Legion Hardware - 5870 Crossfire CPU Scaling @ 2560x1600

[ArchAngel777]

I'm not sure you can just add the 17% x 2, but maybe someone else can add in on this.

If were comparing a 58XX Crossfire to 5970, then this would be incorrect. It would be 17% faster, not 34%.

 

[toyota]

If were comparing a 58XX Crossfire to 5970, then this would be incorrect. It would be 17% faster, not 34%.

well a 5870 crossfire setup gets 34% better framerates than a 5970 at 2560 in Dawn of War 2. so maybe that means when it scales perfectly at 100% it is indeed 34% faster. http://www.bit-tech.net/hardware/graphics/2009/11/24/multi-gpu-round-up/5

 

[soccerballtux]

i think something is wrong with their testing; especially with Ph II X4

What kind of BS is this? Someone explain this to me:

This part of their conclusion makes zero sense:

like duh
--- i don't see real value in this review

Frankly i am going to do the same thing but simplify the hell out of it
- Phenom II X4 vs Phenom II X2 vs Core i7 920 .. and i will overclock and underclock them. i already tested C2D, C2Q and Phenom II X3 and i hate useless repetition. And of course i would test with CrossFired HD 5870s at 925/1300 to make it more interesting; if i am lucky, i can compare scaling results with GTX 480 SLI. Watch for it an a month or so.

Good points. Don't forget to OC the uncore/CPU-NB/L3$. Would be interesting to see the performance in minimum framerates on a 965 with changes in the uncore--2Ghz vs 2.6ghz etc. Since the L3 is what helps with heavily-branched, unpredictable-branch code, speeding it up aught to significantly improve minimum framerates.

 

[apoppin]

Good points. Don't forget to OC the uncore/CPU-NB/L3$. Would be interesting to see the performance in minimum framerates on a 965 with changes in the uncore--2Ghz vs 2.6ghz etc. Since the L3 is what helps with heavily-branched, unpredictable-branch code, speeding it up aught to significantly improve minimum framerates.

i did that the last time i tested 5 CPUs with HD 4870-x3 TriFire
- that is what overclocking is about; *maxing* the performance from the CPU

i have a Ph II 955-X4 that is stable just a hair under 3.9 GHz
- it is the older stepping

 

[ArchAngel777]

well a 5870 crossfire setup gets 34% better framerates than a 5970 at 2560 in Dawn of War 2. so maybe that means when it scales perfectly at 100% it is indeed 34% faster. http://www.bit-tech.net/hardware/graphics/2009/11/24/multi-gpu-round-up/5

I looked through all those pages... Where are you seeing this?
I don't see anywhere on that page where the 5970 is ever beaten by the 5870 Xfire setup. Although it should be beaten by it, but I didn't see that in those benchmarks you linked.

Lets just put it in simple math, because it doesn't really matter as it will never be 34% improvement, because you are 'double counting'.

The 5970 has to be viewed as a Xfire on a single PCB. So it needs to be treated as 2 GPUs. For the sake of comparison, lets pretend everything is indenticle between a 5970 and a 5870 Xfire as far as clock speed, memory bus and memory bandwidth.

GPU 1 = 500Mhz
GPU 2 = 500Mhz

If you scale 100%, you will have equivilent to 1000Mhz

Now, just because you can overclock both GPU1 and GPU2 10% each, doesn't mean performance has increased 20%. It means performance has increased 10%.

Example:

GPU 1 = 550Mhz (10% overclock)
GPU 2 = 550Mhz (10% overclock)

If you scale 100% you have 1100Mhz, which is 10% greater than 1000Mhz. See? You were double counting.

If GPU1 is overclocked 10% and a GPU2 is not overclocked, you have a 5% gain in performance, not 10%.

Example:

GPU 1 = 550Mhz (10% overclock)
GPU 2 = 500Mhz (no overclock)

If you scale 100% you end up with 1050Mhz, a 5% performance gain.

Again, this is in simplistic terms, with fake clock speed examples, assumes 100% scaling and so on. The math principle remains the same whether or not scaling is 100%. Because both the 5870 and 5970 are dual GPU cards, they will 'presumably' scale in a similar or even indenticle manner. However, there are other things that can come into play here, such as the PCI-E bus speed and probably some other things that I can't think of off the top of my head right now.

