ATI X2 VS. CROSSFIRE

[nel89]

Can someone give me an in depth explanation of X2 and CROSSFIRE is there's a difference of performance and how much MEMORY a Windows will see like i.e. I have a 1 and 1 gig on both cards or X2, How much memory will register to windows?

 

[njalterio]

By X2 if you are referring to ATI's card the 3870 X2, that is one card, with two graphical processing units crossfired together. It is crossfire on a stick as many like to call it. In theory it is the exact same as the two 3870's crossfired together, however I believe the 3870 X2 performs a little bit better then two 3870's crossfired although not by much.

If you are using a 32 bit operating system (lets say XP pro 32) then windows will address up to 4 gigs of RAM including the sticks of RAM, graphics card RAM, and if any sound card RAM/motherboard RAM.

So if you only have two gigs (in sticks) of RAM and two gigs of video card RAM, windows will see almost all of it. However if you are using four gigs (in sticks) of RAM and two gigs of video card RAM, windows will only see two gigs of the stick RAM.

If you need to address more than 4 gigs of RAM then you should use a 64 bit operating system (such as XP pro 64)

 

[mihirkula]

There's a difference of HALF FPS between 3870x2 and 3870 CF. So theres no difference really.

http://www.firingsquad.com/hardware/amd_radeon_hd_3870_x2_performance/page12.asp

The only difference is that the 3870X2 runs a little cooler than two single 3870...

 

[dagger]

The only difference is that the 3870X2 runs a little cooler than two single 3870...

Don't you mean the other way around? 2 3870 each with their own 2 slot stock coolers ought to run cooler than 2 of the same gpus sharing a single 2 slot stock cooler.

 

[doomturkey]

The 3870 reference design is a single slot cooler. I would think heat should be about the same from both, and quite frankly, it shouldn't matter unless you are OCing. I myself would go for crossfired 3870 because its cheaper than getting a 3870X2, and you get about the same performance.

 

[stoner133]

When it comes to memory I believe the OP was talking video memory, on the X2 they list it as 1gb of memory but its really 512 mb per GPU when it comes to windows its only going to see 512mb of the X2 when you enable crossfire.

With two 3870's running in Crossfire if you use 3870's with 512mb per card when you Crossfire the cards windows will still only see 512mb total for video memory. But with the new 3870's with 1gb of memory if you Crossfire two of those windows will see a total of 1gb memory.

Running Crossfire does not raise the amount of video memory. It's simply connecting the two GPU's.

 

[stoner133]

Dagger is correct, the two 3870's will run cooler then a 3870X2. The 3870X2 runs warmer because the fan is drawing the heat from the first GPU over the second on the way out the back of the case. Thus the second GPU doesn't get the cooling as well as the first. Two seperate 3870's each have there own fan to cool each GPU.

 

[nel89]

The ATI X2 has a gddr3 memory and a single 3870 has a gddr4 which is a lot faster chips... So, how come an X2 will be better than two 3870?

 

[nel89]

Running Crossfire does not raise the amount of video memory. It's simply connecting the two GPU's.[/quotemsg]


from the way sounded the second memory of the card for crossfire will be wasted.... So, its better to get an X2 than two cards so that there will be no component that will not be use...

 

[njalterio]

There's a difference of HALF FPS between 3870x2 and 3870 CF. So theres no difference really.

http://www.firingsquad.com/hardware/amd_radeon_hd_3870_x2_performance/page12.asp

The only difference is that the 3870X2 runs a little cooler than two single 3870...

I'm not sure where you learned math, but half of 12 FPS is not 8.7 FPS!

Also, that is just one case, you have to look at the overall benchmark results.

Also, I would stick to sources more reliable then firing squad.

 

[rwayne]

Dagger is correct, the two 3870's will run cooler then a 3870X2. The 3870X2 runs warmer because the fan is drawing the heat from the first GPU over the second on the way out the back of the case. Thus the second GPU doesn't get the cooling as well as the first. Two seperate 3870's each have there own fan to cool each GPU.

I disagree with that. Everything I have read leading up the the release of the HD3870x2 states different.

The HD3870x2 has a better cooler on it than a single HD3870. As a result of the better cooling it allowed AMD to increase the clock speeds on the HD3870x2. That is why the reference clock speed on the HD3870x2 is 825MHz as opposed to 775 MHz for the HD3870.

The 50MHz increase in clock speed per core should make it faster than two HD3870s in X-fire mode (Assuming they are both running at the reference clock speed).

 

[dagger]

I disagree with that. Everything I have read leading up the the release of the HD3870x2 states different.

The HD3870x2 has a better cooler on it than a single HD3870. As a result of the better cooling it allowed AMD to increase the clock speeds on the HD3870x2. That is why the reference clock speed on the HD3870x2 is 825MHz as opposed to 775 MHz for the HD3870.

The 50MHz increase in clock speed per core should make it faster than two HD3870s in X-fire mode (Assuming they are both running at the reference clock speed).

Nevermind that, I just assumed the reference design for 3870 is dual slot, because single slot 3870 is so extremely rare. There is only one single slot model on newegg among dozens, and it's not even the cheapest.

 

[rwayne]

What you are saying about the 2nd RV670 Chip running warmer than the first because of the hot air blowing on it from the first heatsink makes total sense.

But for some reason despite that stock HD3870x2 cooler does much better than the single verison and manages to run both chips at a cooler temp.

You would not think that by looking at it but appearently based on AMD's tests that is the case.

Your thought process makes total logic though.

I wish I could find the original article I found on this stating this information. I think it was on Fudzilla or Tom's Hardware.

 

[dagger]

What you are saying about the 2nd RV670 Chip running warmer than the first because of the hot air blowing on it from the first heatsink makes total sense.

But for some reason despite that stock HD3870x2 cooler does much better than the single verison and manages to run both chips at a cooler temp.

You would not think that by looking at it but appearently based on AMD's tests that is the case.

Your thought process makes total logic though.

I wish I could find the original article I found on this stating this information. I think it was on Fudzilla or Tom's Hardware.

In the 3870x2, the first heatsink is aluminum, and second is copper. That's how they try to even out temperature across the 2 cores.

 

[mihirkula]

I'm not sure where you learned math, but half of 12 FPS is not 8.7 FPS!

Also, that is just one case, you have to look at the overall benchmark results.

Also, I would stick to sources more reliable then firing squad.

I'm not sure where you've learned English for you haven't quite caught up with the sarcasm ... nor do you have the ability to read properly, this is a comparison of X2 vs CF, not Single card vs x2.

Also if you think benchies aren't reliable, why don't you post your own benchies... that way no one would worry about reliability since you're extremely trustworthy and oh so reliable with your mathematical and observational skills.

 

[njalterio]

I'm not sure where you've learned English for you haven't quite caught up with the sarcasm ... nor do you have the ability to read properly, this is a comparison of X2 vs CF, not Single card vs x2.

Also if you think benchies aren't reliable, why don't you post your own benchies... that way no one would worry about reliability since you're extremely trustworthy and oh so reliable with your mathematical and observational skills.

Geez way to get defensive!

And excuse me for reading the graph wrong, but the point still stands. Next time please make your sarcasm more obvious. I do think benchmarks are reliable, but you cant come to reasonable conclusions about the performance of a graphics card from one case of specific settings.