eVGA 7900GT C0 SuperClocked babies

[eva2000]

Okay we all know some 7900GTX/7900GT manufacturer oc'd cards had GPU cores which couldn't handle the overclocks they were flashed with.

Question for end users:

Now as an end user and owner of several eVGA 7900GT CO SC cards (arrived today :D ), how does one properly AND efficiently time wise, test for the freezing issue ? Which apps are the quickest to use to pin point whether or not a new 7900GT CO SC card has the freezing issue ? Looping 3dmark2001/2003/05/06 ? Quake4, FEAR, BF2 ?

I just tested my 1st card with ending 4 digits of serial number in low 2000s and out of the box with stock heatsink and no mods, it completed 3dmark05 at 590(610)/1760 easy so at least first card looks promising :)

I'll be thoroughly testing each of my eVGA 7900GT CO SC cards one at a time, so any suggestions or feedback on the above questions would be greatly appreciated

p.s. Anyone experiencing delays in detecting of LCD monitor signal on system boot/reboot up with 7900GT CO SC cards ? On system boot/reboot, it takes 12-15 seconds for me to get a signal to my 20" BenQ FP2091 1600x1200 LCD monitor. I have Aten CS84A KVM switch with analogue to dvi adaptor to connect to 7900GT CO SC's 2nd DVI connector (closest to bottom of PCI slot bracket)

And the answer is no i won't be volt modding them anytime soon :)


Update: June 13th, 2006
Well card #2 the remaining 7900GT CO SC i had from the 3 purchased, ended up with same fault issue as other 2 cards around 4+ weeks ago.... rma'd the card will be getting back a eVGA 7900GT KO SC in it's place...

Hopefully this rma replacement doesn't end up with the same fate as others!

Card #1 and #3 were rma'd for a refund way back... so single card no SLI heh

Wish me luck guys :)

[[cTx]Philosophy]

I got off the phone with EVGA earlier today, and they told me due to the RMA of so many cards,(due to them not performing up to manu spec'd ratings) there experiencing a longer shipment time on them, due to more stringent testing to make sure said cards can do what there stated to be able to do, thats why im in such a long wait for my 7900gtx superclocked :(

So what I guess im trying to say is that there testing them harder for them not to fail before they ship them out now, maybe thats why your look so promising..

[eva2000]

Yes eVGA must be thoroughly testing these out now as this card out of the box just finished 595(615)/1820 3dmark05 - i just hope my other eVGA 7900GT C0 SC does the same it has 4 digit ending serial number of 2170s.


http://fileshosts.com/videocards/7900/eVGA/7900GT_Superclocked/1/photos/7900GT_001.JPG http://fileshosts.com/videocards/7900/eVGA/7900GT_Superclocked/1/photos/7900GT_015.JPG

1st eVGA 7900GT C0 Superclocked
- bios 5.71.22.14.04 dated March 9, 2006 see pic here (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/bios_screens/clocks.jpg). I dumped the bios and uploaded it here (http://fileshosts.com/videocards/7900/mybios/evga/7900GTSC/evga7900GTcoSC.zip) if anyone wants it.
- stock speeds = 550/1580
- Samsung BC14 1.4ns 610 (week 10, 2006) memory chips
- manages 590-595mhz on core and 1860 on mem fine when load temps are 63-66C.
- it managed 600/1860 when ambient temps dropped allowing 58-61C load temps but when load temp rises past 62+ it can't do 600/1860

http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/nvidia_bios_ver.jpg

Out of the box - stock heatsink, no volt mods:
595(615)/1860 was very comfortable and easy to do with FX-60 @2600mhz for 3dmark05 score of 10,686 (http://service.futuremark.com/compare?3dm05=1952109) and 3dmark06 score of 5,891 (http://service.futuremark.com/compare?3dm06=242909) :)

http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark05_595-1860_10686_2_tn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark05_595-1860_10686_2.jpg) http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark06_595-1860_5891_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark06_595-1860_5891.jpg)

@600/1860 with 58-61C load temps
http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark05_600-1860_10719_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark05_600-1860_10719_tnn.jpg)




