[LIVE]TiN&T0lsty goes 3D, QHEX+REX+GTX285 under LN2

[TiN]

Long time no post from me. Now I will fix this flaw with latest oc worklog session :)

So, TiN and T0lsty decided to put some LN2 into old-school rig.

We will bench next hardware this weekend

INTEL QHEX
ASUS REX 601 bios
2x1GB D9GTR DDR3's
ASUS GeForce GTX 285 TOP card for 3Dmark01/05/Aqua for TiN
Radeon 5870 Reference for 3Dmark01/03 for T0lsty :cool:
Allmigthy Enermax Revolution 1250W PSU

As for our torture equipment, we have next:

Xtremelabs.org chiller (subzero) for pretesting (http://kingpincooling.com/forum/showthread.php?t=490)
WEHR-WOLF's CPU waterblock for chiller
Xtremelabs.org MAGNUM Legends Edition copper LN2 pot
XBOX GPU HSF on NB for pretest, Xtremelabs.org AL LN2 NB pot
Kingpincooling.com TEK9 6.66, great credits to Vince for it! :clap:
Stock GPU's cooling for pretesting
Bundle of high-cfm fans
xDevs.com ES VARM
Fluke 87V DMM
60L of Liquid Nitro

Session dedicated to old but not weak still LGA775 rig, this is one of my last LGA775 benchmarks before major switch to 1366 platform.

Also this is first serious LN2 GPU benchmark, with my latest newest TEK9 6.66 GPU pot :) Time to put some serious stress on it to see what it can do.

Schedule is next

15 ARPIL 2010 - Pretest system

Today I've assembled rig with chilled CPU and stock GPU (285) to see what can it do. Long time ago I tested exact these CPU and motherboard with LN and got 6.5G CPU-Z suicide shot, 6.36 SPI 1M and plenty of 6.0-6.2G 3D benchies with old cards. So I hope this REX and QHEX (e8600es lol) can still rock 6.1+GHz in 3D for me.

So first test is max stable FSB possible with current hardware. memories are OCZ FlexXLC D9GTR's.

I went to BIOS and set:

Ai Overclock Tuner : Manual
Oc From Cpu Levet Up : Auto
Oc From Memory Level Up : Auto
FSB Frequency : 333
CPU ratio sitting : 6

CPU configuration :
--------------------------- :
C1E support : disabled
Cpu TM function : disabled
Max CPUID value limit : disabled
Vanderpool technology : disabled
Execute disable bit : disabled
Core multi-processing : enabled

FSB: 640MHz
* CPU clock skew : Auto
* NB clok skew : Auto
FSB Strap to North Bridge : 333
PCIE Frequency : 119
DRAM Frequency : 5:6
DRAM Command Rate : 2N
DRAM Timing Control : Manual

1ST INFORMATION
CAS# latency : : 7
RAS# to CAS# delay : : 7
RAS# PRE time : 7
RAS# ACT time : 22
RAS# to RAS# delay : AUTO
REF cycle time : AUTO
WRITE recovery time : AUTO
READ to PRE time : AUTO

2ND INFORMATION
READ to WRITE delay(S/D) : AUTO
WRITE to READ delay(S) : AUTO
WRITE to READ delay(D) : AUTO
READ to READ delay(S) : 5
READ to READ delay(D) : AUTO
WRITE to WRITE delay(S) : AUTO
WRITE to WRITE delay(D) : AUTO

3RD INFORMATION
WRITE to PRE delay : AUTO
READ to PRE delay : AUTO
PRE to PRE delay : AUTO
ALL PRE to ACT delay : 7
ALL PRE to REF delay : 7

DRAM Static Read Control : disabled
DRAM Dynamic Write Control : disabled

DRAM Skew Control :
------------------------------ :
DRAM CMD skew on channel A : auto
DRAM CLK skew on DIMM A1 : auto
DRAM CLK skew on DIMM A2 : auto
DRAM CTL skew on DIMM A1 : auto
DRAM CLT skew on DIMM A2 : auto

DRAM CMD skew on channel B : auto
DRAM CLK skew on DIMM B1 : auto
DRAM CLK skew on DIMM B2 : auto
DRAM CTL skew on DIMM B1 : auto
DRAM CTL skew on DIMM B2 : auto