 

[SSChevy2001]

well a 5870 crossfire setup gets 34% better framerates than a 5970 at 2560 in Dawn of War 2. so maybe that means when it scales perfectly at 100% it is indeed 34% faster. http://www.bit-tech.net/hardware/graphics/2009/11/24/multi-gpu-round-up/5

Very simple math (((850x2)/(725x2))-1)x100=17.24137931034483%

 

[toyota]

I looked through all those pages... Where are you seeing this?
I don't see anywhere on that page where the 5970 is ever beaten by the 5870 Xfire setup. Although it should be beaten by it, but I didn't see that in those benchmarks you linked.

Lets just put it in simple math, because it doesn't really matter as it will never be 34% improvement, because you are 'double counting'.

The 5970 has to be viewed as a Xfire on a single PCB. So it needs to be treated as 2 GPUs. For the sake of comparison, lets pretend everything is indenticle between a 5970 and a 5870 Xfire as far as clock speed, memory bus and memory bandwidth.

GPU 1 = 500Mhz
GPU 2 = 500Mhz

If you scale 100%, you will have equivilent to 1000Mhz

Now, just because you can overclock both GPU1 and GPU2 10% each, doesn't mean performance has increased 20%. It means performance has increased 10%.

Example:

GPU 1 = 550Mhz (10% overclock)
GPU 2 = 550Mhz (10% overclock)

If you scale 100% you have 1100Mhz, which is 10% greater than 1000Mhz. See? You were double counting.

If GPU1 is overclocked 10% and a GPU2 is not overclocked, you have a 5% gain in performance, not 10%.

Example:

GPU 1 = 550Mhz (10% overclock)
GPU 2 = 500Mhz (no overclock)

If you scale 100% you end up with 1050Mhz, a 5% performance gain.

Again, this is in simplistic terms, with fake clock speed examples, assumes 100% scaling and so on. The math principle remains the same whether or not scaling is 100%. Because both the 5870 and 5970 are dual GPU cards, they will 'presumably' scale in a similar or even indenticle manner. However, there are other things that can come into play here, such as the PCI-E bus speed and probably some other things that I can't think of off the top of my head right now.

yep I looked at the wrong cards.

 

[apoppin]

So is 5870 CF much faster than 5970?
- that is what the entire *premise* of their (maybe faulty) conclusion is based on

 

[soccerballtux]

Good points. Don't forget to OC the uncore/CPU-NB/L3$. Would be interesting to see the performance in minimum framerates on a 965 with changes in the uncore--2Ghz vs 2.6ghz etc. Since the L3 is what helps with heavily-branched, unpredictable-branch code, speeding it up aught to significantly improve minimum framerates.

i did that the last time i tested 5 CPUs with HD 4870-x3 TriFire
- that is what overclocking is about; *maxing* the performance from the CPU

i have a Ph II 955-X4 that is stable just a hair under 3.9 GHz
- it is the older stepping

could you link that?

 

[soccerballtux]

neither of these single out the uncore/L3$/CPU-NB and look at min frames--
http://alienbabeltech.com/main/?p=14309&all=1
http://alienbabeltech.com/main/?p=10085&all=1

 

[apoppin]

neither of these single out the uncore/L3$/CPU-NB and look at min frames--
http://alienbabeltech.com/main/?p=14309&all=1
http://alienbabeltech.com/main/?p=10085&all=1

i just maxed each CPU's performance out; you can certainly see the Phenom IIs (and even the Athlon II in the earlier review) is hanging in with some decent performance - especially once they are overclocked just a bit; there appear to be more differences at stock CPU settings (yes, i think there is some value to O/C'ing).

At the time, the highest resolution available to me was 1920x1600.

My next review will be a shortened version and will use HD 5870 CF and GTX 480 to determine CPU scaling with Phenom II x2 550/X4 955 and Core i7 920. Same three speeds and conditions but now at 25x16 and 19x12.

We already know the differences between the CPUs, now we want to see how they do with faster graphics than HD 4870-X3 TriFire

 

[AzN]

Did this article even have a point?

You shift the GPU bottleneck then CPU starts to matter a lot more.

