HD4850 750/1100 slower than 680/1000 in crysis

B

Bakku

Limp Gawd
Joined
Aug 3, 2004
Messages
317
i've OC'ed my 4850 to 750/1100 so far after the pencil mod. the card is running stable at this speed then i fired up the crysis benchmark, to my surprise, the FPS produced by 750/1100 configuration is about 3-5fps slower than 680/1000 configuration (average framerate). even for the minimum fps, 750/1100 is 1fps consistently slower 680/1000.

this is kinda weird and not sure where went wrong.

btw, i used the AMD GPU Clock Tool to do the overclocking, should i have went the bios overclocking way?

i am at the office right now, will do the test one more time when i get home and post the number up here.
 
Screw Crysis, how did the OC effect your other games?
I wouldn't be surprised to see Crysis go down and the rest (including 3dmark06 even) go up.
 
Bakku said:
i've OC'ed my 4850 to 750/1100 so far after the pencil mod. the card is running stable at this speed then i fired up the crysis benchmark, to my surprise, the FPS produced by 750/1100 configuration is about 3-5fps slower than 680/1000 configuration (average framerate). even for the minimum fps, 750/1100 is 1fps consistently slower 680/1000.

this is kinda weird and not sure where went wrong.

btw, i used the AMD GPU Clock Tool to do the overclocking, should i have went the bios overclocking way?

i am at the office right now, will do the test one more time when i get home and post the number up here.
Click to expand...


you are going to be very bottlenecked by that CPU. its prolly the reason you arent getting anything more from it.
 
When I was running a 2.8GHz Athlon 64 x2, I also lost a small % fps in Crysis while clocking my 8800GT.
The Athlon 64's dont have enough power to drive an 8800GT fully let alone a 4850 and clocking
With an E8400, I clock it +20% from stock and see +20% better performance from stock.
You wont be able to get the full performance at stock let alone clocked.

Try this test:
Clock the card to 80% of stock clocks and see if you get almost the same fps.
Try lower clocks to find out when the fps starts to dip quicker.
This will give you the approximate max performance you can get.
ie 70% of stock clocks = 70% of the max stock performance.

Clocking the card wont help you until you alleviate the CPU bottleneck.
 
dnottis said:
you are going to be very bottlenecked by that CPU. its prolly the reason you arent getting anything more from it.
Click to expand...

Why would a CPU bottleneck cause lower FPS when overclocking a video card when compared with stock speeds?


I've heard overclocking the video card RAM automatically causes the timings to loosen (not sure how true that is). If so, maybe they are loosening enough to cause overall losses.
 
thanks for all the feedbacks.

i know my cpu is not fast enough to push the video card, so i wasnt expecting to see any large jump in performance. in 3dmarl06, i've seen around 200 points increament, but in i find it very weird that the framerate actually went down in crysis, it should at least stay at the same level when compared with pre-oc'ed configuration. i just dont get it, lol.
 
Turn up graphics settings in crysis and bench that oc vs non oc. Maybe shifting more load to the gpu with higher graphics settings will result in fps gains.

edit
Or maybe crysis just hates ati cards. Jury's still out on that one ;)
 
id say a micro code that says

"whoops, ati card... kick 10 fps out"
 
Sounds CPU bottlenecked

A few fps variance is within benchmarking margin of error.
 
A few FPS in Crysis is more than just noise because even a 3 FPS difference can be more than 10% change.
 
Well someone needs to take a good look at benchmarking variance. For all the reviews I've seen none have really taken variance into account but Driverheaven IRC.
 
3DMark06 default benchmark run is 1280x1024, no AA or AF. It's CPU limited even with the faster Intel CPUs. If you want to benchmark newer cards in 3DMark 06, you need to get the Pro version and run at 1600x1200 or higher, preferably with some AA and AF applied.

What settings did you benchmark Crysis at? Same applies here, with no AA or AF, you're going to be CPU limited. The lower result when overclocked was probably due to normal variations (+/- 10% is normal).
 
You must log in or register to reply here.
Back
Top