Ai Clock Twister: AUTO
Ai Transaction Booster: AUTO

-- Manual --

EPU II phase control : full phase

CPU Voltage : 1.500v
Load-line calibration : Disable
CPU PLL voltage : 1.6v
FSB termination voltage : 1.52v
CPU GTLVref(0) : -30
CPU GTLVref(1) : -30
CPU GTLVref(2) : -30
CPU GTLVref(3) : -30
NB GTLVref : auto
North bridge voltage : 1.7v
DRAM voltage : 2.08v
NB DDRVref : +50
DDR3 channel A Vref : +50
DDR3 channel B Vref : +50
South bridge 1.5 voltage : 1.58v
South bridge 1.05 voltage : 1.12v
CPU spread spectrum : disabled
PCIE spread spectrum : disabled

Memory Remap Feature: disabled


Temp on CPU was subzero idle and near +25 under load. I've used 6-7 multis, to keep cpu safe while testing motherboard and rams.

Rams done near 1000mhz 7-7-7-2T 2.08V. NB temps was +50C, and 630 mhz fsb easy in 3D tests, and 640mhz work too but not very stable. Sometime it work ok, sometime would not even pass spi32m.

Suicide shot was done on 661MHz FSB

http://hwbot.org/community/submission/988665_tin_reference_clock_rampage_extreme_661.02_mhz

It's all done with positive temperatures on CPU and motherboard.

So I guess with LN2 on northbridge, and LN2 on CPU i would not hit the wall with motherboard overclock. I need to have this C2D 6.2GHz+ in 3D badly to get decent scores.

http://hwbot.org/signature.img?iid=363905&thumb=false&iehack=.jpg

Some screenie for 630mhz test

http://xtremelabs.org/forums/download/file.php?id=381&t=1 (http://xtremelabs.org/forums/download/file.php?id=381&mode=view)

16 ARPIL 2010 - Pretest / modifications for GPUs

This day will go for GTX285 pretests and mad mods.
TiN will do thingys with GTX, and T0lsty will check his Radeon. :peek:

17 APRIL 2010

10.00 AM GMT+2 begin of LN2 session , R5870+CPU+NB all on LN2
LIVE broadcast on livestream channel

15.00 - 18.00 GMT+2 Little pause

18.00 - 23.00 GMT+2 GTX285+CPU+NB all benching with LN2

Results

TBD

Credits

Kingpincooling.com and personally Vince "k|ngp|n" for TEK9 6.66 :prolleyes:
Xtremelabs.org benchmark team for support :o
ASUS for GTX285,REX support :cool:
Intel for QHEX CPU :D
Enermax for allmighty PSU :up:

Ilya

[TiN]

Anybody's here :D

I made and tested today new experimental shyt for benching :D
I didn't managed to catch our fellow Fermbini with untouchables, but I have another weapon lol :)

Bencher meter :D. officially it's my xDevs.com VARM project unit

It's my own designed 4 channel (only 2 used now) measurement unit.

It's dumb simple.

Power LDO +7...+20V into 5VDC
uC core ATMEL ATmega8L in MLF package (that tiny square chip with 32 pads)
Shiftreg for driving LED displays
Two 4-digit kingbright 7-segment LEDs
few NPNs and passives
AD8547 R2R amplifier as buffer on analogue input, to protect GPU/Vmemory from something bad.
Protective curcuits for inputs
LM335 thermal sensor

Featured and designed but not soldered yet: USB connector and USB bridge chip for remote monitoring and PC-controlled stuff

It's connected to videocard via 3-wire cable, simple GND + Vcore + Vmem, just as you connect sense wires for DMM readout. As for power it take 1 usual +12V FAN 3-pin header on motherboard.

I've calibrated device with my beloved Fluke 87V DMM at 2 points technic :)

So in attach photos in action

As for GTX, i'm testing atm various drivers to find best.
Card oced no problem to 750GPU, 1485 memory on 100% stock cooler, w/o any mods (just meter device hooked)

:)

LIVE pretesting

www.livestream.com/xtremelabsorg

:pcool:

[FatAlbert]

nice! :eek:

dark photos of the Bencher meter in attaches :eek:

waiting for a marriage "Bencher meter" and the untouchable !

simple monster for ln2 :D

[Raja@ASUS]

Very nice stuff Tin.

