Topic Title: Odd Dirt3 FPS scaling AMD FX-8
Topic Summary:
Created On: 02/24/2012 06:54 AM
Status: Read Only
Linear : Threading : Single : Branch
Search Topic Search Topic
Topic Tools Topic Tools
View similar topics View similar topics
View topic in raw text format. Print this topic.
 02/24/2012 06:54 AM
User is offline View Users Profile Print this message

Author Icon
abundant threading
n00b

Posts: 4
Joined: 02/16/2012

Hi guys, new here.

I don't understand this http://www.techpowerup.com/reviews/AMD/HD_7970_CPU_Scaling/9.html

@ 1024 x 786 the FX-8 is getting around 120FPS with the 2500K around 200FPS.

As you can see the FX remains about the same @ 120FPS or so.. throughout the test and in the last test the FX and 2500K are level pegging.

Now you could put that down to the FX bottle-necking the GPU and the 2500K being much better at handling that GPU.

Now, Techpowerup did not publish a test higher then where the FX matched the 2500K, but i have found one...

Here http://amdfx.blogspot.com/2012/02/dirt-3-revisited-again-by-request.html

They boosted the AA to 8x from 4x in the techpowerup link and the results are to me a little surprising, clock for clock suddenly the FX overtakes the 2500K and is still there about consistent with what it has been in the first link.

What is going on there?



-------------------------
Nothing yet

Edited: 02/24/2012 at 07:12 AM by abundant threading
 02/24/2012 08:52 AM
User is offline View Users Profile Print this message

Author Icon
Hardwood
Dungeon Master

Posts: 4001
Joined: 01/27/2008

Same ol same ol...
Most cpu reviews use low res gaming for their benchmarks. The reality being when it comes to hi res gaming it's the GPU that counts and CPUs are all the same or have insignificant differences. That's a impressive OC for the FX too.

-------------------------

#1: Phenom II C2 955BE @ 3.6, Antec Kuhler 920, Gigabyte GA-MA790FX-DS5, Asus 5870/V2, ASUS Xonar D2X, 4 x 1gig OCZ2N1066SR2GK @800, Antec Truepower New 750, Arc Midi, Vistax64
#2: MSI A88X-G45, AMD A10-7850K, IGP, AMD 2x4GB 2133 (R938G2130UK), Antec 750 True, 128GB Sandisk SSD, WD Black 1TB, Win8.1pro x64
#3: Athlon 64 X2 4200+, AGP 1950pro, Audigy 2, Asus A8V-D, 2gig ram, Enermax 465P, Antec900case. Now on Win8.1pro cat 10.2

 02/24/2012 09:37 AM
User is offline View Users Profile Print this message

Author Icon
abundant threading
n00b

Posts: 4
Joined: 02/16/2012

Originally posted by: Hardwood Same ol same ol... Most cpu reviews use low res gaming for their benchmarks. The reality being when it comes to hi res gaming it's the GPU that counts and CPUs are all the same or have insignificant differences. That's a impressive OC for the FX too.

 

Thanks

So the FX beating the 2500K (by a large margin) is neither here nor there... and the FX bottle-necking the GPU before the 2500K stands?



-------------------------
Nothing yet
 02/24/2012 04:50 PM
User is offline View Users Profile Print this message

Author Icon
Earnhardt
Case Modder

Posts: 816
Joined: 09/11/2008

who the heck plays at 1280x768 anyways..lol

 



-------------------------

AMD Phenom II 1100t @4.0Ghz /Asus sabertooth 990fx/2x Gigabyte 3G 7950's in crossfire/500G samsung 840 ssd/16G G-Skill Ripjaw 1600/2-samsung 24" monitors 1920X1200/Corsair AX1200W psu/W7 pro 64,G19 keyboard,Roccat kone pure mouse,Cooler master HAF932,Corsair H100i water cooler,Zalman ram cooler.Sound Blaster X-Fi Titanium Fatal1ty Champion Series,Logitech Z5500 speakers,Roccat Kave 5.1 headset.

 02/24/2012 05:01 PM
User is offline View Users Profile Print this message

Author Icon
abundant threading
n00b

Posts: 4
Joined: 02/16/2012

Right... gaming on a retro rig lol



-------------------------
Nothing yet
 02/25/2012 04:21 PM
User is offline View Users Profile Print this message

Author Icon
Thanny
Elite

Posts: 1296
Joined: 07/13/2009

Don't make the mistake of thinking that a CPU bottleneck means that the CPU is slowing the GPU down. 

All that means is that the rate at which the CPU can calculate snapshots of the virtual world is lower than the rate at which the GPU can render that snapshot into an image.

The CPU's workload is not affected by screen resolution, so what you see at a low resolution like 1024x768 (I note in passing that I'm old enough to still find that statement odd - I paid, many moons ago, over $500 for a graphics card capable merely of displaying 1024x768 pixels in 24-bit color, with its huge 4MB frame buffer) is basically the maximum speed the CPU is capable of running that game at.

As resolution goes up, the GPU needs to do more and more to keep up, and if said resolution goes up high enough (or you enable sufficiently demanding modes of anti-aliasing and anisotropic filtering), the GPU will no longer be able to keep up with the CPU.

Things get complicated when you start offloading to the GPU work that's typically done by the CPU.  There could easily be situations where the GPU has capacity to spare at low resolutions, making the game run faster with offloading, while at higher resolutions the speed would be better if the offloaded work was done by the CPU instead, leaving the GPU more time to render the additional pixels.

The results referred to are a bit poor for showing all this, since both aspect ratio and AA settings change at the lowest resolutions.

 

 02/25/2012 05:16 PM
User is offline View Users Profile Print this message

Author Icon
abundant threading
n00b

Posts: 4
Joined: 02/16/2012

Thanks, not that i fully understand all of that but if i have got the gist of it i would say that both tests are true in that the FX can only handle D3 @ 120 FPS max, where as the 2500K can do 200.

Yet that is not to say the 2500K is better, in fact once the work load on both the GPU and CPU is high the FX can maintain its 120FPS up to the point where the 2500K falls behind.

so in conclusion the 2500K is simply better at low GPU workloads and is over taken by the FX in high workloads, in short, low settings 2500K = better, high setting FX = better.

that about right?



-------------------------
Nothing yet
 02/27/2012 05:27 AM
User is offline View Users Profile Print this message

Author Icon
Thanny
Elite

Posts: 1296
Joined: 07/13/2009

I don't see where the 2500K falls behind at all.  The tiny difference at 2560x1600 is just noise in a situation where the GPU is clearly the limiting factor.

The chart shows that Bulldozer is limited to around 125fps, while Sandy Bridge can do at least 200fps.  The CPU is the bottleneck until 2560x1600 with 4xAA for Bulldozer, while the GPU is the bottleneck starting at 1280x1024 with 2xAA for Sandy Bridge (or is already the bottleneck at 1024x768, which we can't confirm without numbers from a lower resolution).

It's not quite so neat, of course, as you can see more readily from the rather volatile Nehalem numbers.

Overall, though, there's no question that Sandy Bridge is faster than Bulldozer.  AMD simply hasn't caught up since Intel's mobile team developed Core 2, and arguably would have never been ahead if Intel hadn't taken a side trip with the awful Netburst architecture.

 

Statistics
84124 users are registered to the AMD Support and Game forum.
There are currently 4 users logged in.

FuseTalk Hosting Executive Plan v3.2 - © 1999-2014 FuseTalk Inc. All rights reserved.