Why the hell didn't they use any AA in any of these benches? Oh right 2x 5870 just isn't powerful enough for AA nor 3 year old games. :/

Steve who writes for legion is a dumb ass. I remember he made head line news a while back for being an idiot.

 

[toyota]

Did this article even have a point?

You shift the GPU bottleneck then CPU starts to matter a lot more.

Why the hell didn't they use any AA in any of these benches? Oh right 2x 5870 just isn't powerful enough for AA nor 3 year old games. :/

Steve who writes for legion is a dumb ass. I remember he made head line news a while back for being an idiot.

well it did show that many cpus are quite poor to pair up with that gpu setup in the first place. 2560x1600 is pretty freaking gpu demanding as it is and some of those cpus looked quite poor especially in the min framerate department. it backs up xbit in showing that the Phenom 2 X4 cant deliver in the min framerate department compared to even the Core 2 Duo in many cases.

I wish they would have used more games. I also would liked to have seen 1920 used in this test along with some AA used at the different resolutions. that still would not help the cpus that are clearly outclassed for this level of graphics cards though.

 

[AzN]

well it did show that many cpus are quite poor to pair up with that gpu setup in the first place. 2560x1600 is pretty freaking gpu demanding as it is and some of those cpus looked quite poor especially in the min framerate department. it backs up xbit in showing that the Phenom 2 X4 cant deliver in the min framerate department compared to even the Core 2 Duo in many cases.

I wish they would have used more games. I also would liked to have seen 1920 used in this test along with some AA used at the different resolutions. that still would not help the cpus that are clearly outclassed for this level of graphics cards though.

A guy with 2x 5870 or 5970 is going to play with AA. That's a fact.

It's like testing GTX 260 @ 800x600 without AA. It's silly! There's no point!

 

[apoppin]

well it did show that many cpus are quite poor to pair up with that gpu setup in the first place. 2560x1600 is pretty freaking gpu demanding as it is and some of those cpus looked quite poor especially in the min framerate department. it backs up xbit in showing that the Phenom 2 X4 cant deliver in the min framerate department compared to even the Core 2 Duo in many cases.

I wish they would have used more games. I also would liked to have seen 1920 used in this test along with some AA used at the different resolutions. that still would not help the cpus that are clearly outclassed for this level of graphics cards though.

no, the article actually doesn't show anything
- we suspect the Phenom II s were even set up incorrectly

he makes ridiculous assumptions in his conclusion assuming the 5870 CF is a lot faster than HD 5970

So what did you get out of it other than the testing needs to be completely redone?

 

[toyota]

A guy with 2x 5870 or 5970 is going to play with AA. That's a fact.

It's like testing GTX 260 @ 800x600 without AA. It's silly! There's no point!

I agree with that and I just said he should have added some AA results. if the test would have been at 1680 and 1920 then you would have complained they didnt use 2560. again though it clearly showed that some cpus are not adequate for that level of gpu power. adding AA isnt going to help the really slow cpus out anyway. 15-20fps minimums suck whether you have AA on or not. I would never personally use 5870 crossfire or a 5970 on anything less than an i5 750. that cpu is cheap and will give you basically 100% of what any graphics card setup can do especially if oced to around 3.6.

 

[soccerballtux]

neither of these single out the uncore/L3$/CPU-NB and look at min frames--
http://alienbabeltech.com/main/?p=14309&all=1
http://alienbabeltech.com/main/?p=10085&all=1

i just maxed each CPU's performance out; you can certainly see the Phenom IIs (and even the Athlon II in the earlier review) is hanging in with some decent performance - especially once they are overclocked just a bit; there appear to be more differences at stock CPU settings (yes, i think there is some value to O/C'ing).

At the time, the highest resolution available to me was 1920x1600.

My next review will be a shortened version and will use HD 5870 CF and GTX 480 to determine CPU scaling with Phenom II x2 550/X4 955 and Core i7 920. Same three speeds and conditions but now at 25x16 and 19x12.