[TiN]

Pretesting GTX285 done

QHEX 600*7 4200MHz
REX aircooled
GTX285 stock cooled 100% stock volts / NO MODS @ 756/1764/1485
WinXP

http://www.xdevs.com/oc/gtx285/drivers.png

Different drivers test. 182.50 best so far AFAIS

http://www.xdevs.com/oc/gtx285/lods.png

Lods testing on 182.50 driver

[pro]

thats how 3d01 testing NEEDS to be done, every point counts mate...

brilliant effort :)

[packet]

182.50 ftw...nice work Tin :)

[TiN]

Sooo...live started :)

http://www.livestream.com/xtremelabsorg

Everybody welcome

Schedule

Intel QHEX LN2
ASUS REX LN2
Radeon 5870 LN2
GeForce GTX 285 LN2
Enermax Revo 1250W
3dmark03/05/01 damage

Hope this day will be not bad for some results...

[FatAlbert]

TiN don't forgot kicks ass with iron on this GTX :D

[TiN]

As we described, me (TiN) and my teammate T0lsty did some runs with old-school Intel E8600ES + ASUS Rampage Extreme.

First day we started at 10AM with 5870 card, 3 pot system (cpu,nb,gpu) and tested until 14.30. This was almost first time in life while we test 5870 under ln2, so just first results was not very bad, and T0lsty got about 50 global hwboints. We used 16L dewar on this results.

http://hwbot.org/signature.img?iid=365125&thumb=false&iehack=.jpg

http://hwbot.org/community/submission/989565_t0lsty_3dmark_2001_radeon_hd_5870_107885_marks

http://hwbot.org/signature.img?iid=365137&thumb=false&iehack=.jpg

http://hwbot.org/community/submission/989555_t0lsty_3dmark_2003_radeon_hd_5870_110946_marks

Then was little break, and I continue benchin alone. Same day evening tested E8600 with nb under ln alone, to keep sure that system will work nice while my GPU LN2 bench. That was done with stock aircooled 8800gs. CPU done spi1-2m runs at 6.2-6.21, suicide shots at 6.35-6.36, and 3d tests run ok at 6.1-6.15. CPU have CB at -125С, and for 6G it need temp range from -95 to -120С.

I finished tests at 1AM, disassembled and heated system and go sleep.

Settings for all runs for REX:


Ai Overclock Tuner : Manual
Oc From Cpu Levet Up : Auto
Oc From Memory Level Up : Auto
FSB Frequency : 600-612
CPU ratio sitting : 10

CPU configuration :
--------------------------- :
C1E support : disabled
Cpu TM function : disabled
Max CPUID value limit : disabled
Vanderpool technology : disabled
Execute disable bit : disabled
Core multi-processing : enabled

FSB: 640MHz
* CPU clock skew : Auto
* NB clok skew : Auto
PCIE Frequency : 119
DRAM Frequency : 5:8
DRAM Command Rate : 2N
DRAM Timing Control : Manual

1ST INFORMATION
CAS# latency : : 7
RAS# to CAS# delay : : 7
RAS# PRE time : 7
RAS# ACT time : 22
RAS# to RAS# delay : AUTO
REF cycle time : AUTO
WRITE recovery time : AUTO
READ to PRE time : AUTO

2ND INFORMATION
READ to WRITE delay(S/D) : AUTO
WRITE to READ delay(S) : AUTO
WRITE to READ delay(D) : AUTO
READ to READ delay(S) : 5
READ to READ delay(D) : AUTO
WRITE to WRITE delay(S) : AUTO
WRITE to WRITE delay(D) : AUTO

3RD INFORMATION
WRITE to PRE delay : AUTO
READ to PRE delay : AUTO
PRE to PRE delay : AUTO
ALL PRE to ACT delay : 7
ALL PRE to REF delay : 7

DRAM Static Read Control : disabled
DRAM Dynamic Write Control : disabled

DRAM Skew Control :
------------------------------ :
DRAM CMD skew on channel A : auto
DRAM CLK skew on DIMM A1 : auto
DRAM CLK skew on DIMM A2 : auto
DRAM CTL skew on DIMM A1 : auto
DRAM CLT skew on DIMM A2 : auto

DRAM CMD skew on channel B : auto
DRAM CLK skew on DIMM B1 : auto
DRAM CLK skew on DIMM B2 : auto
DRAM CTL skew on DIMM B1 : auto
DRAM CTL skew on DIMM B2 : auto

