2.72Ghz Conroe, X1900CF @ 712/738 takes AM3 WR on air

[Carfax]

AM3 is a shader intensive benchmark, so obviously it would get higher scores on a X1900CF platform which has 48 pixel shaders per core.

Conroe is an amazing CPU, but I'm sure the score is mostly attributed to using the R580..

[freecableguy]

AM3 is a shader intensive benchmark, so obviously it would get higher scores on a X1900CF platform which has 48 pixel shaders per core.

Conroe is an amazing CPU, but I'm sure the score is mostly attributed to using the R580..

Go ahead and ask anyone you know with the right cards to set their x1900 cards to 712 core and 738 mem and see if they can beat me. The cards are not what made this score possible. It was the CPU.

[deathman20]

Congrats on that clock. Makes me wish I didn't upgrade this past fall but oh well somethings always greener around the corner

[Carfax]

Go ahead and ask anyone you know with the right cards to set their x1900 cards to 712 core and 738 mem and see if they can beat me. The cards are not what made this score possible. It was the CPU.

OK I believe you.. I didn't think anyone had benched with X1900CFs, but then again, synthetic benchmarks don't really interest me..

So, the only other explanation must be that the entire benchmark itself can fit into Conroe's L2 cache, and combined with Conroe's extensive execution resources, enables Conroe to get more work done in a fraction of the time it takes the K8..

AM3 is probably INT intensive (most game code is), which explains why Conroe and Dothan do so well.

If there's one area where Conroe just absolutely rapes the K8 (and every other processor), it's in INT..

[RyderOCZ]

Tony and I were on the phone and for S's & G's just ran 3D05 with 2 x X1800 in CF at identical clocks to FCG's cards (X1800 and X1900 have same number of shaders and pipes...X1900 is mostly just a die shrink)
Opty 165 Clocked to 2.9+ GHz, OCZ PC4000 EB 2GB kit at 267MHz 3-3-2 timings and only scored 15,3xx. So you can see the CPU making up the 1500+ points

[Carfax]

Tony and I were on the phone and for S's & G's just ran 3D05 with 2 x X1800 in CF at identical clocks to FCG's cards (X1800 and X1900 have same number of shaders and pipes...X1900 is mostly just a die shrink)
Opty 165 Clocked to 2.9+ GHz, OCZ PC4000 EB 2GB kit at 267MHz 3-3-2 timings and only scored 15,3xx. So you can see the CPU making up the 1500+ points

Sorry you're wrong there dude. The X1900 core has three times the amount of shader units as the X1800 does..

They have the same amount of pipelines though..

[deathman20]

Tony and I were on the phone and for S's & G's just ran 3D05 with 2 x X1800 in CF at identical clocks to FCG's cards (X1800 and X1900 have same number of shaders and pipes...X1900 is mostly just a die shrink)
Opty 165 Clocked to 2.9+ GHz, OCZ PC4000 EB 2GB kit at 267MHz 3-3-2 timings and only scored 15,3xx. So you can see the CPU making up the 1500+ points

X1900's have 48 shaders not the 16 that the X1800 has. And didn't think the X1900's had a die shrink yet, thought they where using the 90nm dies.

[Vapor]

And, the total score is what counts....a lot of what this is coming from is actually a VERY high GPU score because Conroe removes all the driver overhead.

[ex2cib]

FCG whats the max & min fps for that run?

[freecableguy]

OK I believe you.. I didn't think anyone had benched with X1900CFs, but then again, synthetic benchmarks don't really interest me..

So, the only other explanation must be that the entire benchmark itself can fit into Conroe's L2 cache, and combined with Conroe's extensive execution resources, enables Conroe to get more work done in a fraction of the time it takes the K8..

AM3 is probably INT intensive (most game code is), which explains why Conroe and Dothan do so well.

If there's one area where Conroe just absolutely rapes the K8 (and every other processor), it's in INT..

i can guarantee you that the AM3 benchmark does not fit into the Conroe's 4MB L2 cache. stop trying to rationalize this with unfounded theories. why is it so hard for people to accept the power of Conroe? i getting so sick of posts like this, i'm about ready to stop making this info public and just share it with my friends who i know are intelligent enough to look at it in the right light....

Edit: in fact, someone load AM3 into memory and tell me the size....I'm going to say somwhere in the range of 16-18MB is resident at any given time. then someone look at the size of the program on disk...I don't know...100MB? 4MB of memory is a lot for a CPU's cache, but remember, it's just that....4 MB....not a lot in actuality. bottom line, knock off the "fitting in cache" statements. enough of this crap.

[Vapor]

He might be onto something...(well, not really, but hear me out, lol).

One thing we're not seeing with Conroe is the really HUGE jump in CPU scores that we were expecting (a la Pi times and overall scores in 3D). But overall 3D scores are going up dramatically.

This means one of two things...either the graphics subsystem (chipset and vid cards) are actually faster (they're not--in fact, they're SLOWER than on RD580), or that the driver overhead that plagues CrossFire (and SLI), is being completely manhandled by Conroe, allowing a lot more of the CPU time to be used for the benchmark.

Maybe the driver's 'work' is handled within the cache, requiring little memory access? That would require a lot MORE memory access for the actual benchmark though, so I guess there goes that theory :-/

EDIT: in conclusion: Conroe's good

[Carfax]

i can guarantee you that the AM3 benchmark does not fit into the Conroe's 4MB L2 cache. stop trying to rationalize this with unfounded theories. why is it so hard for people to accept the power of Conroe? i getting so sick of posts like this, i'm about ready to stop making this info public and just share it with my friends who i know are intelligent enough to look at it in the right light....

I can rationalize anything I want.. I'm not forcing or peddling my theories to anyone, so stop telling me what to do..

