21s SuperPi 4.8Ghz 323 1:1

[Charles Wirth]

Intel P4 3.8Ghz ES
DFI Lanparty 925XE-T2
OCZ 600ADJ
OCZ DDR2 4200 3-2-2-8
ATI Radeon x800xt
Chilly cascade -124c
Nanotherm PCM
Prescott patched

http://www.xtremesystems.org/forums/attachment.php?attachmentid=24416

I also ran unpatched.
http://www.xtremesystems.org/forums/attachment.php?attachmentid=24417

Ooops, fixed title too.

[Kunaak]

red X for me :(

[CTKP]

red X for me :(
21 sec patched
23 sec unpatched ;)

[L0$t Pr0PhEt]

Very nice! You think you are able to do 22s?

[bias_hjorth]

Very VERY impressive :slobber:
Does it have anymore in it?

[Charles Wirth]

Yea, working on it now.

[Crow]

WTG Fugger...that is looking sweet :D

[enzoR]

hes got till 1.9vcore :D

[bias_hjorth]

Man 22 is just around the corner. My god that board is sweet :up:

[olfen]

that dfi board is really great

[eva2000]

very nice ram too :D

[agenda2005]

VERY NICE RESULTS 323MHz 1:1! Looks impressive for DDR2. Can you post some other benchmarks like pifast and show everest read/write and latency?

[k|ngp|n]

Awsome job Fugger :toast:

22sec unpatched would be sick!

[charlie]

awesome job!!

[The Byter]

the future is yours :)

we want 19sec

[Bulldog20]

Nice going Fugger :)

You finding the DFI better?, what was the max fsb on your Abit?

Also I remember afew of you saying that the FSBVTT would not go very high without errors. Well I change my 570j and now the board runs perfect upto 1.65, not tryed any more. Is there the same problem on the DFI or don't they offer FSBVTT.
Funny thing is the chip is the same date and Batch number and clocks exactly the same :rolleyes:
Cheers

[blander]

very nice :toast:
wr unpatched spi is 22s, right?

[Kanavit]

that is a sick sick score, nice job Fugger! wow, 21 sec. is that a world record?

[Ref]

No, there's not patched PI competition, but the best is FUGGERS 20 sec, but he doesn't have a screenie.
REMEMBER it's the unpatched score that counts.
:slobber:

[D_o_S]

Awesome :slobber:

[bachus_anonym]

those new toys are performing very well I see....
I know that you're only going for SuperPi 1M and PIfast for now, but I'll be waiting for some 3D or maybe at least 8MB SuperPi run....

Keep pushing it, Charles :thumbsup:

BTW, change thread title... I almost "not-clicked" it when I saw 4.8GHz :D

[uclajd]

REMEMBER it's the unpatched score that counts.If you have an AMD that is. Don't even get me started on this. Why can't both Intel and AMD people run the "patched" version?

Funny how a version which ignores an entire aspect of the Prescott's arcitechture "counts," but the version which actually uses SSE3 doesn't "count."

Seems to me there should be versions for each processor which take full advantage of their respective designs. Spitting out pi numbers the fastest should be the goal, not using a version which cripples aspects of one of the competitors.

This is why I don't do pi.

Nice job Fugger.

[Charles Wirth]

I asked Futuremark to make a pi bench. I let them know of our current problems with the patches.

They felt no need and seen no advantage to adding another benchamrk to the arsenal. Should we start a petition for FM to listen to our plee?

[Charles Wirth]

a few needed changes

new keyboard, less typo's

added 8k btu ac

added 7k rpm fan to nb

The maximum mhz capable with the 600adj is 6ghz for running memtest at 1.65v. I could boot at 1.7v but when I went for 6Ghz the power supply would shut down. I tried a PCI video card and still could not get stable enough in windows to complete a pi.

I swapped psu's to the Enermax 651w

The mobo will memtest @ DDR700 (350FSB) 3-2-2-8 1:1 2.3v

Would you guys like to see 5.6Ghz memtest at that speed?

[Charles Wirth]

I would like to see an A64 get close that

OCZ DDR2 PC4200 EB Plat Rev2 @ DDR700 3-2-2-8 1:1

http://www.xtremesystems.org/forums/attachment.php?attachmentid=24482

[reject]

holy :banana::banana::banana::banana: fugger, you own

[hipro5]

He,he,he.......4500+ unbuffered memory.....Keep on....INTEL rocks(as always did)....... :D :toast:

[kristos]

Chilly cascade -124c
Nanotherm PCMI had been told that nanotherm wasn't good for phase change rigs because it starts acting like an insulator at negative temps :confused:

I take it that was bogus then? :)



btw, have you got a screenie with Process Timer and superpi unpatched? just to know how close you are to that 22s :D

[Charles Wirth]

I am really liking the Nanotherm now, I had perfect core contact. evident in the lack of thermal material present when I seperated the block and cpu. I can dismantle immediatly after shutting down. Seperation has never been easier.

Shamless Nanotherm plug.

[agenda2005]

I would like to see an A64 get close that

OCZ DDR2 PC4200 EB Plat Rev2 @ DDR700 3-2-2-8 1:1

http://www.xtremesystems.org/forums/attachment.php?attachmentid=24482

A64 has been showing those speed even with DDR1, but not with those kind of timing. A64 got very close here with 1:1 timing
http://xtremesystems.org/forums/showthread.php?t=53051.
but with loose timing 2.5-4-4-
DDR 720 here http://xtremesystems.org/forums/showthread.php?t=48626
but with similar loose timing 3-5-5-

[hipro5]

A64 has been showing those speed even with DDR1, but not with those kind of timing. A64 got very close here with 1:1 timing
http://xtremesystems.org/forums/showthread.php?t=53051.
but with loose timing 2.5-4-4-
DDR 720 here http://xtremesystems.org/forums/showthread.php?t=48626
but with similar loose timing 3-4-4-

What about memtest unbuffered MB/s.......?...... ;)

[enzoR]

39 sec. memtest isnt stable i dont think...