Ai Clock Twister: AUTO
Ai Transaction Booster: AUTO

-- Manual --

EPU II phase control : full phase

CPU Voltage : 1.800v-1.925v
Load-line calibration : Disable
CPU PLL voltage : 1.59v
FSB termination voltage : 1.47v
CPU GTLVref(0) : -30
CPU GTLVref(1) : -30
CPU GTLVref(2) : -30
CPU GTLVref(3) : -30
NB GTLVref : auto
North bridge voltage : 1.7v
DRAM voltage : 2.10v
NB DDRVref : +50
DDR3 channel A Vref : +50
DDR3 channel B Vref : +50
South bridge 1.5 voltage : 1.57v
South bridge 1.05 voltage : 1.12v
CPU spread spectrum : disabled
PCIE spread spectrum : disabled

Memory Remap Feature: disabled


Next day, sunday 18 april, i've begin modding green frog GTX285 for LN. Before it I never experience any GT200-cards oc even on air, if not count brief and junky tests of flawed Gigabyte GTX275 SOC.

http://i44.tinypic.com/2eefpki.jpg

From the very beginning i've traslated all stuff on livestream, and jmke was kind enough to post even news about my crappy modding. :D
http://www.madshrimps.be/vbulletin/f22/xtremelabs-shows-modding-overclocking-skills-live-71504/

Modding was not fast, i needed to do some datasheet messing for Vgpu PWM controller (teh ISL6327).

Was done next mods:

Vgpu 1.125V stock >>>> 1.525V FIXED by new VID code
OCP overral >>>> ~55A/phase
Vmem 2.1V stock >>>> 2.19V+ (with VR)
Added nice caps (tantalum + Chemicon)
I've physically disconnected TEMPFAULT pins from monitoring chip from card PCB to prevent FAULT assertion.
(not sure if that help coldbug out, never tested this, just made in advance)


http://hwbot.org/signature.img?iid=366289&thumb=false&iehack=.jpg

http://hwbot.org/signature.img?iid=366290&thumb=false&iehack=.jpg

Sorry ppl, i dont have much time these weeks, so not describin every mod in details. Be sure, it's not because i want to keep secrets or similar bad stuff, just lack of time. I will describe details later for sure, and try to answer exact quiestions.

Then i insulated card with blutack both sides, all pcb except VRM area, and started mounting Kingpincooling.com Tek9 6.66.

http://hwbot.org/signature.img?iid=366313&thumb=false&iehack=.jpg

As I saw, side for GT200/G80 chips was not flat. It's very wavy, even feelable by finger touch. :( Brushed that with abrasive paper, but not much luck. left as is and keep mounting.

On MB NB was installed alu small pot to keep X48 chipset solid -20...-50 during all tests. This mobo done 661Mhz fsb no problem, so more than enough.

http://hwbot.org/signature.img?iid=366311&thumb=false&iehack=.jpg

Rig started..
Before that i tested card on stock cooler 100% and it done good with clocks 750\1760\1485 gettin 88k in 3d01 with cpu at 600х7.

http://hwbot.org/signature.img?iid=366312&thumb=false&iehack=.jpg

temp -30 on bad tempmeter (real temp 20-50c lower from value), 800mhz gpu...ok
-50 , 800gpu in nature runs, ok

here riva monitoring begin doing crap showing ripple on Core temperature from -55 to +78 and then quadzillion degrees celsius :D

http://hwbot.org/signature.img?iid=366300&thumb=false&iehack=.jpg

Nice, that i disconnected pins TEMPFAULT from PCB. lol. :D Same stuff was on stock Gigabyte GTS 250,and after -55 card dont go 3D at all. Here it goes, so maybe that help to get rid of CB?.

-60, 850 - ok
-70 - 850 - ok
-80 (realtemp i think near -150) - 850 - ok
-100 (real near -170)

wow :) buuugzzz :) everything on image screwed :D:rofl:

Then started to oc itself. booted cpu at 6030mhz for rock stable runs and go 3d01 with gpu800mhz, drams 1485 which was nice on stock air for allday-long runs. got 109k..saved 850mhz.....freeze, black screen on start...reboot...850mhz again.....freeze......ramped temp to -30....800mhz....freeze! ...reboot..830...freezee....10 more freezes-reboots....half of second 16L dewar empty now.....1 hour of trying w/o ANY results AT ALL.ok...

