New Products Prev. ASUS P5Q3 Deluxe WiFi-AP (Intel p45  - 1600 FSB DDR3) New Products Next
Pictures Prev. Pictures Next
Fusion-CS-Intel-775G2
What's your view on SSDs?
Unfortunately as from 07/07/08 V-Solutions LTD has ceased trading.
In this article we have tested a crossfire X1800XT setup from ATI against the Nvidia's 7800GTX512 in SLI. Both rigs were watercooled and benchmarked pretty well. It seems that ATI finally managed to produce something to put it back in the game.

Date: 15/01/2006 Discuss in forums Author: The Editor

Back in October, we discussed the x1800 graphics cards from ATI, showing the watercooled potential and 3dmark05 benchmark scores. Since then we have been waiting for ATI’s master cards to see how they will do in Crossfire compared to Nvidia’s SLI.
However, in mean time Nvidia tried to make a mess of the ATI’s launch by releasing their 7800GTX512, where a single card scores better than a single X1800XT.
Fortunately, we had the opportunity to play with a few pairs of watercooled 7800GTX512 in SLI and the best results we had were just below 14800 in 3dmark05. The test system was an Asus A8N32 SLI with an overclocked AMD 4800x2 and 2x 1GB Corsair DDR433 CL2. The results can be found here:
http://service.futuremark.com/compare?3dm05=1620240

Internal setup:

External view:


To be honest we felt somewhat deceived by Nvidia and whatever our attempts were, we just couldn’t get any higher and that is only about a 40% increase over the single card results.

Well, this week, a month later than we expected finally we have the x1800 master cards that allowed us to run a 2x X1800 crossfire setup. The PCBs are very similar one to each over except that the master card has a few more silicon chips and a crossfire processing unit. The GPUs and memory seem to be exactly the same:

Slave card PCB:

Master Card PCB:


As you can guess, we have watercooled them with the same VGA blocks from Alphacool and only difference between the Master and Slave card is the actual crossfire chip that we’ve cooled with a Zalman aluminium heatsink as it can’t be covered by the actual waterblock:

Watercooled Master Card:

Well, next is the system setup. We’ve managed to implement our Liquocool Antarctic D in a Gigabyte 3D Aurora with some modifications to the actual case design. The dual Radiator was mounted on the back after dremeling some bits of the case and the single Radiator has been placed on the front after dremeling off the HD Cage.

Other Hardware used:
Mobo: Sapphire Crossifre
CPU: AMD4400 @ 244x11 = 2684Mhz
Memory: Corsair DDR550 CL 2.5 (488 MHz @ 2.5-3-3-7)
HDs: 2 x 74 GB Raptors in Raid 0

Next we had the cooling setup with our 24hours pressure testing of the loop:

Pressure testing:

Benchmarking/Overclocking:
The initial 3dmark05 score was 11888 without any CPU overclocking/memory tweaking.
Then we raised the CPU bit by bit until we reached the maximum the board and CPU allowed and then we tightened the timings as much as we could.
The result were a linear increase with the coefficient of 0.81 in the actual 3dmark score – i.e. every 1% increase of the CPU resulted in 0.81% increase in the gaming benchmark. So the new score was 14009! http://service.futuremark.com/compare?3dm05=1705024  – that is 17.84% increase from just the CPU overclocking and memory optimisation! (That was primed for 3 hours to check the stability)

From the past experience we know that the GPU/GDDR overcloking resulted in a 1 to 1 linear increase in performance and we were able to achieve a 7% increase from just overclocking of the x1800xt graphics without going into any extremes – just to 700/1600. We thought that we will get a 7% increase from that too and theoretically the scores of crossfire x1800 should’ve been 14989. We got excited as to beat our SLI 7800GTX512 scores of 14771 and decided to sacrifice a weekend to play around with the GPU and GDDR overclocking. All this resulted in numerous windows installations with many unsuccessful attempts to overclock using various overcloking utilities including the ATI control panel. On the second day (Sunday) we gave up as even the ATItool wasn’t successful in our attempts – it just resulted in worse scores.

Final Compared Results:

As the results show, ATI does better in the first tests, but still trails the Nvidia due to the bigger gap in the 3rd test. So the conclusion is simple: If the crossfire overclocking problems are rectified (which will take a while like with the single cards) then we clearly have a winner here and some might not like it, but this is the fact:

2 x X1800XT in Crossfire perform the same as the Nvidia’s 2x7800GTX512 for a lot less money!

Update:
19/01/06 Yesterday after the release of the 6.1 drivers we tried to run the tests again, but this time with a overclocked FX60 @ 2.85Ghz and low low memory timings - 2-3-2-6 this time it scored 14716 3dMark05 points!