939 vs AM2 :: SuperPi 32M / 1M @ 3GHz personal best showdown...

[3oh6]

recent conversations in various threads here at XS has brought up the fact that AM2 is slow. if you have graduated from 939 to AM2 then you know what i am talking about...AM2 is slow. that isn't fair, it can be faster but requires primo gear and the gains are marginable.

that said, AM2 is only marginably slower than 939. i finally have a few AM2 motherboards, CPUs, and a pair of memory kits through the workbench and so i figured i was armed with enough knowledge of my newer AM2 gear and take a legitimate run at my personal best 1M and 32M Super Pi times at 300*10=3GHz. here is how it shook out after a few hours messing with the AM2 rig:

Systems: 300*10 = 3000MHz for both setups
Socket:  AM2                    |  939
MB:      Asus M2N32-SLI         |  DFI LP Ultra-D
CPU:     A64 3200+ (512KB L2)   |  Opteron 146 (1MB L2)
RAM:     Corsair 2048-8500 C5   |  OCZ 1024-5000 Plat. (TCCD/5)
GPU:     HIS X800GTO            |  HIS X800GTO
OS:      Windows XP SP2         |  Windows XP SP2
PSU:     OCZ GamerXStream 700w  |  SilverStone ST56ZF

Obviously 939 has a slight advantage in the two setups above, depending on opinion of course, due to the larger cache. Other than 300*10, there are no rules except to get a faster time using whatever means necessary.

Socket 939:
1M - 26.984 seconds_______________32M - 23m 51.687
____

Socket AM2:
1M - 26.985 seconds_______________32M - 23m 51.625
____
These numbers are of course a work in progress but I thought that it was interesting that AM2 is nipping at 939s heels after only a month or two of being out. Of course this is only Super Pi and at a restricted processor speed but I like these types of comparisons.

Please post your own Super Pi 1M and 32M runs at 300*10, especially if you are running AM2. I really want to see how others crack into the 26s with AM2. I will be trying to get the OCZ PC7200 plats I have to run 3-3-3 at DDR856 (holy voltage required batman) and see what it spits out for times in a few days.

[Name Change]

The 1mb cach for 939 is prob making it faster since AM2 only has 512 cach. Best if you had a 3200+ at same speeds from each socket. . gj tho.

[bachus_anonym]

Very impressive times, 3oh6

I'm gonna try to dig out some screens for the comparison but it's not going to be 3GHz, though I think 2.9GHz and I will post it later, if I can join... My 3200+ will not pass 32M @ 3GHz with cooling I have on and I got too lazy to figure out AM2 mounting for my single stage. It's being resting for 3 weeks now

But, I remember so far only one person showing off 1M at below 27s @ 3GHz and that was Kinc - 26.907 @ DDR2-1200 5-3-4-4.0 with Corsair's TWIN2X1024-8500. It's still below the best 1M & 3GHz times on S939 but not very far...

[fhpchris]

Very impressive times, 3oh6

I'm gonna try to dig out some screens for the comparison but it's not going to be 3GHz, though I think 2.9GHz and I will post it later, if I can join... My 3200+ will not pass 32M @ 3GHz with cooling I have on and I got too lazy to figure out AM2 mounting for my single stage. It's being resting for 3 weeks now

But, I remember so far only one person showing off 1M at below 27s @ 3GHz and that was Kinc - 26.907 @ DDR2-1200 5-3-4-4.0 with Corsair's TWIN2X1024-8500. It's still below the best 1M & 3GHz times on S939 but not very far...

Thats only single Pi on dual core....

[3oh6]

Thats only single Pi on dual core....

that counts in my books. this isn't about stability, just shooting for 26 seconds at 300*10. its fun because you don't always need crazy cooling just a solid chip and decent cooling. on 939s a decent pair of BH5s, a lan party, and a bench rig next to a window in canadian winters got you in the door. with AM2 its a bit different with the ram required costing a fortune. after that its just a game of tweaking ram and os though and that is damn fun.

thanx for the link BA, i knew it had to be something stoopid like 4-4-3-5 @ DDR1200 and kinc to get it done. having just spent the last 48 hours with the 2048MB version of the same kit as kinc i truly have to bow down to that memory manipulation...crazy. i think i need an FX to find a chip with a decent memory controller, my 3200+ is doing alright but i have found its limits already and something tells me that it won't ever be running 2GBs at DDR1200 4-4-3-5

[nn_step]

you forget to mention that the AM2 latency and timings suck ass. try DDR2-800 3-3-3-8 or 4-4-4-12

[CHaynes112]

Has anyone broken into 26's with 2GB of RAM? I am talking about 3.0GHz and air cooling. I can hit 27's all day long, but I can NOT get this thing to break into 26's I still haven't given up though

[3oh6]

well it's either DDR750 / 1000 / 1200 for options until we get other multipliers (if that is even possible with AM2) so maybe i will see what i can get at 3GHz DDR1000 4-4-3-5. i think i ran DDR750 at 3-3-3-x and it wasn't anything close to 26s but i can't decisively remember. that rig is running something else right now but when its done i will put it back in 1M mode.

i still think that everything i have found thus far with AM2 leans towards higher clocks giving best super pi numbers. it will just be a matter of getting those timings down at DDR1200, like kinc did with lethal amounts of voltage i am sure.

