HD 3870 Crossfire Scaling: X38 vs P35

[MeltDown]

http://www.xcpus.com/forums/graphic-...38-vs-p35.html

Yeah I know, I've been working on this since Christmas. Anyway, the basic gist of the article is up now - I've still got a few more changes (and one graph to fix) but you can take a look now.

[SirRobin4ever]

Great article! I've yet to see one that compares this using 3870s as of yet. Now I have!

I hate to be the one that always criticizes, because I do the same thing w/ my mini reviews...but on the final graphs comparing the overall performance at the end of each page should really have the resolution on it if at all possible. If thats a lot of work, don't bother, its minor really. Again, congrats!

[Eurasianman]

Man, I'm glad I got an HD 3870X2!

Good job Meltdown! Even though it took you months

Nonetheless...

[MeltDown]

I hate to be the one that always criticizes, because I do the same thing w/ my mini reviews...but on the final graphs comparing the overall performance at the end of each page should really have the resolution on it if at all possible. If thats a lot of work, don't bother, its minor really. Again, congrats!

Yeah it's a pain but I'm doing it anyway... due to my mediocre skills with Excel all resolutions have to be added via post-processing with PDN.

[ABRaquel]

Great review.

I would love to see the results with CrossFireX in use, that is if you have a 3870X2 :P

[rugrat2553]

Great review and hope to see other sites linking to it.


EDIT: Nevermind, I finally read what the caption says

Thanks Randomizer

[randomizer]

On the X38 board, Crysis ran very well with medium settings (and yes I'm aware that I have a copy/paste error in the graph). Next the same test was repeated on the P35 board.

[Clue69Less]

Awesome work Meltdown.

[MeltDown]

OK I added the scaling efficiency charts and corrected the Crysis charts (I had copied the max frame rate from the next loop into the field for average framerate at 1024x768). The resolutions on the comparison plots will have to wait until tomorrow before I finish them.

[baligavinod]

very nice melty. was worth the wait

[MeltDown]

Great review.

I would love to see the results with CrossFireX in use, that is if you have a 3870X2 :P

I unfortunately do not have a 3870X2, but if I get a hold of one once CF-X drivers come out I will try to add it to this (at least on the X38 board, I'm pretty sure the P35 will cripple it).

[YMAA]

Nice work. Looks like for those who want Crossfire we now have proof that X38 is the better option in GPU-limited games. But it's also good to see that current P35 users aren't that far behind.

[prozac26]

You know Melty,

I had the pic as my avatar, now that I've veen reminded of it...

"ATI, The way its meant to be played"

[Eurasianman]

Meltdown, I notice some minor things.

First off, in the majority of your graphs, when you show the graphs comparing the P35 vx X38 in crossfire, you're missing the resolutions.

Secondly, in your conclusion, the 2nd paragraph has a minor typo:

For people who have X38 boards and want to run Crossfire, the performance testing showed it to be a very effective way of improving performance. For people with P35 boards, the only way two HD 3870s in Crossfire would really be recommended is if you already own one HD3870 and want a way to add performance in GPU-challenging games like Crysis or Oblivion. Otherwise, we would recommend instead of two HD 3870s that these users consider a more powerful single graphics card such as an HD 3870X2 which provides Crossfire-level performance on a single card. Consider that the price difference between a P35 board and an X387 board is greater than the price difference between two HD 3870s compared to a single HD 3870X2, the combination of a P35 and an HD 3870X2 looks more appealing for new system builds unless Crossfire performance in the future is important.

Other than that, I could run some tests for you with a Quad and the HD 3870X2, granted, I can't do it with the CPU at 4.0 GHz

[Simul4ti0n]

Looks like the 38x0 series is where it's at. Very nice article. I might have to trade up my 8500GT for a 3870x2

[1Tanker]

Nicely done review Melty. CF/SLI are just like multi-core CPU's. Some apps/benches put the extra core(s) to good use, and some don't. At least in the CPU department, driver optimizations aren't a determining factor.

[MeltDown]

Meltdown, I notice some minor things.

First off, in the majority of your graphs, when you show the graphs comparing the P35 vx X38 in crossfire, you're missing the resolutions.

Secondly, in your conclusion, the 2nd paragraph has a minor typo:


Other than that, I could run some tests for you with a Quad and the HD 3870X2, granted, I can't do it with the CPU at 4.0 GHz

First point: I know, I'm going to update those tonight. Sirrobin already pointed that out as well.

Second point: FF's spellchecker didn't catch that one since it doesn't know X38 from X387 and I relied on it a bit too much it seems . Fixed and thanks.

[rugrat2553]

Meltdown, I notice some minor things.

First off, in the majority of your graphs, when you show the graphs comparing the P35 vx X38 in crossfire, you're missing the resolutions.

Secondly, in your conclusion, the 2nd paragraph has a minor typo:


Other than that, I could run some tests for you with a Quad and the HD 3870X2, granted, I can't do it with the CPU at 4.0 GHz

Get in your truck and drive to Georgia, I will be taking one out tonight.

[Eurasianman]

A QX9650?!

*Leaves work and speeds off to Georgia*

[rugrat2553]

A QX9650?!

*Leaves work and speeds off to Georgia*

I didnt say you could have it. Maybe use and abuse it for the comparison and Forum Wars. Then IT WILL BE RETURNED

*readies arsenal for when Eurasianman does not return*