How Many Cores?!?!?!

[s7e9h3n]

Not bad for 2Ghz and default Vcore

@ Chew - Why 1.55? I should be able to take down the AMD 1024WR next run

[chew*]

[.jpg[/IMG]

Not bad for 2Ghz and default Vcore

@ Chew - Why 1.55? I should be able to take down the AMD 1024WR next run

becasue 1.55 is the only version recognized by hwbot.

32m A ways to go for Ripping though. Bet it goes down at 2.3-2.4g

wP32M 04.304sec Pheelon AMD Opteron 8382 2612.1 1.20 Air

[stangracin3]

dear god that thing is a beast

[64NOMIS]

64NOMIS says this comes out from the matrix.

Hmph.

With this, who needs the matrix? You can just create a local server and run your own locally!

[Gautam]

becasue 1.55 is the only version recognized by hwbot.

32m A ways to go for Ripping though. Bet it goes down at 2.3-2.4g

wP32M 04.304sec Pheelon AMD Opteron 8382 2612.1 1.20 Air

And not only that, it scores better too.

You're going to be in the top 5 worldwide for sure in 1024M once you run it. Is this enough to bring out the dual-head autocascade or what?

[justapost]

Thanks for the clue

I found the register and am in the process of editing all 24-cores - this is turning out to be quite an ordeal I need a tool which will script the process.

I'm only changing them to 2Ghz for now just so I can make sure it's stable before pushing this thing hard....

You can use variables

>Local1=0xc0010070
>Local2=[value]
>cpu 1
>WRMSR Local1 0 Local2

Or you can read the value from the first core and copy it to all others

>cpu 1
>Local1=0xc0010070
>RDMSR Local1
>cpu 2
>WRMSR Local1 LocalB LocalA

RDMSR <register#>
Read CPU MSR register, LocalA, B, C, D=EAX, EBX, ECX, EDX

WRMSR <register#> <EDX> <EAX>
Write CPU MSR register

If CnQ is active it's better to use the p-state-0 register.

[Mechromancer]

This K10Stat program is pretty cool :

@ DEFAULT Vcore:

So you are not able to change the Vcore with K10stat? That makes me very sad .

[chew*]

So you are not able to change the Vcore with K10stat? That makes me very sad .

Why would it make you sad?

This is beta hardware running with software that wasn't designed for it....we are lucky running the program alone doesn't crash the rig......

Servers are always packed with overclocking features

[s7e9h3n]

becasue 1.55 is the only version recognized by hwbot.

32m A ways to go for Ripping though. Bet it goes down at 2.3-2.4g

wP32M 04.304sec Pheelon AMD Opteron 8382 2612.1 1.20 Air

Haven't yet got 1.55, but here we go @ 2.6Ghz

[justapost]

Can you please run stream once again at 2.6GHz?

[chew*]

LOL rocked it congrats on WR man sick.

Bet my estimation was close.

What a comeback thread you knocked this one out of the park

Here's the times he beat guys , no longer current Wprime Wr's

wP1GB 01min55.988sec Pheelon AMD Opteron 8382 2612.0 1.20 Air
wP32M 04.304sec Pheelon AMD Opteron 8382 2612.1 1.20 Air

[s7e9h3n]

You can use variables

>Local1=0xc0010070
>Local2=[value]
>cpu 1
>WRMSR Local1 0 Local2

Or you can read the value from the first core and copy it to all others

>cpu 1
>Local1=0xc0010070
>RDMSR Local1
>cpu 2
>WRMSR Local1 LocalB LocalA

RDMSR <register#>
Read CPU MSR register, LocalA, B, C, D=EAX, EBX, ECX, EDX

WRMSR <register#> <EDX> <EAX>
Write CPU MSR register

If CnQ is active it's better to use the p-state-0 register.

OK...what program are you using to change multiple registers @ a time. I've always used CrystalCPUID and did it with the MSR editor, but that obviously isn't working out too well now that there are 24 cores. Also, do you have any suggestions where to find the VID register? It's unclear to me which register K10Stat is writing to.