Has anyone broken into 26's with 2GB of RAM? I am talking about 3.0GHz and air cooling. I can hit 27's all day long, but I can NOT get this thing to break into 26's I still haven't given up though

that is basically what i am doing here, trying to at least. 2GB, 3GHz...post up your screenshots. include sys info for ram timings too if possible please.


**edit**
turns out my memory does actually work to some extent, no pun intended or implied. DDR1000 4-4-3-5 only gets 27.7xx and 27.4xx using copy waza. it looks like i am going to have to perform a little soldering work and see if i can't make one of these two kits run DDR1200 at 4-4-3-x which, if stable, would plow through 26s. thanx again for Kinc screenshot BA, a little push is all i needed to kill some sticks this weekend. 2GB 3GHz 26s...here i come, now where is my ice bucket

[fhpchris]

well it's either DDR750 / 1000 / 1200 for options until we get other multipliers (if that is even possible with AM2) so maybe i will see what i can get at 3GHz DDR1000 4-4-3-5. i think i ran DDR750 at 3-3-3-x and it wasn't anything close to 26s but i can't decisively remember. that rig is running something else right now but when its done i will put it back in 1M mode.

i still think that everything i have found thus far with AM2 leans towards higher clocks giving best super pi numbers. it will just be a matter of getting those timings down at DDR1200, like kinc did with lethal amounts of voltage i am sure.


that is basically what i am doing here, trying to at least. 2GB, 3GHz...post up your screenshots. include sys info for ram timings too if possible please.

The question now seems to be, "can you run DDR1000-1200 or 1T at -70c?"

We also need to see how well they are scaling, Kinks 1.6v @ 3000 does not look very good. I could swear he was running higher voltage on phase, so he just didnt forget to change it to something lower...

[nutcase]

Oh, I think my AM2 4800+ is doing well

nothing fantastic yet.. Too bad the chip is Badly cold bugged.

[3oh6]

The question now seems to be, "can you run DDR1000-1200 or 1T at -70c?"

We also need to see how well they are scaling, Kinks 1.6v @ 3000 does not look very good. I could swear he was running higher voltage on phase, so he just didnt forget to change it to something lower...

hahaha, DDR1200 at -C just doesn't strike me as something that will be a common occurance, we'll have to see. this 3200+ i am running in my tests runs 2.95GHz with 1.57v but takes 1.65v to stabalize 3GHz (prime/3d/memtest/superpi). i desperately need to get a pot and run some dry ice on it and see how it reacts to the cold. its memory controller seems quite strong thus far.

as for this 26 second goal, consider that accomplished. guess i just had to put another few hours into it...man i am a loser first post updating now...

[Eldonko]

AM2 would be better w/ a 1mb cache. Performance is very close though, nice comparo.

[Dani]

Is AM2 that slow

S939 3GHz SP32M @ 23m39.xxxs.

[3oh6]

Is AM2 that slow

S939 3GHz SP32M @ 23m39.xxxs.

that is a monster run mate and yes it is that slow.

in order to touch 939 times with BH5 i need to get 4-4-x-x stable at DDR1200. something tells me i have sticks that can do it but just not the memory controller just yet. a 5.5 divider for memory would do the trick too. i would also love for a 3GHz 1MB L2 single core AM2 to fall from the sky so i can do a legit comparison because i am sure the cache is playing a small role.

[misteroadster]

nice compare, interresting.

I think AM2 with 2x1Mo L2 should do better with about 2 or 3 1/10 sec as always with A64.

[Zebbo]

Is AM2 that slow

S939 3GHz SP32M @ 23m39.xxxs.

No its not, you just have "too many" tweaks in use there

AM2 4800+ @ 3GHz paired with Mushkin Redline XP2-8000 = 23m 39.031s

[Eldonko]

Nice Mr. Zebbo

[john3]

were u comparing an opteron to an athlon 64? no wonder the s939 was at am2s heels. i wonder how the x2 3800+ would do against its am2 counterpart.

[Absolute_0]

Much closer than i expected, especially for something as memory-dependent as SP.

[3oh6]

AM2 4800+ @ 3GHz paired with Mushkin Redline XP2-8000 = 23m 39.031s

hot damn, that is a nice run my friend...win 2k?

Much closer than i expected, especially for something as memory-dependent as SP.

i was expecting more responses like this i think. despite being totally different platforms as far as memory goes, i have managed to use basically the same tweaks, same type of OS install, same technique, bangin kits of memory to work with and get virtually identical results.

i am just glad i was able to get 26s with AM2 without a volt mod on a motherboard and the DFIs haven't even dropped yet

[Zebbo]

hot damn, that is a nice run my friend...win 2k?

Win XP Pro (non 64-bit)

[Dani]

No its not, you just have "too many" tweaks in use there

AM2 4800+ @ 3GHz paired with Mushkin Redline XP2-8000 = 23m 39.031s

Yes I know
..but still same performance between these setup

[tictac]

Yes I know
..but still same performance between these setup

Agreed...

[metalazzo]

very nice times
finally with a lot of tweaking, AM2 is not sucking that much

[Obi]

bugger, I just bought AM2...... Can you feel the difference in normal day to day use?

Insane