Anyway, what does my theory have to do with NOT accepting the power of Conroe?

All the cache does is make sure the execution units are fed at a very fast rate..

I also believe I stated that Conroe rapes the K8 in anything INT related, and it just so happens that games tend to use INT code more than anything else.

Conroe has 3 64-bit ALUS, and 3 SSE2 units solely devoted to INT.. Thats alot of firepower..

Again, how am I in any way not accepting the awesomeness of this processor?

Edit: in fact, someone load AM3 into memory and tell me the size....I'm going to say somwhere in the range of 16-18MB is resident at any given time. then someone look at the size of the program on disk...I don't know...100MB? 4MB of memory is a lot for a CPUs cache, but remember, it's just that....4 MB....not a lot in actuality.

The actual EXE is going to be much smaller I'd think.. It's similar to the Quake III benchmarks.

[Harshal]

FCG, dont let thing like dat get to you! Congrats, also you are doing some nice work to let people know how Conroe is going to be

[RyderOCZ]

Sorry you're wrong there dude. The X1900 core has three times the amount of shader units as the X1800 does..

They have the same amount of pipelines though..

I sure crossed that up somewhere....oh well.

[Carfax]

Maybe the driver's 'work' is handled within the cache, requiring little memory access? That would require a lot MORE memory access for the actual benchmark though, so I guess there goes that theory :-/

EDIT: in conclusion: Conroe's good

Graphics drivers use alot of SSE2 instructions, and Conroe is an SSE2 monster, which could explain why it's so fast in gaming..

This was theorized at one point over at Aces.

[SoulGG]

i can guarantee you that the AM3 benchmark does not fit into the Conroe's 4MB L2 cache. stop trying to rationalize this with unfounded theories. why is it so hard for people to accept the power of Conroe? i getting so sick of posts like this, i'm about ready to stop making this info public and just share it with my friends who i know are intelligent enough to look at it in the right light....

Edit: in fact, someone load AM3 into memory and tell me the size....I'm going to say somwhere in the range of 16-18MB is resident at any given time. then someone look at the size of the program on disk...I don't know...100MB? 4MB of memory is a lot for a CPU's cache, but remember, it's just that....4 MB....not a lot in actuality. bottom line, knock off the "fitting in cache" statements. enough of this crap.

Bleh don't penalize everyone for the few bad eggs that doubt the power, Everyone will always have bad or good opinions about something. I think some are entitled to it but others just act ignorant. Keep us posted man I can't speak for everyone but i'd like to know what it's capable of before buying it

[iterations]

i can guarantee you that the AM3 benchmark does not fit into the Conroe's 4MB L2 cache. stop trying to rationalize this with unfounded theories. why is it so hard for people to accept the power of Conroe? i getting so sick of posts like this, i'm about ready to stop making this info public and just share it with my friends who i know are intelligent enough to look at it in the right light....

Edit: in fact, someone load AM3 into memory and tell me the size....I'm going to say somwhere in the range of 16-18MB is resident at any given time. then someone look at the size of the program on disk...I don't know...100MB? 4MB of memory is a lot for a CPU's cache, but remember, it's just that....4 MB....not a lot in actuality. bottom line, knock off the "fitting in cache" statements. enough of this crap.

Don't stress over chumps. There are plenty of technology enthusiasts out there who greatly appreciate the access to info that you are providing. Keep up the good work, and gratz on the WR.

[Carfax]

the 4megs and prolly tight latency is causing this uber result, i believe yonah wouldn't come close to this result..

The cache has something to do with it, but I think it's more that Conroe is such an SSE monster..

Graphics drivers use alot of packed SSE2 instructions, and as we all know, Conroe just whoops ass when it comes to SSEn..

[ex2cib]

a quick score comparison between FCG's conroe and Fugger's 955XE

Fugger's score:


X1900XTX 700/900 CF
955XE @ 5.8ghz
http://www.xtremesystems.org/forums/...01&postcount=1


FCG's score:


X1900XTX CF 718/738
conroe @ 2.72ghz


The Cpu scores are almost identical here,
Fugger 16.6k
FCG 16.8k

but, FCG's cards are slower than Fuggers and but still end up with a higher graphics score, which would probably mean, like its already been said, that conroe can help reduce the bottleneck by using CF.

then for fun, conroe has a higher CPU score than a 955XE at more than double the CPU speed.

[railer]

AM3 will use atleast 80MB of Vmemory. Last time i checked with Rivatuner it was in this range. Ant 4mb" fith into cache theroy" is so frkn wrong only people that have no idea how cpu works can say something like that.

[SoulGG]

I was not addressing you in general. He's not forced to display anything unless he wants to so we should show some gratitude towards him for giving us the inside scoop. All I meant by that statement really.

Edit: You do not want to discourage one that is helping us all gather more information about it than you can shake a stick at

[Rovtar]

congratz from my side

[Carfax]

AM3 will use atleast 80MB of Vmemory. Last time i checked with Rivatuner it was in this range. Ant 4mb" fith into cache theroy" is so frkn wrong only people that have no idea how cpu works can say something like that.

Um, I'm not talking about the entire program .

I'm talking about the EXE..

[iterations]

Um, I'm not talking about the entire program .

I'm talking about the EXE..

I doubt you have any clue how a CPU works if you think the size of the EXE matters in the least when it comes to L2 cache usage in this benchmark.

[VulgarHandle]

it's nice to see conroe performing the same/better than an older intel at half the speed, really, awesome...

but the claims of 4ghz on air are still unfounded

don't get me wrong though, i'd LOVE to see 4ghz on air, hell, i'd LOVE to see 4ghz on phase

FCG, FTW!!! get the phase cookin!!!