[Trunks]

[RE4 Merchant]

I'll Buy it at a High Price!

[/RE4 Merchant]

[s7e9h3n]

Can you please run stream once again at 2.6GHz?

-------------------------------------------------------------
Function Rate (MB/s) Avg time Min time Max time
Copy: 47981.1671 0.0134 0.0133 0.0134
Scale: 47920.4232 0.0134 0.0134 0.0136
Add: 45246.5847 0.0212 0.0212 0.0213
Triad: 45260.7544 0.0213 0.0212 0.0214
-------------------------------------------------------------

[justapost]

OK...what program are you using to change multiple registers @ a time. I've always used CrystalCPUID and did it with the MSR editor, but that obviously isn't working out too well now that there are 24 cores. Also, do you have any suggestions where to find the VID register? It's unclear to me which register K10Stat is writing to.

All this code are scripts for rw-everything
http://jacky5488.myweb.hinet.net/

P-State Registers are described here starting at p 389
http://support.amd.com/de/Processor_...28_5-28-09.pdf

0xc0010062 is the current p-state and hold all info for the FID/DID and VID of the CPU and NB. For the NB only the VID can be changed however.

Update: Wring register it's 0xc0010070 not ...62
Can be the VID is not unlocked there is an msr register for CPUMAXVID somewhere gotta check the BKDG. If there is an MAXVID defined you can not go above that voltage via vid's only via voltage offsets on the pwm controller chip.

The control register 0xc0010063 can be used to select an specific p-state. In that case the values of this p-state (0xc0010064 for p-state-0) gets copied in the p-state control register. The same happens if CnQ switches p-states.

Easiest should be you use k10 stats to change the control and p-state-0 register and use an rw everything script to copy those values from cpu1 to cpu2-cpu24.

Thanks for the stream run, that looks more DDR3 1333 alike.

In case you forgott here's info how to change the NB FID.
http://www.xtremesystems.org/forums/...226283&page=14
Must be changed on D24-D27 for dual magny-cours I assume.

[Gautam]

Haven't yet got 1.55, but here we go @ 2.6Ghz

Hey, I see photoshopping in the CPU-Z. That calls for a dq and a ban.

[G H Z]

s7e9h3n you always have your hands on something interesting and this rig looks like a real kick to play with. Wanna see how far you can push this...just don't blow up the work server

[EnJoY]

This is getting really exciting. I'm very impressed. Can't wait to see more.

[s7e9h3n]

This is getting really exciting. I'm very impressed. Can't wait to see more.

Don't know how much more I can do....I have to travel to Europe for work for the next week and a half. After I get back, it's on

BTW, benching @ 3ghz @ 1.1Vcore crashes the system. I'll give it one more push before I have to leave.

[s7e9h3n]

s7e9h3n you always have your hands on something interesting and this rig looks like a real kick to play with. Wanna see how far you can push this...just don't blow up the work server

No worries...got more than one

[Brother Esau]

s7e9h3n....been a while since I have seen you last

Whats the ETA on AMD / ATI new Server & Workstation Motherboards? I have been patiently waiting for like 2 years now

[s7e9h3n]

s7e9h3n....been a while since I have seen you last

Whats the ETA on AMD / ATI new Server & Workstation Motherboards? I have been patiently waiting for like 2 years now

I've been hearing Q1 2010



This is my last run for the next week and a half. You guys will have to sit tight until then...unless, of course, someone else decides to show their dual duodecuple system I'm almost positive the cpu's are throttling themselves due to the heat. The numbers - at least the 1024m - doesn't seem to line up with the previous benches:

[TehPwn]

One word WOW! This is insane, it looks like AMD is winning in the gpu area with the 5870 series and now 12 core cpus! Haha im happy for AMD looks like they pwn all now :P I knew AMD would be in the lead again one day lol

[Utroz]

Thats so sick!!!!

[Chumbucket843]

how much ppd will this sucker get?