XFX 8600GT SLI beats 8800GTS 320MB OC and Costs Less

Status
Not open for further replies.
OP
H

Harvik780

ToTheBeatOfUrHeart
^^I scanned for artifacts using Artifact Tester 5 and Ati tool 0.26(Latest).There were no artifacts at all.I have overclocked the cards to 700Mhz Core/ROP domain and 1566Mhz Core/Shader Domain.The memory is overclocked to 860Mhz.
I am really interested in getting the Artic Cooling's MX-2 and see how much more i can go with that.

I also tested a lot of latest games.
Crysis
UT3
Gears Of War
Bioshock
COD4
NFSPS demo.
 

darklord

Cyborg Agent
Hmm, but could you post some scores ?
Why i am asking is , i want to see how efficiently SLI scales with clocks.That would be interesting to see.
One suggestion,
While benching with overclocked cards, it will be a pain to keep both the cards cool, till the time you bench, you can keep 2 x 120mm high CFM fans over the 2 cards which will keep things cool.
This is possible only if you are benching on a benching setup and not if you have the PC assembled in a Case.
What do you use to OC the cards ?
Why not test what happens with CPU clocking too ? as in how much does SLI benifit from overclocking the CPU alone.
 
OP
H

Harvik780

ToTheBeatOfUrHeart
^^There you go
*img118.imageshack.us/img118/8917/8600gt002so4.jpg

The max OC with CPU i got was 2.8Ghz(primarily because of 667Mhz RAM)max OC on GPU without Artic Cooling MX-2 was 702Mhz/864Mhz/1566Mhz.
There's probably no way i can get an HSF cooler to fit in my SLI setup nor a passive cooler is available easily(even tried from matrix 3d LTD).Water coolers are probably hard to source here in India....found that after looking for Corsair nautilus 500(i guess).

Here are the test results including Scaling in SLI at different resolutions and Clocks.

ROP-540/MEM-700/Shader-1188,Resolution -1024*768,CPU-2.8Ghz
Score( Single ) - 6114 Score( SLI ) - 9896
Scaling - 1.619x

ROP-702/MEM-864/Shader-1566,Resolution - 1024*768,CPU-2.8Ghz
Score( Single ) - 7432 Score( SLI ) - 11480
Scaling - 1.545x

ROP-540/MEM-700/Shader-1188,Resolution - 1280*1024,CPU-2.8Ghz
Score( Single ) - 4861 Score( SLI ) - 8177
Scaling - 1.682x

ROP-702/MEM-864/Shader-1566,Resolution - 1280*1024,CPU-2.8Ghz
Score( Single ) - 6018 Score( SLI ) - 9825
Scaling - 1.633

ROP-702/MEM-864/Shader-1566,Resolution - 1280*1024,CPU-2.33Ghz Stock
Score ( SLI ) - 9225

The software i used for overclocking my GPU(s) was ntune because most other softwares currently only support experimental SLI overclocking support.I have downloaded ATI tool 0.27 Beta 4 and will overclock using it and show some more results.
Since my cards are not that big i already use two 80mm fans to blow directly at the GPU heatsink to keep things cool just in case.
As u can see we have some interesting results here.At higher resolutions with higher overclocking we get higher scaling in SLI,so SLI is pretty much for high res gaming.If u have SLI setup u would wanna game at resolutions over 1280*1024.
Also we can see that overclocking the CPU gave me a score jump of precisely 600.

With this setup i can play crysis at High with no tweaks at 1024*768 with 2x AA.
 
Last edited:

darklord

Cyborg Agent
Wow !! very detailed testing there harvik, good job !!;)

I am saying, keep a very high CFM 120mm fan next to the SLI bridge over the 2 cards, parellel to the board, will give you slight headroom for better clocks IMHO.
Also IF possible check if the thermal paste on the core is making good contact with the heatsink, most of the times i have found, it doesnt do it.AS5 will help you cool the card a bit.
I guess you are using the P5N-E SLI mobo for testing, its based on the 650i SLI chipset so you can run the RAM in unlinked mode which takes memory out of the OC equation and can push the CPU even more.
BTW, i hope you are cooling the PWN area with 80/120mm fan right ? If not, do that, keeping the PWM cool, helps in improving stability.
 