Booted on STOCK cpu, stock gpu , run msi kombustor (furmark)...ah,,triangles droppin.... :parghh:

Mate from OCA on livestream chat suggests to disasm card and check issues with water/ice. ok, doing that.

http://hwbot.org/signature.img?iid=366314&thumb=false&iehack=.jpg

Very little snow on bostix...heating all card to +80C...assemble with towel btw pcb and kflex...
starting...

cpu back to 6GHz
run 3d01....looks ok...-30C ...ok...-50ok,,,freq gpu to 800 - ok....850...freeeze....FUUUUUUU, evil anger here , death rays go from my eyes....:mad::headshake:

time on clock near 22.00 ...second 16L dewar is empty now..taking last 34l one. Flashing bios from EVGA GTX285 SSC rev03. testing...-40 gpu 800 - ok...dram left at 1300 stock...850mhz - nature run ok....900mhz gpu, temp -50...ok...second run at 900...ok...third run on 900...ok...950mhz - black screen just after i press "Run benchmark"....

http://hwbot.org/signature.img?iid=366288&thumb=false&iehack=.jpg

rebooted...run all suite on 830mhz gpu....91k, ok....set gpu clock 911mhz run nature....ok 116k. saved...not a single artifact

Just pressing again "run benchmark"....again black/green/blue screen...wtf

rebooting.....runnin same again
lobby done. drago done...cars....carlo ok...carhi - 20fps from start, 3-4 fps later, 10fps overral test O_O. run nature just after this carhi....1000+fps everythign ok.....again carhi - 9fps...WTF? :shrug:

Killing 3dmark2001 install...runnin fresh setup...it tellin me "no Direct3D device detected. check your resolution and blablabla". and not installing...lol....reinstalled forceware....same....:peek:

booting win2008, second os. installing driver, cpu at 6G , everything ok. runnin 05 mark - 33.8k at 800gpu, ok. Installing 3dmar2001 from it and going back to XP...3dmark working ok now...

Then most interesting part here. p:(

those who oced low-end card know effect when memory OR gpu ocing good, but when try to oc together vmem and gpu - oc is much worse. Thats because of lack of power or similar stuff.

testing this on this GTX285 with two external 6p PCI-E connectors with my main Enermax 1250W PSU. :baffled:

vmemory clock to 900mhz (400mhz below stock 1300). testing gpu...800 - ok....850 - ok, 900 - ok, 950 - ok, 1000 -few runs ok, third run artifacts...

runnin memory to 1500mhz - 670mhz gpu ok, 700 ok, 750 ok, 800 - solid black screens. WTH is that on pre-fermi top card? :pmad:

3am on clock. im totally tired, it's cold on balcony where my benchtable placed, cant think properly because of lack of oxygen and lots of ln2 poured, so rebooting with cpu 6120 to make final runs.

setting gpu clock at near 800, mem lowered and just run.

wow , 118k :D thats best :\


saving that crap and going rest.

So ALL finished results from 10+ hours nonstop pouring are here:


http://kingpincooling.com/forum/attachment.php?attachmentid=248&stc=1&d=1271718746


So that's this sad story. But i did not broken card (but had real anger to do that!) and maybe some continue will occur. Just need to sort out some ukrainian FEDEX crap to receive something veryvery nice for mad experiments.

So for first try i have now:

24th 3d01 http://hwbot.org/community/submission/990561_tin_3dmark_2001_geforce_gtx_285_118322_marks
345th 3d05 http://hwbot.org/community/submission/990566_tin_3dmark_2005_geforce_gtx_285_35781_marks


Credits:

ASUS, Intel, Enermax
Personally cyclone, T0lsty for support :)
Thanks for Kingpincooling.com and personally to Vince "k|ngp|n" Lucido (http://kingpincooling.com/forum/showpost.php?p=7065&postcount=40)
George_oc from Hellas Overclocking Team for support online
jmke
Mate from OCA for support livestream

[George_o/c]

It was my pleasure dude! :)

[Punx223]

glad to be there for ya tin... it was fun watching ya beat on teh 285.... too bad it was being a pain in the ass.

[TiN]

Some more mad benchin was done today with T0lsty.

Q4EG + 5870 and Q4EG + 9800GX2 under LN2 both CPU+GPU.

http://www.xtremelabs.org/img/hwbot/hwbot_x2t.png

[Kal-EL]

Nice grab in 05 guys ;)