facebook rss twitter

Review: CrossFire and SLI dongle-less / bridgeless performance analysis

by Steve Kerrison on 18 September 2006, 09:11

Tags: ATi Technologies (NYSE:AMD)

Quick Link: HEXUS.net/qagq3

Add to My Vault: x

Final thoughts

First things first. Bridgeless SLI... why bother? If the golden fingers are there, use them. On high end cards it clearly makes a difference, and with SLI there's no faffing about with master cards, so it's really no big deal. Still, if there's a problem with the bridge, we've shown the performance impact, which is in some cases substantial.

For CrossFire, there's a little more appeal to not having a dongle. It's not the dongle itself that's the problem, it's the expense of a CrossFire Edition master card. Take the X1900 GT, which at the time of writing weighs in at around £120-150, depending on bundle, cooling and what not. A CrossFire edition card to go with that could be £220 or more. So we must ask if the sacrifice made with dongle-less CrossFire is worth the ~£100 saving? It all depends on whether you want the most from your multi-GPU solution, or if all you really need is a boost that's good value for money. If you're after the latter, we reckon a pair of X1900 GTs might be the wiser option.

Problem is, on our i975X test bed, CrossFire performance and indeed the impact of not having a master card varies so wildly, that the benefit you'll see really will depend on what games you play. We'd like to think ATI can make performance differences a little more consistent between dongled and dongle-less CrossFire, but we'll have to wait and see.

Regardless of the outcome, we have hopefully investigated something that's of interest to our readers. This reviewer was certainly intrigued by the outcome of the tests. The bottom line is SLI works best with the bridge, and there's no good reason to be without it. CrossFire, however, does have some scope for dongleless usage, unless ATI has any other plans in the pipeline...



HEXUS Forums :: 11 Comments

Login with Forum Account

Don't have an account? Register today!
a quick comment…

afaik ATI when mentioning dongleless crossfire said that a motherboard with 2 16xspeed pci-e slots would be required (not 2 8xspeed slots like the x975 chipset) as this would give enough pci-e lanes for the bandwidth to not be an issue.

As you tested a board with 2 16xspeed lanes (Nforce590) against one with 2x 8speed lanes surely this test is not totally fair.

Although I realise Core2Duo does not have any crossfire mobos with dual 16xspeed slots until RD600 arrives, you could have used an amd-based system as although it would have been slower overall it would imo show a fairer comparison.

e.g. the Asus ASUS M2R32-MVP vrs the Asus M2N32-SLi?

just my 2 pence, otherwise an interested review :)
thanks for your input YorkieBen
YorkieBen
until RD600 arrives
no-one should hold their breath on this…

with the AMD-ATi merger, politics appear to have already kicked in… leaving third party manufacturers, who had planned, and counted on quickly bringing to market products based on ATi RD600 very frustrated indeed…

and further, some of these manufacturers are interpreting the signs that RD600 may never even see the light of day…

if you've not yet read this elsewhere, then remember where you heard it first… :)

so much for the (apparently AMD led) rhetoric about the future plans of the merged company being fully ‘inclusive’ of other technology providers (presumably such as Intel and NVIDIA)… as a long term plot, i sure didn't buy it for a minute… and i suspect that the landscape will be very different in 24 months or so…

cheers,

PD
If anyone thinks that the merger will force people to pair AMD + ATI and Intel + nVidia, I think they are quite short-sighted.

It would be killer for the companies to effectively lock people in more then they are now at the same time as narrowing their potential customer base.

I am sure for now, there are some confused people and that is the reason the RD600 seems up-in-the-air, but give it a couple of months and i think we will be back on track :)
I find it a little odd when looking at the Crossfire Splinter Cell results you suggested the dongle may not be performing as well as it could, despite getting an increase of 97% over the single X1900GT. Seems like it's more a case of the non-dongled version doing well than the dongle doing bad to me.

Still overall was quite interesting. :)
I completely agree having 8x8 vx 16x16 is unfair, thry it on a RD580 board you will find crossfire improves around 15+%, or so the case was with dual X1600XT cards in 16x16 board then over a 8x8 which people tested a ASUS MPV Vs and ASUS MPV32 sk939. At the moment this is not apples to apples as nvidia have a 100% bandwith peformance increase.