OP
H

Harvik780

ToTheBeatOfUrHeart
^^Thanks
I also wanted to install a 120mm fan but the problem is that its not available easily at my place.I'll be ordering for a 120mm fan from theitwares now.
Also the thermal pastes are definitely not properly applied by default.I have checked the thermal pasting on both the cards and they were a bit awwy,..very thin layer of paste on the core chip.I didn't play much though as i am waiting for my Artic Cooling's MX-2.
Well i tried the unlinked mode but i encountered problems running in unlinked mode.With Ram and processor linked orthos did not encounter errors for 5Hrs.
But as soon as I unlink the ram and processor i have problems passing the orthos test even when i set the clocks of the ram to lower values than 800Mhz and the processor is at 2.8Ghz.
I haven't played much though as i was looking for the sweet spot with both processor and ram and since this config does not have a cooler installed.
For the PWN area yes I have already installed an 80mm fan had to do a bit of modding though to install one.
 
OP
H

Harvik780

ToTheBeatOfUrHeart
Yup my processor is a bottleneck.U can make that by looking at the CPU score i have provided.
I again tried ocing my processor...set my ram to 667Mhz and processor to 3.0Ghz and the system did not boot.So its clear that EXX50 series is not a good overclocker.
 

assasin

Banned
^^^ why dont u try ocing ur ram too?maybe it cud help oc the proccy.even i've oced my ram from 667 to 800MHz and its running perfectly stable.

also why dont u try with ur E6600.it sud easily oc to 3.6GHz at 1.41 vCore.
 
Last edited:

deathvirus_me

Wise Old Owl
:p .. or be satisfied at 3.0 GHz at 1.256V :p ... anyway .. whatever it is , SLi is not good from any angle .. benchmark doesn't mean a **** anyway .. and in case a game doesn't support SLi properly , u're literally stuck with one card ...
 
OP
H

Harvik780

ToTheBeatOfUrHeart
assasin said:
^^^ why dont u try ocing ur ram too?maybe it cud help oc the proccy.even i've oced my ram from 667 to 800MHz and its running perfectly stable.

also why dont u try with ur E6600.it sud easily oc to 3.6GHz at 1.41 vCore.

I have already overclocked my Ram to 800Mhz.
I'll oc the e6600 sooner when i get hold of MX-2.

deathvirus_me said:
:p .. or be satisfied at 3.0 GHz at 1.256V :p ... anyway .. whatever it is , SLi is not good from any angle .. benchmark doesn't mean a **** anyway .. and in case a game doesn't support SLi properly , u're literally stuck with one card ...
Most people who get SLI do not play all the games.They buy it for one particular game(s) which has good visual fidelity and obviously supports SLI.
 

go4saket

9437077259
Guys, what do you mean by SLI...
Moreover, I am planning to buy a XFX 8600GT for my system. Is this card going to support all the mordern games till now and will it work fine for the games supposed to be released in a year time...
 
OP
H

Harvik780

ToTheBeatOfUrHeart
^^SLI is nvidia's technology of improving 3D APP performance by using two similar Graphics crads simultaneously.
Don't buy 8600GT 256MB or 512MB its too weak even to run games that are currently available.If ur low on budget get and 8800GT 256MB for Rs 9K.Its a kicka$$ product for resolutions lower than 1280*1024.
 

juggler

In the zone
Harvik780 said:
^^SLI is nvidia's technology of improving 3D APP performance by using two similar Graphics crads simultaneously.
Don't buy 8600GT 256MB or 512MB its too weak even to run games that are currently available.If ur low on budget get and 8800GT 256MB for Rs 9K.Its a kicka$$ product for resolutions lower than 1280*1024.

8800GT 256MB for 9k hmm what is the source for this price??
 

fun2sh

Pawned!... Beyond GODLIKE
one question guys!!! for SLI tech we need same graphic card or even two different graphic card wil also work together??? my cousin yesterday bought 8500gt in 4.5k though i told him to 8400gs coz it was cheap n he didnt wanted to invest in graphics card a lot.
n yeah he has bought a new config yesterday n givin us a headache. currently i m creatin a new thread related to problem
 

darklord

Cyborg Agent
its hard to quantify the difference but memory plays a very important role in performance,thats for sure.
 
Status
Not open for further replies.
Top Bottom