• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

3870 Crossfire.

killersheep34

killersheep34

killersheep34

Associate
Joined
3 Jan 2008
Posts
182
Location
Devon, UK
Hi there, relatively new to the forum but been reading for a couple of months now.

I have just gotten hold of 2 x Radeon 3870's, installed them, and on the whole, I'm not really impressed.

The retailer I have ordered the cards from ballsed up the order first off, basically I ordered 2 of Powercolor 3870 PCS units, and recieved 1 PCS unit, and 1 standard unit. The difference is, the PCS runs at 800/1200 core/mem wheras the non PCS version runs at 777/1126 (or there abouts). This is in hand, but I chucked the 2 in the system anyway to test crossfire performance.

Have done a complete re-install of win xp 32bit.

System specs are:
C2D E6750 @ Stock (2.66ghz)
2gb 1066mhz OCZ memory
Gigabyte GA-P35-DS4
Corsair HX620 PSU

Results at the moment are coming in like this:

Crysis benchmark (i assume this runs at what the game settings are set to)
1440 x 900 no aa
Average FPS single card: 28.44
Average FPS crossfire: 32.31

HL2:LC
1680 x 1050 maxed settings:
Single: 144.49
XFire: 145.34

CS:S
1680 x 1050 maxed settings:
Single: 150.43
XFire: 150.86

3dmark06
1680 x 1050 no aa
Single: 9492 marks
XFire: 9486 marks (this is running in 1680 x 1050 mind)

3dmark 06
1280 x 1024 no aa (default benchmark)
Single: 10366
XFire: 11990

Does this seem to be about right? Initial impressions are not very good to be honest but I'm not sure if this is because either:

a) The cards are different clock speeds and perform quite differently (i also benched the cards seperately, the faster one has about 1500 3d marks more on default settings)

b) My motherboard only supports 16x PCI-e slot 1, 2nd slot 4x

c) I'm doing something completely wrong in setting this up

d) I'm picking games with no CrossFire Support (although I'm sure I saw HL2:LC being benched on these forums with XFire before)

What are your thoughts guys?

Thanks in advance.
 
The PCS version (stands for Professional Cooling System) has the Copper Heatsink on as standard instead of the standard Slot Cooler. It runs nice and cool, but it's noisy as hell!! Not sure why they call it "Silent"!
 
something looks pretty wrong there.... even with 4x you should up your 3Dmark noticably, as it generaly gives a nice boost in Crossfire/SLi.

Are you sure you have the drivers right, and there are no BIOS tweaks or jumpers needed to enable CF? Have you enabled CF in the Cat control panel (not sure if it does this automaticaly).
 
OcUK Call the Powercolor PCS a:

PowerColor ATI Radeon HD 3870 XT Extreme Silent 512MB GDDR4 TV-Out/Dual DVI/HDMI (PCI-Express) - Retail

I have just updated my BIOS and will report back with results.

Thanks
 
Updated Bios to the latest Version (F10) for Gigabyte GA-P35-DS4 Rev 1. No change.
I had already enabled crossfire via the control panel.
I am running Drivers 7.12 downloaded fresh from ATI
I am unaware of any CF jumpers that need to be in place.

Anyone any other Ideas?
 
I wouldn't be impressed with those scores personally...

HL2:LC all settings max, 1680x1050 I get 180fps with my 7950GX2 which is about the same performance as a single 3870 when SLI is working perfectly on my card... even at 1920x1200 I get 145.38fps.

11.3K in 3D Marks 06 would be the top end score for my card without extreme overclocking/cooling... you should be getting atleast 15K in crossfire at 1280x1024 0x FSAA imo... and more like 16-17K.

Wonder if the PCI-e bus is saturated... try increasing the PCI-e frequency a little bit to say 110 and see if it shows any increase in scores... you might be better off returning the cards and waiting for the 3870X2 so you can run it on the pci-e slot at x16.
 
Last edited:
Firstly... now you have crossfire you NEED to overclock your cpu. The setup benefits from faster clock speeds as it now has the crossfire overhead to deal with and feed info to 2 cards instead of just one.

I know the co you mean, and they did the same to me twice. RMA it and get a collection arranged so you don't pay (using their online system), and get the PCS version you wanted. In the end I stuck with the standard card because it was a christmas present, otherwise I would have RMA'd it again.

As for the actual problem, it looks like you have a config error or something. What happens if you declock the first card manually (or OC the slower card to gets clocked in-sync). [and yes I know it 'should happen' automatically]

Matthew
 
Bald-Eagle22 said:
No it seems about right .

I have 2x3870 On P5K Premium 1x16 and 1x4 with a mildly clocked E6600 @ 3.2 .

3d06 On stock gives me 13,085 in Vista 64Bit
Click to expand...

When i get 11860 odd with my OCed Powercolor 3870 (single card) paired with my 3.4Ghz C2D, 12k-13k is very low for crossfire, even with 16x/4x

http://www.anandtech.com/video/showdoc.aspx?i=2990&p=3

Shows there is limite difference between 8x/8x and 16x/4x in 3DMark06, with crossfire, and those scores were from drivers 7 months ago on the then "new" 2900XT..... which has had noticable improvements in results due to driver updates which will have carried forward to the 38xx series.

In short the scores are pretty darn low.

An aside, has anyone else noted the aprox £20+ drop in 3870 price on OCuk? at sub £150 they are starting to look even tastier!
 
Give your CPU a bit of a clock and you'll see a noticeable improvement in your score.

Don't be too disheartened by folks with quads reporting big 3dm06 scores, their CPU's will be worth 3 to 5000 points alone depending on the clock speed.

I initially got disappointing results with SLI, only about 2000 points over a single card, but that was with an E2140 at 3Ghz. Clocking that to 3.4Ghz got me another 1500+ points. I'm sure it won't be any different with Crossfire.

Game performance was where it mattered though, and at any clock it was faster than a single card in Crysis and COD4.
 
There has to be something wrong with your setup.

On the 7.12 I got a very scalable result of around 35-40% performance boost on my system in each test with 2x 3870 (both running stock speeds but different manufacturers). In fact I've been reading on the net and indeed this forum of some people achieving even better virtually doubling frame rates in some games which is exceptionally good.

It's also worth noting my stock speed 3870 outperforms your system in 3Dmark06, yet with your cpu I would have thought it shouldn't!

That said I sold the 2nd card because I simply didn't need it, in CoD4 I run 90fps capped constant at max detail with just 1 card, I simply don't need a 2nd right now.
 
11660 with my system below and thats a single HD3870XT! :D

Right ok I will share my experiences with the dreaded 16x/4x combo! The JUMPER SETTINGS WILL BE WRONG! :D I had to change mines over which took about 20mins to find out what to do and then 1min to actually do! :p My 2600XT crossfire setup was running poorly and I wondered why, it was actually the fact it worked in XP but not in Vista which proved it was wrong too. So once I changed the jumper settings to 4x I got a pretty nice boost on the setup! I think it jumped from 5000 in 3DMark to 9000! So it was all good! You will probably be running on PCi-E 1x at the moment hense the low score so get that sorted and things should look much better. Also clock that CPU!
 
But that score was achieved with a Quad at 3.5Ghz, so not a fair comparison at all. Your CPU score will be more than twice that of a dual core at the OP's clock.

You'd probably get 15,000 on your system with another card, but not on the OP's...
 
You must log in or register to reply here.
Back
Top Bottom