================================

2nd eVGA 7900GT C0 Superclocked
- bios 5.71.22.14.04
- stock speeds = 550/1580
- Samsung BC14 1.4ns 610 (week 10, 2006) memory chips
- doesn't seem as good as first card looping 3dmark05 already hit a limit on core - not passing at 590(610)mhz so far with 63-65C load temps
- limit of this card on the core seems to be 580-583mhz core and 1840mhz on memory with 62-65C load temps which are close to same load temps at stock 550/1580


================================

3rd eVGA 7900GT C0 Superclocked
- bios 5.71.22.14.04
- stock speeds = 550/1580
- Samsung BC14 1.4ns 610 (week 10, 2006) memory chips
- has consecutive serial number with the 2nd card
- lowest load temps of the three eVGA 7900GT CO SC cards, @550/1580 load temp is 61-62C, idle temps around same 44-46C
- this card has similar core clocks to that of 2nd card but slightly better on core, around 585-595mhz core on this card




.

[t024484]

Question for end users:

Now as an end user and owner of several eVGA 7900GT CO SC cards (arrived today :D ), how does one properly AND efficiently time wise, test for the freezing issue ? Which apps are the quickest to use to pin point whether or not a new 7900GT CO SC card has the freezing issue ? Looping 3dmark2001/2003/05/06 ? Quake4, FEAR, BF2 ?

The most demanding and at the same trime the fastest program to test the cards stability (freezing issue) that I know is the "Comanche 4 Demo" benchmark.
When your board is not stable it will freeze almost in seconds, where other programs take many minutes.

[Highland3r]

EVGA seem to be doing a good job with these cards :D
Have a 7900 GT CO (not superclocked version) and it runs 550/900 stable on stock cooling/volts.
Had mem up to 990 earlier vmodded, Persivores still working on OCP for the card though :( Kicking in at ~ 1.3v atm :(

[DILLY]

great Score.:slobber: :fact: :woot: :slap: :p: :) :( :slobber: :p: :cool: :rolleyes: :confused: :nono: :slapass: :clap: :banana: :eek: ;) :lsfight: :gamer: :exclaim: :censored: :wiggle: :with: :worm: :fart: :rehab: :coffee: :bananal: :explode2: :kitten: :sick: :wth: :bs: :hehe: :banana3: :YIPPIE: :cow: :shock: :kitten2: :moon: :argue: :spam: :hrhr: :wtf: :chicken: :horse: :kittenhat :wave: :rocker: :brick: :thumbsup: :bounces: :eleph: :bows: :pimp: :shock2: :frag: :bounce: :cussing: :banana2: :wierd: :bday: :stop: :shoot: :confused2 :surf: :smoke: :bday2: :lol2: :kissbutt: :toast2: :sofa: :shakes: :hitself:

[cMw]

link to c4demo : http://www.techpowerup.com/downloads/57a/c4demo.exe

btw eva thats a great card so far for hitting 10.6k with those air clocks

im sure a little more juice from the fx60 and ram will break 11k :D

[eva2000]

Thanks for the tips guys...

Seems 1st card
- manages 590-595mhz on core and 1860 on mem fine when load temps are 63-66C.
- it managed 600/1860 when ambient temps dropped allowing 58-61C load temps but when load temp rises past 62+ it can't do 600/1860

@600/1860 with 58-61C load temps
http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark05_600-1860_10719_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark05_600-1860_10719_tnn.jpg)

[madgravity34]

Thats a nice card. What core rev. is yours? Also week? I just got my card in and its an 0610 core, week 10 of 2006 and we have the same week ram modules too lol. Where'd you get yours from? I got mine from tigerdirect before they sold out.

[eva2000]

no idea what core week is, still straight of the box and haven't lifted gpu heatsink off yet

Seems 2nd eVGA 7900GT CO SC isn't as good as first card looping 3dmark05 already hit a limit on core - not passing at 590(610)mhz so far with 63-65C load temps

Bought mine from http://www.shadowentertainment.com.au/home.php?cat=274 in Australia :)

[madgravity34]

Cool, well hopefully I get hit that 600 core mark. I just recieved the card today and after I verified it booted into windows I took of stock HSF and put in under water. I wanted to test it without ram sinks first so its only the core that has better cooling. So far idle is 40C and load is 41C lol. Almost no delta with this tiny core.

[eva2000]

So you never tested your card on air/stock hsf ??? why not ? I find it strange that folks sometimes skip this... I always thoroughly test the cards i buy on air/stock heatsink to know what they can do first :)

First time i used water on GPU was with X1800XT which was a great success... so probably water cool the 7900GT later on just need to buy a 2nd Maze4 LP block as i only have one which is on my X1800XT right now.