[agenda2005]

What about memtest unbuffered MB/s.......?...... ;)

Well, I will answer your question first and then give my opinion about what I observed here.
The memtest unbuffered uses computation that occurs in a short burst and therefore ignores the weakness of the overall system architecture.
Here is the reason why you see the difference between P4 and A64 results.

P4 uses inclusive L2 cache arcitecture; while A64 uses an exclusive L2 cache.

Inclusive cache is faster and have lower latency but suffers alot of limitations.
Exclusive cache is slower , but more flexible to increase.

The consequences is that L1 cache in P4 has to be really small 8KB or 16KB to be effective.
L2 cache in P4 can not be extremely large otherwise diminishing returns will set in. This is why intel cannot increase the P4 L1 cache arbitrarility. I wil leave details and not go into engineering details.

For A64, with exclusive L2, you can have a bigger inclusive and fast L1 cache(128KB, 64KB data, 64KB instruction) and L2 cache can be as big as it can if you can get good yield and if price doesn't bother you. The consequences is slower L2, but the L2 latency can be masked with a victim buffer between L1 and L2 caches. Memtest doesn't see the victim buffer and therefore gives a wrong impression of the overall system performance.
If you want a good benchmark for system memory, there is nothing that rivals Sisoft Sandra as it uses at least 128MB-512MB depending on your system memory.

Don't be carried away with those numbers you see on memtest if you can back it up with real world performance.

As for my observation. I used Intel and AMD systems (Intel for mobile and AMD for workstation). Some people like the megaHURTZ from Intel P4's but that's fine. As for me I will not get stuck with a brand name except if I work for them, or they pay me some money to do so.

[althes]

great job.

[hipro5]

Well, I will answer your question first and then give my opinion about what I observed here.
The memtest unbuffered uses computation that occurs in a short burst and therefore ignores the weakness of the overall system architecture.
Here is the reason why you see the difference between P4 and A64 results.

P4 uses inclusive L2 cache arcitecture; while A64 uses an exclusive L2 cache.

Inclusive cache is faster and have lower latency but suffers alot of limitations.
Exclusive cache is slower , but more flexible to increase.

The consequences is that L1 cache in P4 has to be really small 8KB or 16KB to be effective.
L2 cache in P4 can not be extremely large otherwise diminishing returns will set in. This is why intel cannot increase the P4 L1 cache arbitrarility. I wil leave details and not go into engineering details.

For A64, with exclusive L2, you can have a bigger inclusive and fast L1 cache(128KB, 64KB data, 64KB instruction) and L2 cache can be as big as it can if you can get good yield and if price doesn't bother you. The consequences is slower L2, but the L2 latency can be masked with a victim buffer between L1 and L2 caches. Memtest doesn't see the victim buffer and therefore gives a wrong impression of the overall system performance.
If you want a good benchmark for system memory, there is nothing that rivals Sisoft Sandra as it uses at least 128MB-512MB depending on your system memory.

Don't be carried away with those numbers you see on memtest if you can back it up with real world performance.

As for my observation. I used Intel and AMD systems (Intel for mobile and AMD for workstation). Some people like the megaHURTZ from Intel P4's but that's fine. As for me I will not get stuck with a brand name except if I work for them, or they pay me some money to do so.

I know exactly what you meen............Don't worry........I just wanted to see an unbuffered compared to the BH-6/5 chips......... ;)

[reject]

i thought fugger meant the L1 and L2 bandwidth, it is massive

[macci]

Fugger, Excellent result there w/ DDR2 and fast timings. :toast:

What sort of unbuffered Sandra mem bench can you get w/ those settings?
'the stilt' has been testing the ocz DDR2 too and he managed to hit some nice speeds too (w/ asus mobo):
sandra unbuf @ DDR716 3-2-2 (http://www.akiba-pc.com/OCZ-4200EB/Unbuff_Tweaked.PNG)
(268fsb and 3:4 divider), I'm guessing that the lowish FSB is holding it back a bit and running 1:1 at those speeeds would give even better result.

[Charles Wirth]

Working my way up to 4.75k unbuff

I have a 4.55k posted in the "first look" thread. I been holding back on the unstable stuff. But I will post a little something special just for you macci.

Gimme a few minutes to fire up the hard drive.

[olfen]

4.75k unbuffered, thats good, so at what fsb is that at, still 3-2-2 or?

[Charles Wirth]

The 4.5K was done at 333FSB 1:1 3-2-2-8

But the attached file is better, you can see my desktop junk too. I usually cut it out to save bandwidth and screen space.

[Carnajo]

Really now, you think Far Cry or whatever your game of choice may be is any more enjoyable at 100fps instead of 90fps, please, but I can tell you that multi-tasking is more enjoyable with HT :)

[FallenAngel]

great job as usual, congrats FUGGER

[uclajd]

Fact is A64 is the gamers CPU and no matter how much speed you put into those P4s the architecture is inefficient. You got a CPU that needs to run twice as fast and costs twice as much in order to simply compete with an A64...I thought this was the overclocking/pi forum, not a gaming forum. :rolleyes:

But since you brought it up, this cpu (http://209.35.180.29/emotes/4555primecrop.jpg) cost $220 new, and I hit 7317 (http://service.futuremark.com/compare?3dm05=551977) in 2005. Looking for a $110 Athlon that can beat my scores... :lol:

So when do you test the 6XXs FUGGER?

[Charles Wirth]

Cleaned up and booted him from this section. Unreal