SkullTrail anyone???

[barneybear]

There is very little interest in Sandra CPU Synthetic Benchmarks, the results can be easily predicted beforehand.

To you. most synth benchmarks can be predicted beforehand, however its always nice seeing them. or so i thought. but I can say the cinebench results posted earlier in this thread on the V8 are absolutely slaughtered.

The most interesting benchmark is memory bandwidth test, specifically multi-threaded memory benchmark.

True, I could show them, would you like to predict the results beforehand?

[Cronos]

True, I could show them, would you like to predict the results beforehand?

This is not predictable. All I can say it won't be anywhere near it should be.
I can very roughly guess ~8500 for dual-thread read bandwidth and ~5000 for single-thread read.

Sandra CPU synthetic benchmarks results are equal to (Total Frequency) x (Some constant factor ). If you know the factor for this specific CPU core, you can predict any benchmark result.

[barneybear]

This is not predictable. All I can say it won't be anywhere near it should be.
I can very roughly guess ~8500 for dual-thread read bandwidth and ~5000 for single-thread read.

Sandra CPU synthetic benchmarks results are equal to (Total Frequency) x (Some constant factor ). If you know the factor for this specific CPU core, you can predict any benchmark result.

ok, so based on your logic, can you tell me the speed the skulltrail was running in the sandra picture I posted, should be easy enough.

[y2kbugger]

Skulltrail on H2o...



No not my, but I though I'd share this with you all. Found it at xbitlabs.com.

walang ano man

that is the sickest setup i have ever dreamed of having!

[Cronos]

ok, so based on your logic, can you tell me the speed the skulltrail was running in the sandra picture I posted, should be easy enough.

~ 4 GHz

[barneybear]

~ 4 GHz

spot on 10x400 (1.3volts)

[Cronos]

spot on 10x400 (1.3volts)

Don't forget to test the rig with Linpack 64bit. 32bit results are of little interest as well.

Theoretical x8 (64bit double) floating point peak performance of Core2 is equal to 4x Speed (in Gigaflops). For 8 cores at 4GHz, we can estimate it as 4 x 4GHz * 8Cores =128Gigaflops.
See how much of this theoretical peak you can get. I was able to get ~88% on my Q6600@OCed. This can only be achieved with 64bit Linpack, 32bit produces much lower results.

[barneybear]

unfortunately with the alpha evaluation bios I have, the core voltage increases are extremely tempremental and in fact only the overvolt seems to work properly but it increases the volts to almost 1.5v (around 1.48 from 1.25), and as im using zalman coolers im a little unsure as to whether that is safe. I did boot at 4.2ghz with that voltage however the temperatures went over 70c under full load.

bit risky I think, would need watercooling. Will speak to intel this week about an update bios to fix the pesky vcore issues ive been experiencing.

[Cronos]

barneybear
Can you at least show Sandra and Everest memory bandwidth ?
Better yet Right Mark Multi-Threaded Memory bench....

[barneybear]

barneybear
Can you at least show Sandra and Everest memory bandwidth ?
Better yet Right Mark Multi-Threaded Memory bench....

sure. Right mark multi threaded memory test.

On the Intel V8 system (2x5365's @ 3.0ghz, with 16GB crucial fully buffered 667mhz ram 8x2GB)

each cpu averages around 20100.0

total average around 157000.0mb/s

Skulltrail system (X9775x2 sitting at 3.6ghz atm stock voltage) [/B] 2x2gb FB's 800mhz (should get another 2 sticks from kingston tomorrow)

each cpu averages around 23500.0

total average around 190000.0mb/s

Anything else?

[Cronos]

Anything else?

What you have measured is cache read speed

Set at least 50MB size for each test thread. Set Read/ With Prefetch and Prefetch distance =256 for each thread. Repeat the test for 1,2,4,8 threads.

By the way, try to determine which core each thread is assigned to. All cores belonging to the same CPU package share the same FSB and are limited to ~1/2 of total bandwidth. At least, for 5000X/P chipsets.

[Cronos]

Just noticed you have only 2 sticks, this is not interesting at all
The memory controller has 4-channels, so it requires 4 DIMMs to operate at full speed.

[barneybear]

here is a cinebench at 3.6ghz

[barneybear]

Just noticed you have only 2 sticks, this is not interesting at all
The memory controller has 4-channels, so it requires 4 DIMMs to operate at full speed.

yeah, I know im waiting on 2 more sticks to come, sorry !

[LaMpiR]

ing amazing pic! It looks awsome dude!!!

[Don_Dan]

yeah, I know im waiting on 2 more sticks to come, sorry !

Could you please run Everest memory benchmarks ( now and when the 2 other sticks have arrived ). Are you allowed to share it?

[DarthBeavis]

Will be building it next week . . .have the monitors setup for it tho
TripleHead2Go is working . . .not all that I dreamed of but it is good enough:

I think it is something like 3072 x 768 . . .limited by the resolutions of the Westys.

[Scooby]

That is AWESOME DarthBeavis, I thought the TripleHead2Go was designed for three regular LCDs (1280x1024) and not wide screens, but I see you have it going.

Sorry for the off topic guys, hey DarthBeavis you mind if I PM you for more info?

[LaMpiR]

This is awsome
What gpu power should you have to make crysis go smooth in this?

[moloko]

I don't generally post too much on XS or on forums in general but after seeing this thread, I figured I'd share some pics with you all.

First up, the initial setup of the platform on air:



Some quick tests around the 4GHz mark - max benching on air in the humid and hot conditions was around 4.4GHz:





Next was setup for sub-zero:



The messy testing area - excuse the mess, I've been crazy busy:



Two pots on the pair of CPUs. Pots used were Kayl rev 2.0 and 3.0 pots:







With some dry ice in the pot and acetone added to the mix the end result was a 5.33GHz frequency to have some fun with.

wPrime 32M world record:



Unfortunately after completing the 1024M run (also WR) the HDD spat the dummy and corrupted so the 1024M result was lost and didn't get to do a PCMark run Oh well, always another weekend for that

I'd like to thank:

Ms. Intel Australia for the support
pro for being a fun bench buddy: hip-hop + benching + beers = win

This weekend hopefully get another session organised with a few of the Team.AU boys in Sydney. With any luck we'll get some spare time to throw this setup under LN2 but no promises - soz.

Cheers peeps and I hope you all enjoy


Team.AU

[barneybear]

what kind of voltage settings are you using? im on the newest bios from intel and no matter what core voltage setting I use it actually wont increase it. I see you are still on 684 bios. you should ask intel for the restricted login info for the new one.

regardless of the setting you alter in the performance/processor override, the processors always remain at 1.25 core volts (checked by looking in hardware monitoring within the advanced tab). The overvolt setting works (CPU0 and CPU1 voltage offset) however it pushes the core voltage up from 1.25v to 1.48v which seems rather high to me? There is a section called "reference voltage overide" and they are all defaulted to 67. have you any info on this?

[DaMulta]

I sooo wonder how this will do with cheaper CPU's installed with it.

I might be able to come up with the funds for the board and two of the lower in CPU's, but no way will I be able to come up with the funds for the two unlocked chips at 1grand and up.

[barneybear]

i think the estimated retail is around 1300-1500 each.

[DaMulta]

i think the estimated retail is around 1300-1500 each.

So about a 4 grand setup, and 7 grand from dell(if dell even will have this).

[Dualist]

So about a 4 grand setup, and 7 grand from dell(if dell even will have this).

If someone will pay it, someone will supply it.