R9 290/i5-2500k: Peculiar benchmark behavior

Toggle sidebar Toggle sidebar
R

Randomite88

Distinguished
Feb 22, 2011
7
0
18,510
First off, here are my relevant system specs:

CPU: i5-2500k (OC to 4.3 GHz)
Mobo: Asrock Z75 Pro3
GPU: Gigabyte R9 290 Windforce (OC to 1030/1400 up from 947/1250) Cat. 14.7 beta
RAM: 8Gb DDR3-1600
PSU: Antec HCG-850W

I recently upgraded to the R9 290 and have been running some benchmarks to make sure the card is stable and operating correctly. I have been comparing these benchmarks to various recent reviews I have found online (I have overclocked the 290 to the stock Sapphire Vapor-X 290 frequencies to better match results). For the most part, the results I have gotten have lined up with reviews, e.g. Crysis 2, Tomb Raider, Metro Last Light, Bioshock Infinite, Unigine Heaven 4.0, 3dMark Firestrike, among a few others.

However, when I ran the built-in benchmark for Hitman Absolution, the results I got were drastically different from the reviews. With all settings on Ultra (except MSAA toned down to 2x from 8x), I got 58-59 FPS, while a reviewer (LANOC) got 77 FPS with the exact same settings. LANOC's result lines up with the kind of performance other reviewers have seen.

When I tinkered around with settings, e.g. dropping the preset to high, maxing out or turning off MSAA, I would only get an avg FPS deviation of ±3 FPS.

I really don't understand this behavior. Could a GPU opreate normally on most benchmarks but get significantly lower performance than usual on others?

I would really appreciate your input on this.
 
Sort by date Sort by votes
T

Traciatim

Distinguished
Nov 11, 2006
2,057
0
20,460
58-59 and not moving when you adjust settings really seems like it's just syncing to the refresh of the monitor though. Are you sure there isn't something in the catalyst setup that's overriding the settings in the game that's forcing it to turn V-Sync on? If you record the frame times and your min/max/average when you change the settings do the min/max/average all stay about the same, all shift at once, or does just the max/average drop when you turn the AA and stuff back on?
 
Upvote 0 Downvote
dwatterworth

dwatterworth

Honorable
Dec 5, 2012
1,535
0
12,460


Adding to Traciatim, that Raptr program that seems to be standard with the AMD install also can override settings, adding more possible issues.
 
Upvote 0 Downvote
R

Randomite88

Distinguished
Feb 22, 2011
7
0
18,510


Thanks for the reply. The FPS does not remain constant through the test. jumps up to around 80, and goes down to 48-50 in parts. CCC isn't forcing VSync (all settings are set to application dependent), and I do not have the Raptr application installed.
 
Upvote 0 Downvote
R

Randomite88

Distinguished
Feb 22, 2011
7
0
18,510

Raptr application is not installed (unchecked it at driver install) and the CPU used for the benchmarks was an i7-3960x, however according to this link the scaling between processors shouldn't be so drastic.
 
Upvote 0 Downvote
dwatterworth

dwatterworth

Honorable
Dec 5, 2012
1,535
0
12,460
Have you recorded the temps during your benchmarking runs? Any chance the GPU is throttling? Setting the OC to the Vapor X configuration is very ambitious, those Vapor-X / Toxic cards get binned pretty heavily to reliably hit their clock rates and stay in spec temps.

Some benchmarks/programs hammer hardware differently than others. Just take a look at the CPU benches that all load the CPU up to 100% but you get drastically different temperatures between the lot.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom