My pearly 940X4 and a 4850X2 + 4850

[razor2025]

Interesting results. Not too surprised though, 4850x2 has shown to edge out 4870x2 in certain high-res, high-memory situations (especially at your res and AA levels). Good luck with your setup, definitely looks killer.

 

[Antman56]

Originally posted by: razor2025
Interesting results. Not too surprised though, 4850x2 has shown to edge out 4870x2 in certain high-res, high-memory situations (especially at your res and AA levels). Good luck with your setup, definitely looks killer.

Thanks, I appreciate it. O yeah, by 4850X2 you meant 4850X3, right?

 

[MegaWorks]

Radeon 4850 owners if you want to CF make sure you have bios v.105 or higher to avoid the Blue screen of death!

To flash your bios use Winflash 2.0.1.5 and Radeon Bios Editor to force the bios.

I hope this is important information to thread Antman56.

 

[Elfear]

Originally posted by: Antman56


That's interesting. If I had your setup, I would get a cheap "drop in" quad Q-something Intel and overclock it the solid high 3 GHz or even a low 4 GHz range. The difference you read about, in that review link you posted, is the difference of frame rates undetectable by the human eye. I would be surprised if an i7 would increase my frame rate in Far Cry 2 with 8xAA enabled. If anything... I don't have two+ GPUs running in Crossfire to do 1920x1080 with anything less than 4xAA. A non-AA benchmark would not apply to me. Surely those frame rates would be a good bit lower with 4x or 8x AA enabled (Though we are talking about a Tri-SLI 280s). Either way, I could not justify skipping a simple sub-200 dollar upgrade for a 600+ dollar one... just to have a higher frame rate that I couldn't even "see" (anything > 60 FPS) when AA is disabled.

Just a suggestion, not an attack. If money is not an object, then definitely get an i7.

Also, I can tell that considering a Phenom II wouldn't be optimal either. You are already setting on a setup with higher performance potential (with a CPU upgrade + overclocking) than any Phenom II has to offer. I have a Phenom II because I already had a AM2+ motherboard, so my upgrade is similar to one I think you should make.

Edit: I needed to fix some typos

You make some good points but what amazed me about the benchmarks I linked to was the domination the i7 had at even the highest resolution. There are lots of reviews out there where processor A will be xxfps ahead of processor B but it is always at unrealistic resolutions like 800x600. This is the first time that I can remember when high resolutions showed such a disparity between processors and that in a pretty graphics intensive game.

I guess my thinking is that even though the framerate is acceptable for both processors at 2560x1600 (what I use), when situations arise where framerate dips into the noticeable range, the i7 will be a much better choice. It should be noted however that tri-SLI has greater cpu overhead than two cards would require. I'd like to see a processor shootout at various resolutions using single and dual cards and see how things end up.

 

[Antman56]

Originally posted by: MegaWorks
Radeon 4850 owners if you want to CF make sure you have bios v.105 or higher to avoid the Blue screen of death!

To flash your bios use Winflash 2.0.1.5 and Radeon Bios Editor to force the bios.

I hope this is important information to thread Antman56.

Actually, this is a big helper... I have been posting this on ATI's CrossfireX forum because a trillion people are having crossfire problems. Thank you very much for the link.

 

[Denithor]

The DMC4 benches show nearly perfect CF scaling & only minor differences between stock & overclocked cpu/IMC/GPU. I know you don't have the card but it would be interesting to see what kind of fps you'd get with 4870 x1/2/3 on your CPU setup. Even with three 4850s you might still be a bit GPU limited (top end doesn't go up much when adding extra cpu muscle).

The FC2 minimum fps graph is the most interesting to me. You are clearly GPU limited with the cards at stock speed (OC the GPU - perfomance goes up nicely). This holds true except in the case of the overclocked IMC, apparently the extra bandwidth allows the card to process data more efficiently or something.

Is it possible to OC the IMC without overclocking the core speed on the PhII? If so it would be interesting to see this benchmark repeated with stock CPU/GPU and an overclocked IMC (as high as you can get it stable).

It's also worth noting - at the top end, with the 3.53/2.28 there's not much difference between the stock vs OC GPU. That tells me you've become CPU limited to some degree. Wonder how an i7 would push that trio of cards?

 

[SunnyD]

I have to wonder if your 790GX's 2x PCIe 8x (when in Crossfire) might be limiting you somehow. I'm not saying it is, I'm just wondering.

 

[Antman56]

Originally posted by: Denithor
The DMC4 benches show nearly perfect CF scaling & only minor differences between stock & overclocked cpu/IMC/GPU. I know you don't have the card but it would be interesting to see what kind of fps you'd get with 4870 x1/2/3 on your CPU setup. Even with three 4850s you might still be a bit GPU limited (top end doesn't go up much when adding extra cpu muscle).