[madgravity34]

I did that mainly because the card was loading around 65-68C and I figured since the cards is going to be watercooled anyways, I was going to stick it on there and see how far it goes at stock volts.

[eva2000]

I see let me know how it goes on water... i assume you're using maze4 you listed in your sig ? Guess i'm more interested in the improvement of water over air, but you didn't test on air heh

[eva2000]

hmmm interesting 3rd card mem can do 1880mhz!!! so for SLI i have 1st card 595-600mhz core/1860mhz mem, 3rd card 585-595mhz core/1880mhz mem

So which is better as first PCI-E slot card for SLI ?

[eva2000]

Here's 1st eVGA 7900GT CO SC card with some overclocking

@stock 550/1580
3dmark05 = 9,830 (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark05_550-1580_9830.jpg)
3dmark06 = 5,588 (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark06_550-1580_5588.jpg)

3dmark05 @589/1880 = 10,781 (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark05_589-1880_10781.jpg)
3dmark06 @590/1892 = 6,005 (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC1/3dmark06_590-1892_6005.jpg)

Scacrificed 5-6mhz on GPU core for 20-32mhz more on memory heh

All stock heatsink no mods :D

here's very first 2x 256MB eVGA 7900GT CO SC SLI @stock 550/1580 with cards #1 and #3

3dmark05 = 13,937
http://service.futuremark.com/compare?3dm05=1954630

http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GTSC_13_SLI/3dmark05_550-1580_13937_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GTSC_13_SLI/3dmark05_550-1580_13937.jpg)

Strange though see in pic my control panel with 84.21WHQL drivers in SLI mode, the temperature display option has gone missing ?

3dmark06 = 8,704

http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GTSC_13_SLI/3dmark06_550-1580_8704_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT3x/10x/260-260-3325-7-14-2223_1.33-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GTSC_13_SLI/3dmark06_550-1580_8704.jpg)

Thought i'll pull this one out of the hat and no it ain't 3dmark05 score :D

@590(610)/1827 = 9,683
http://service.futuremark.com/compare?3dm06=245316

http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC13_SLI/3dmark06_590-1827_9683_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_8F5/7900GT_SC13_SLI/3dmark06_590-1827_9683.jpg)

[TiTON]

Those are some very comprehensive Screen shots :) I like them.. makes it easy to see how your system is setup :)

I am interested in seeing you go to water or better cooling. I wonder if these cards have a "warm" bug or not.

SAD

[LexDiamonds]

Great testing and impressive results Eva! I like the pics of the cards laid out.

It does appear that a Overclocked 1900xtx can match or exceed 2 overclocked Gts in 3dm05 at least. Food for thought.

[eva2000]

Thanks Lex, yeah thought that particular set of pictures would do the cards justice :D

Interesting i swapped the 2 cards around in SLI, so #3 card in PCI-E #1 and #1 card in PCI-E #2 slot furthest away from cpu socket, and idle temp reported by ATI Tool is 3-5C lower than with card #1 in first slot

Definitely card #3 runs way cooler than the other two and SLI proves it....

single card max
card #1 = 595-600mhz core and 1860mhz mem (can do 1880-1892mhz mem if i drop core to 590mhz)
card #3 = 585-595mhz core and 1880mhz mem

SLI max (PCI-E slot 1/2)
card #1/#3 = 590mhz core and 1827mhz mem (> 1827mhz mem will result in lower 3dmark06 HDR/SM3.0 scores)
card #3/#1 = 585mhz core and 1840mhz mem (this config results in 6C cooler load temps then card #1/#3 config!)

[eva2000]

okay strange i'm getting same 3dmark06 deep freeze locks http://www.evga.com/community/messageboard/topic.asp?TOPIC_ID=14044 ever since i changed from 84.21WHQL drivers to 84.43 beta drivers... it happens even at stock 550/1580 speeds single card and sli. No probs with 84.21WHQL initially but now changing back to 84.21WHQL i get same tearing/lock up at stock 550/1580 speeds ??? :confused:

with 84.43 beta drivers, in winxp event viewer i see

nv: silent running: rising temperatures caused level transition from L2 -> L1
nv: silent running: rising temperatures caused level transition from L1 -> L0

i don't get this message with 84.21WHQL but still get 3dmark06 deep freeze tearing

??

edit: confirmed this issue is happening on card #1 (the best clocker too :( )

[baracus2k]

Great testing and impressive results Eva! I like the pics of the cards laid out.

