Wildly varying FPS @ any given clock frequency.

Status
Not open for further replies.

RichM499

Daemon Poster
Messages
783
Location
Spaceballs Mega-Maid
I can't for the life of me figure out what the **** is going on here.

I decided to play around with a file server machine I have in the other room.
It was an extremely cheap build, and the video card was a left-over. I wanted to see how far I could overclock the processor, and then test performance.

Well I guess I should have done it one step at a time and tested performance at every step, but, well.
I am using Counter Strike Source stress test with everything maxed out, 8xaa 16xaf @ 1600x1200.

The specs of the machine are:
Radeon 4870
Athlon II x2 240
ASRock N68-S UCC
2gb DDR2 667 PNY 5-5-5-13 1.8v @ DDR2 667 4-4-4-12 2.0v

Everything is fine with this machine as far as stability. The machine just holds some of my files for my personal network.

OK This is the weird part. When I tested performance with the CSS stress test and the above settings at 3.64ghz I got 70~ FPS. That seemed to be quite low considering this game is old as dirt, so I decided to lower the overclock to 3570 (by decreasing the HT from 260 to 255, the multi is 14 in case you were wondering) and I got 101fps.

Does this make sense to you? It sure as **** doesn't make a lot of sense to me

So I lowered it to stock 2.8ghz. 90.05fps
3.43ghz - 100fps
3.50ghz - 80fps???
3.616ghz - 80fps
3.43ghz - 80fps (this is the 2nd time because at this point I'm just confused)


What's going on? How is this possible? Why are my framrates varying so wildly? Any ideas?

Oh and just FYI these aren't exact results. If you'll give me some sort of true CPU benchmark I can quickly run at any given speed I'll report exact numbers to you to aid in helping me figure out wtf is up with this system.

Thanks in advance.

Edit: Update:
SiSoftware Sandra Processor Arithmetic test
Aggregate Arithmetic Performance: 19.36GOPS
Dhrystone ALU 20.44GIPS
Whetstone iSSE3 18.29GFLOPS
@ 2.8ghz

Aggregate Arithmetic Performance: 19GOPS
Dhrystone ALU 19.85GIPS
Whetstone iSSE3 18.3GFLOPS
@ 3.35ghz

Aggregate Arithmetic Performance: 19GOPS
Dhrystone ALU 19.7GIPS
Whetstone iSSE3 18.3GFLOPS
@ 3.5ghz

What the ****?

Edit: Update #2

Super Pi mod 1.5xs 1m

30.831s @ 2.8ghz
25.426s @ 3.35ghz
24.508s @ 3.5ghz
23.527s @ 3.616ghz

THIS makes more sense. I'm sure those numbers would drop dramatically if I'd close all my browsers and turn off Aero, but I just wanted to see a clear and defined performance curve. Finally I get a performance curve. This still doesn't help me understand the SiSoftware and CounterStrike results though....

EDIT: Update #3

3dmark Vantage CPU test 1 AI test.

2.8ghz : 472.27 PLANS/S 1st | 479.52 2nd | 477.72 3rd.
3.07ghz : 532.45 PLANS/s 1st | 523.27 2nd | 524.14 3rd.
3.5ghz : 606.75 PLANS/s 1st | 610.44 2nd | 608.03 3rd.

So this makes sense to me. It looks like a clear indicator that the higher frequency is giving better results.

Do you guys have any idea why CS:S and SiSoft are giving me those nutty results?

I think that CS:S may be GPU limited at those settings, but then again that game is so old. Maybe the 8xaa and 16xaf combined with all of the other settings turned on as high as they will go are taxing this 4870.


Any opinions or ideas?
 
This is premature and not really a bump, but uh.. 72 views and no one has any ideas at all?

No one?

I mean it's fine if this is some freak occurance or something but I've been here since 2005 and never once have I come across a problem that you techies couldn't solve for me.

Did I stump the great Tech Forums?
 
1-3.jpg


I guess it's not as stable as I thought it was. Prime95 never had any issues. I'll drop the FSB by 2 or 3 and try again until it's stable.




2-3.jpg


Well it seems stable at 3.58ghz with 1.51v.

I had this chip clocked 80mhz slower before when I tested CSS with the same vcore when CSS and Sandra gave me those wierd numbers.
So I'm still confused......
 
Status
Not open for further replies.
Back
Top Bottom