The FC2 minimum fps graph is the most interesting to me. You are clearly GPU limited with the cards at stock speed (OC the GPU - perfomance goes up nicely). This holds true except in the case of the overclocked IMC, apparently the extra bandwidth allows the card to process data more efficiently or something.

Is it possible to OC the IMC without overclocking the core speed on the PhII? If so it would be interesting to see this benchmark repeated with stock CPU/GPU and an overclocked IMC (as high as you can get it stable).

It's also worth noting - at the top end, with the 3.53/2.28 there's not much difference between the stock vs OC GPU. That tells me you've become CPU limited to some degree. Wonder how an i7 would push that trio of cards?

It's pretty nice to be able to put some numbers with my tweaks. The IMC definitely helps in CPU bottlenecked situations. You did notice that I ran the Phenom 2 at 3.5 with the IMC overclocked and IMC at stock clock, right? That should give you an idea of the effect...

Furthermore, the only games that really show great crossfire scaling throughout are the ones that use more than 2 CPU cores (makes since). Obviously, two 3.53 cores are simply not enough to keep Three 4850's busy. Utilizing all four cores, Devil May Cry 4 and Far Cry 2 in this review , gives the Phenom leverage it needs to push these cards to a WAY harder.

*Half-Life saw a nice performance jump when I overclocked the IMC compared to its equivalently clocked 3.5GHz non-IMC modded competitor*

An i7 with my setup would be interesting... that is for sure. I am currently getting a GPU score equal to a Geforce 295 in Vantage and I am getting some incredible frame rates at the high resolutions (even in Far Cry 2) with some seriously taxing settings applied. No complaints... at all.

 

[Antman56]

Originally posted by: SunnyD
I have to wonder if your 790GX's 2x PCIe 8x (when in Crossfire) might be limiting you somehow. I'm not saying it is, I'm just wondering.

To be honest... I really doubt it. If so, I would think it would be within 5 percent (definitely not worth the price difference of getting a 790FX with a R750 Southbridge)

 

[Antman56]

In my corresponding thread at Tomshardware.com, a fellow forum member cjl posted his results at the same settings as me using an i7 965 with a 4870X2. Interesting to say the least.

Originally posted by: cjl at Tomshardware.com
OK, for a run near the beginning of FC2, I get the following (fully cranked, 1920x1080):


Stock clocked i7 965 (3.2) + stock clocked 4870x2:
Min: 37
Max: 61
Avg: 49


Oddly enough, upon reopening the game to run another test at 4.0GHz, I find that the 1920x1080 option has vanished, so here's a run at 1920x1200:


Min: 30
Max: 55
Avg: 44.6


I'll try to get a 4GHz run in at 1920x1080 though - if I get it to work, I'll post the results.


(these are all 60 second averages btw...)


EDIT: still can't get in a 1920x1080 run at 4GHz, but here's one at 1920x1200, 4GHz, 4xAA (rather than the 8x AA used above - all other settings identical):


Min: 47
Max: 108
Avg: 68


These may be the settings to use for actual gameplay - the 4x is a noticeable bump in FPS, without a noticeable drop in quality.

My frame rates are slightly higher at 1920x1080 with 8xAA and 16AF, but I know I don't have to processor power to push my cards as hard as they could be. Now if he had my GPU setup with his processor... that would be serious.

On my Phenom 2 machine, I would not get as high of a frame rate as him at 1920x1200 resolution with only 4xAA. The stress on the GPUs is substantially reduced, thus making the processor more dominate in pushing up the frame-rate. I'm sure my frame rate would not increase that dramatically (since my processor is already maxxed out).

The main reason I tested this resolution was to demonstrate my TriFire potential in a GPU-stressed situation (where my processor would be the root of weak scaling).

 

[Denithor]

What I was getting at earlier - I think the IMC bandwidth has a large impact on GPU performance (probably keeps the GPU cores fed better & running smoothly instead of in bursts as data trickles in for processing). That's why I wondered if you can OC the IMC without overlocking the CPU core speed (3GHz, 2.28GHz IMC) and also just see how high you can push the IMC period and test there.

 

[Antman56]

Originally posted by: Denithor
What I was getting at earlier - I think the IMC bandwidth has a large impact on GPU performance (probably keeps the GPU cores fed better & running smoothly instead of in bursts as data trickles in for processing). That's why I wondered if you can OC the IMC without overlocking the CPU core speed (3GHz, 2.28GHz IMC) and also just see how high you can push the IMC period and test there.

I apologize for the delay. I am going to go ahead and give that a shot. I'll you know how it turns out.

 

[Denithor]

There is some additional interest on this topic (over here). If you're still around we'd like to see some further work...

 

[soccerballtux]

Very interesting.