Super Talent DDR3-1333 Testing

[BenchZowner]

Let the DDR3 adventures begin!

I will be testing the Super Talent DDR3-1333 2GB Kit ( W1333UX2G9 ) this week, but won't publish a review before I get a better BIOS for the P5K3D 'cause I feel there's more juice to it, and some more BIOS tweaking to be done

The sticks that I've received use Infin... erm Qimonda chips, as shown bellow:

First test ( quickie ) at DDR3-1066 trying to find the lowest ( strict ) timings possible ( not found yet )

validation link ( http://valid.x86-secret.com/show_oc?id=207903 )

If you have any request ( e.g. SuperPi with specific CPU Freq / RAM timings... ) , etc feel free to give me a shout

In the search for the lowest timings possible at DDR3-1200

validation - http://valid.x86-secret.com/show_oc?id=207993

SuperPi @ 3.6GHz coming later on

Will test them thoroughly after I finish two reviews that should be done as soon as possible ( estimated sticks testing/review date is this Monday/Tuesday ).

Will run some 'fast' tests with the sticks later, not thorough stuff though, but still some runs ( i'm focusing on finishing the writing of a ddr2 roundup again )

Take 1:

SuperPi 1M:
CPU @ 3.6GHz ( 360MHz FSB , 10x Multi )
RAM @ DDR3-1440 ( 1:2 ) @ 9-8-8-21 @ 1.9V

Validation - http://valid.x86-secret.com/show_oc?id=208038

Take 2 & 3:

System setup:
CPU @ 3600MHz ( 10x Multi , 360MHz FSB )
RAM @ DDR3-1440 ( 1:2 Div ) @ 9-8-8-21 with totally untweaked sub-timings
GF8800GTX Stock

Driver settings for highest performance
24/7 WinXP Setup ( actually a specially scr3wd up edition, used with 3 different mobos )

AquaMark 3 ( no LOD )

3D Mark2001SE:

ORB : http://service.futuremark.com/compare?2k1=9214760
Details:
Test order : Nature, Drag Low, Car High, Lobby High, Lobby Low, Drag High, Car Low
LODs:
CL 3
CH 9
DL 8
DH 5
LL 5
LH 6
N 4.9

Check back later on for more

 

[lopri]

Super Pi 1M @3.60GHz. Can it make it under 14 secs?

 

[lopri]

NM. Didn't realize your links were not correctly written! DDR3 isn't as bad as I thought it to be at this early stage. An advantage from the 1M shot is that you can run 1:2 thanks to the high frequencies. 1:2 should be (at least in theory, and probably in practice too) faster than 4:5, 3:4, 3:5, etc.

Reference point:

975x: 9x400 / DDR2-800 (3-3-3) --> 14.0xx secs (P5W-DH default other than the big 4s)
680i: 9x400 / DDR2-800 (3-3-3-1T) --> 14.0xx (EVGA 680i default other than the big 4s)
P965: 8x450 / DDR2-900 (4-4-4) --> 13.9xx (DFI P965-S default other than 'Fast' setting in the BIOS. But this board is very slow in 32M for whatever reason)

 

[lopri]

Oh. Does Everest recognize DDR3? If so, can you run quick Cache & Latency test?

 

[BenchZowner]

More tests, and your requests will be fulfilled in a matter of hours

 

[BenchZowner]

Another update

Finding the tightest ( best ) timings for this memory kit, because I'll need them for the P5K3 Deluxe review this weekend

Here's a screenie with various benchmarks ( SuperPi v1.5 XS 1M Calculation, WinRAR v3,70 Beta 7 Benchmark, Everest Ultimate v400 Memory Read )

System Configuration:
Asus P5K3 Deluxe
Intel Core 2 Extreme X6800 @ 3600MHz ( 400MHz FSB, 9x Multiplier )
Super Talent DDR3-1333 @ 8-6-5-12 @ 1.90V

http://www.benchzone.com/Bench...sb-3600mhz-1333ram.jpg

CPUz Validation - http://valid.x86-secret.com/show_oc?id=208468

Oh, and here's your request ( run at the previously mentioned system configuration [ in this post ] )

http://i7.tinypic.com/61le78m.png

 

[BenchZowner]

Another addition
Untweaked SuperPi 32M with best possible timings ( BIOS timings only tweaked ) [ totally wrecked everyday OS ]

http://i17.tinypic.com/4yp361u.jpg

 

[lopri]

The Everest Cache-Latency shot shows very similar values to P965 platform. (975X and 680i show different patterns) I should say the 32M number is extremely impressive considering that DDR3 has just been introduced to the mass market. That is a way better number than I expected to be, regardless of tweaks.

Said that, it's still not enough to replace DDR2 - more than anything due to the prices. Are your sticks retail-purchased, or engineering samples? Also curious under what FSP strap you're running 9x400? Assuming 1066 strap (I am, that is), how many dividers are available? I would think running tightest possible @1:2 will give you the best results due to less overhead. Can you test different dividers with tightest configurations? I know it's going to be a lot of work then maybe 1:2 vs something else.

 

[BenchZowner]

Will do everything you asked for, and even more, after I finish the review of the P5K3 Deluxe mobo which I have to have out by Monday
After that, I will be reviewing this memory kit, so every limit will be explored

 

[BenchZowner]

Same config ( only forgot to re-adjust 1 sub-timing ) fresh WinXP Pro x86, only copy-waza tweak

http://i8.tinypic.com/5z5n6lx.jpg

Will be back for more later on