It does appear that a Overclocked 1900xtx can match or exceed 2 overclocked Gts in 3dm05 at least. Food for thought.
thing is ati optimizes their drivers for 3dmark05.... i remember when i had a x1800xl and scored around 9.5k... my 7800gtx 256 scored 9k but yet the 7800gtx beat the x1800xl in every game i had by 25-40%

[situman]

SO who's going to buy their non superclocked cards when everyone KNOWS all the good cards are already cherry picked?

[madgravity34]

Eva, I would try cooling the memory better. I was having that problem too until I sinked up the memory. If you touch them while running 3dmark they get extremely hot!! I used sinks and have the card 24/7 stable (so far) at 585/900, but I haven pushed the memory further and the core artifacts at 590. Without memory sinks I was seeing tearing at around 850 mem lol.

[eva2000]

thanks will try a fan over the card and see what happens..

it strange 84.43 betas might have done something to my card #1 as it now has tearing at stock speeds in 3dmark06 deep freeze test even on 84.21WHQL but no probs on card #3 which never saw 84.43 drivers only 84.21WHQL

[guess2098]

7900GT.... eVGA is going to relase some good one, super super oc card soon

[eva2000]

plucked in card #2 and looping 3dmark06 at stock fine so far... so looks like card #2 and #3 with consecutive serial numbers in low ending 4 digits of 2170s are fine and only card #1 has tearing issues at end of 3dmark06 deep freeze loop tests.



interesting i just plucked in both remaining cards i have

PCI-E slot #1 = card #3 idles at 37-41C (usually in SLI slot #1 is always hotter)
PCI-E slot #2 = card #2 idles at 41-44C

when i had card #1 + #3 in SLI, 84.21WHQL drivers didn't slow temperature panel in nvidia control panel to be able to see both card temps

but with card #3 + #2 in SLI, the temp control panel exists and can view both card temps

[eva2000]

what's wrong with this score with SLI with 2x 256MB eVGA 7900GT CO SC @550/1580 stock speeds :lol:

http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_7F5/7900GT_32_SLI/3dmark06_550-1580_10280_tn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_7F5/7900GT_32_SLI/3dmark06_550-1580_10280.jpg)

http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_7F5/7900GT_32_SLI/3dmark06_550-1580_10294_tn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/250-250-3325-7-14-2223_1.48-1.3-1.5-2.64_1552_ds8dds2_7F5/7900GT_32_SLI/3dmark06_550-1580_10294.jpg)

[eva2000]

proper score :)

best i can do so far with card #3 + #2 in SLI (held back on mem clocks for now to make sure these 2 cards are okay :)

3dmark06 @580(600)/1780 = 9,627
http://service.futuremark.com/compare?3dm06=249573

http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/255-255-3325-7-14-2223_1.57-1.3-1.5-2.64_1552_ds8dds2_7F5/7900GT_SC32_SLI/3dmark06_580-1780_9627_tnn.jpg (http://www.fileshosts.com/DFI/NF4_SLI_D/results/FX60/HC200EX/OCZ/PC4000EB/2_1/1114_3/LDT4x/12x/255-255-3325-7-14-2223_1.57-1.3-1.5-2.64_1552_ds8dds2_7F5/7900GT_SC32_SLI/3dmark06_580-1780_9627.jpg)

[s e t h]

do you have some serious software tweaks going there eva2000?
i've got a pair of 7900gt's at 710/1900 with an X2 3800+ at 2.5ghz and some ddr500 bh-5 and my sm2 score is way lower than yours. (4105)
my sm3 is is 4338

even if i plug your cpu score in to my theoretical calculator i'm still 300 points short of your score.

are you using the latest drivers or the 84.21's for that ^ score

maybe a decent cpu would help me get higher sm2/sm3 scores.

anywa your scores are absolutely flying there bud. a couple of mods will see you take the WR and i'll never catch you then :P

[cronic]

nice score eva2000, looks like this is the week my world records in 05 and 06 fall. 05 was already taken on the 14th. oh well, i have a new "weapon" coming this week.

Nice 10294 in o6. great score and congrats on knocking me off the top.

