @ 4.8GHz with cfx or sli systems with gtx 770 and up or r9 280 and up)

maybe its not bottlenecking but i know there has to be something that is causing those spikes on i think amd system.

i did find a video of a guy with an fx 8350 OC (but i forget what it was) and he had 2 R9 290's or 290x's
his video lasted 30mins and was smooth as silk so i really just dont get it.

i have seen alot more intel sli systems that were smooth as silk but only like one or two amd systems.

im leaning on pairing the intel 4770k with gtx 780 sli but i sure wish it was R9 290 instead because they cheaper with almost the same performance (when you look at frame rates alone)
as we all know that dont tell the whole story.
when there is action or an intense scene weird things show up and thats the stuff i trying to stop.
good info on the higher speed ram.
what resolution are you gonna be using in BF4? BF4, right?
 
1 - 14 of 14 Posts
Top