We already know the differences between the CPUs, now we want to see how they do with faster graphics than HD 4870-X3 TriFire

yeah but you guys didn't overclock the L3 cache at all which is of utmost importance when you push the Ph2 past ~3.2ghz.
2.6ghz or 2.8ghz on the L3 $ seems to be the sweet spot for 3.4-4Ghz

 

[apoppin]

yeah but you guys didn't overclock the L3 cache at all which is of utmost importance when you push the Ph2 past ~3.2ghz.
2.6ghz or 2.8ghz on the L3 $ seems to be the sweet spot for 3.4-4Ghz

i did overclock some individual settings but i mostly aimed for the highest core speed possible. i also did not absolutely tweak every possible setting for i7 or C2Q

However, even with a "disadvantage" at 3.8 to 3.9 GHz, the Phenom IIs do rather well against their competition.

For my next article i will pay particular attention to L3 cache settings.
Thanks!

 

[AzN]

I agree with that and I just said he should have added some AA results. if the test would have been at 1680 and 1920 then you would have complained they didnt use 2560. again though it clearly showed that some cpus are not adequate for that level of gpu power. adding AA isnt going to help the really slow cpus out anyway. 15-20fps minimums suck whether you have AA on or not. I would never personally use 5870 crossfire or a 5970 on anything less than an i5 750. that cpu is cheap and will give you basically 100% of what any graphics card setup can do especially if oced to around 3.6.

Why 2560x1600 and 1680x1050? Why not 800x600 if he's trying to find CPU scaling?

All steve did was shift the GPU bottleneck to CPU to come up with some silly assumptions. Not to mention some of these games are nearing 3-4 years old.

His silly minimum fps seems way off to me in some of these benches especially Crysis and WIC. Crysis just doesn't scale well with faster CPU nor does faster CPU make any difference in a GPU limited situation. I have that games I know I tested multiple platforms which CPU doesn't do diddly squat.

 

[RussianSensation]

So is 5870 CF much faster than 5970?
- that is what the entire *premise* of their (maybe faulty) conclusion is based on

Check out xbitlabs review of GTX480 which includes the 5970: http://www.xbitlabs.com/articles/video/display/geforce-gtx-480_6.html#sect0

5970 does NOT produce consistently good minimum framerates, often being slower than GTX480 and even 5870 while destroying both in averages. Saying 5870CF is significantly faster than 5970 is incorrect. But 5870 CF appears to be significantly faster than 5970 when it comes to minimum frames. Since higher minimum frames improve playability, 5870 CF is a faster setup as a result. We have known for years that a faster CPU improves minimum frames, but no one seems to bring this point up.

 

[RussianSensation]

Did this article even have a point?

Why the hell didn't they use any AA in any of these benches? Oh right 2x 5870 just isn't powerful enough for AA nor 3 year old games. :/

Even without AA, some of these CPUs are providing horrendous mins. Sorry not everyone wants to play with 15 mins and 8AA. I'd rather take 30 mins and 0AA.

 

[RussianSensation]

His silly minimum fps seems way off to me in some of these benches especially Crysis and WIC.

With all due respect, other websites have shown similar results: http://www.xbitlabs.com/articles/video/display/radeon-hd5870-cpu-scaling_4.html#sect0

World in Conflict 2560x1600 8AA/16AF 5870 CF
Core 2 Duo 4.1ghz = 23 min / 71 avg
Phenom II X4 4.1ghz = 17 min / 67 avg
Core i7 4.1ghz = 37 min / 93 avg

Also Core 2 Duo 2.7ghz is giving 10 fps in Crysis.

 

[soccerballtux]

With all due respect, other websites have shown similar results: http://www.xbitlabs.com/articles/video/display/radeon-hd5870-cpu-scaling_4.html#sect0

World in Conflict 2560x1600 8AA/16AF 5870 CF
Core 2 Duo 4.1ghz = 23 min / 71 avg
Phenom II X4 4.1ghz = 17 min / 67 avg
Core i7 4.1ghz = 37 min / 93 avg

Also Core 2 Duo 2.7ghz is giving 10 fps in Crysis.

He didn't overclock the L3 cache either. NB speed is different from CPU-NB.
Bench sites these days don't know anything.

 

[biostud]

<3 my i5-750

 

[nyker96]

the intel i5 i7s are just incredible in gaming. looks like the best combo nowadays is a i5 750 with something like 5970 for serious gaming. I've always known my x4 620 is weak at gaming but the benchmark seem to suggest the cache is really slowing it down. i5 750 looks quite promising as my next upgrade.