[s e t h]

scuze the off topic eva2000.

cronic - if i may ask. are you running LOD tweaks or not?

[cronic]

scuze the off topic eva2000.

cronic - if i may ask. are you running LOD tweaks or not?

no, no tweaks at all

[eva2000]

flytek, yup using 84.21WHQL no game to try 84.43 beta after failure of card #1 heh

i think fx-60 @3060mhz is helping me there - that is probably equivalent to a X2 3800+ @3260mhz ;)

cronic, 10,280 and 10,294 are bugged runs only 9,683 with card #1+#3 and 9,627 with card #3+#2 are proper runs... bugged runs notice the cpu score ;)

oh and if you clock your mem too far, you can lower your scores more so in HDR/SM3.0 deep freeze test, if you clock your gpu core to higher can lower your scores particularly in canyon test in 3dmark06

[cronic]

yeah i see the insane cpu score now.

[eva2000]

Yeah your 3dmark orb rank is safe :D

[eva2000]

hmmm now with card #3 + #2 in SLI config, 3dmark06 deep freeze is fine looping with no tearing

BUT trying 3dmark2003 i found that Troll Lair GT3 test is showing tearing.. and even one instance of locking up.... lowered the default memory from 1580mhz to 1500 and no tearing! so seems card #3 has issues with 3dmark2003 GT3 when in SLI.. could be heat related ???

aborting out of 3dmark2003 GT3 Troll Lair looping with ATI Tool and nvidia panel on GPU2, I see temps of only 43-46C

still using 84.21WHQL drivers with system specs same as listed above

[s e t h]

i must say that one of the two cards i have definately has weaker memory than the other one.
so you might have one weak ram chip there somewhere.
also if you try the comanche demo from techpowerup you may see that your overclocks are actually far from stable.
i found doing the resister pencil mod, in the 7900 vmod sticky, that boosts the switching frequency made my cards a lot more stable in comanche.
the vmem can also be done with a pencil fairly easily too.

and @ cronic - respect for not using the LOD tweak sheeite. at least we know what the playing field is ;)

[kite7]

SO who's going to buy their non superclocked cards when everyone KNOWS all the good cards are already cherry picked?

Atleast I can still hit 1.8ghz memory and 560mhz core with my standard 7900gt

[sup3rfly]

hello there i read about the "freeze" issue with the 7900gtx.... i just wanna know if there is any problem with the 7900gt oc as well???

[eva2000]

Yeah issue can affect 7900GT CO non-SC but to a lesser extent

To confirm my issues, i will do a fresh winxp pro sp2 install on a spare 3.46Ghz 955XE/Abit AW8-Max system I have here as well and test each card individually to confirm if the issue is driver/system specific...

finished setting up fresh winxp pro sp2 install for 955XE/Abit AW8-Max and am doing basic system stability tests first now. I made a back up image of OS partition before video drivers were installed so i can restore a backup image to try fresh/different drivers easily :)

System:

P4 3.46Ghz 955XE with Swiftech G4 Storm H20
Abit AW8-Max 16 bios
2x 512MB Corsair PC5400UL v1.5 3-3-2-8 DDR2
80GB Seagate 7200.7 PATA hdd
LiteOn 165H DVD-Rom
FDD
650W Silverstone SLI psu

WinXP Pro SP2
84.21WHQL drivers

http://fileshosts.com/abit/aw8_max/results/955XE/SwiftechG4Storm/Corsair/5400ULv15/268_284/16bios/17x/204-307-3328-1.2875-1.65-2.1/stressprime2004_quad_load_tn.jpg (http://fileshosts.com/abit/aw8_max/results/955XE/SwiftechG4Storm/Corsair/5400ULv15/268_284/16bios/17x/204-307-3328-1.2875-1.65-2.1/stressprime2004_quad_load.png)

to test card #1

Result:
Initially, 3dmark06 deep freeze looped a few loops at stock 550/1580 without any tearing issues. I thought this was maybe due to no 84.43WHQL driver interference or the fact that on Abit AW8-Max my memory Corsair PC5400ULv1.5 needs chipset voltage for motherboard at 1.65v. Then loop it for longer 30 minutes and notice slight tearing issues in deep freeze 3dmark06 looping (not as bad as on FX-60/DFI NF4 UT SLI-DR rig).

I stop the looping and load winxp event viewer and notice the 'nv rising temperatures caused level transition...' messages - which according to mvktech.net's Johan means thermal throttling occured due to high temps. Now I had ATI Tool logging temps in background too so thought to compare the timestamps of when the event viewer logged the messages with the temps at that particular time in ATI Tool logs.

Notice anything strange about this screenshot of 9:55:16 pm time stamp and the temp recorded ? It's at the lowest temp of 42C ???? highest temp logged was like 60C

http://fileshosts.com/abit/aw8_max/results/955XE/SwiftechG4Storm/Corsair/5400ULv15/268_284/16bios/17x/204-307-3328-1.2875-1.65-2.1/7900GT_SC1/thermalthrottle/42C_throttle_weird_L2toL1_tnn.jpg (http://fileshosts.com/abit/aw8_max/results/955XE/SwiftechG4Storm/Corsair/5400ULv15/268_284/16bios/17x/204-307-3328-1.2875-1.65-2.1/7900GT_SC1/thermalthrottle/42C_throttle_weird_L2toL1.png)

maybe faulty temp sensors on the cards or bios/driver related or both ?



more tests to come..


seems tearing in looping deep freeze 3dmark06 test at stock 550/1580 for card #1 got pretty bad once gpu load temps hit 67C - ATI Tool log (http://fileshosts.com/abit/aw8_max/results/955XE/SwiftechG4Storm/Corsair/5400ULv15/268_284/16bios/17x/204-307-3328-1.2875-1.65-2.1/7900GT_SC1/thermalthrottle/bad_tearing_550-1580_30xloop_67Cmaxload.png) :(

lowering mem clocks below rated 1580mhz speeds reduced or fixed tearing issues .. so i say gpu heat is affecting memory clock stability ?????

[eva2000]

Update:

card #2 is still solid been running it all day today with no probs at all i've backed off from the max the card does 580-583mhz core and 1880mhz mem to settle for 570-576mhz core and 1760mhz mem for testing so far not one problem at all :)

3dmark06 = 6,033
http://service.futuremark.com/compare?3dm06=259776

http://fileshosts.com/abit/aw8_max/results/955XE/SwiftechG4Storm/Corsair/5400ULv15/268_284/16bios/21x/231-345-3328-1.6375-1.65-2.1/7900GT_SC2/3dmark06_575-1760_6033_tn.jpg (http://fileshosts.com/abit/aw8_max/results/955XE/SwiftechG4Storm/Corsair/5400ULv15/268_284/16bios/21x/231-345-3328-1.6375-1.65-2.1/7900GT_SC2/3dmark06_575-1760_6033.png)

so

card #1 = faulty
card #2 = OKAY
card #3 = faulty

heh

[madgravity34]

Damn that sucks. I probably need to push mine harder, I have been too busy upgrading. Today I recieved my Expert and hopefully 2morrow I get my new tubing in to re-run my loop. Once thats done and I get the pc booted Ill just play with the card. The only thing I can say is that at stock when I initially tested it I never had any problems with 03/05/06 and I must have done 15 runs of 05 and 10 03 runs. Only ran 06 twice. Eva, did you get a chance to checkout the core rev's of your working gt?

[revenant]

ok, so you're running all these tests with no vmods implemented, correct? didn't see anything about any, but wanted to ask. :)

[madgravity34]

Yea he has done no vmods yet.

[eva2000]

no mods at all just stock hsf :)

no idea what's on the cores i did reapply Artic Ceramique on card #3 didn't help but forgot to take note of the core i think it was 610 dated as well same as memory chips

anyway packed both cards for rma, sending back to my local retailer :)

[madgravity34]

Damn, I have the same 0610 rev. on my card. Since you are keeping that #2 card do you think you could do me a BIG favor and take a peak under the HSF? hehe...

[eva2000]

well card #3 is only 1 digit above card #2 in serial numbers so i assume probably same core date if they all have 610 samsung memory as well

[revenant]

wow - two out of three bad.. not the best odds. I am getting two 79gts today.. well, I hope at least one of them if good. I have been cardless for too long. ;)

[eva2000]

Update: June 13th, 2006
Well card #2 the remaining 7900GT CO SC i had from the 3 purchased, ended up with same fault issue as other 2 cards around 4+ weeks ago.... rma'd the card will be getting back a eVGA 7900GT KO SC in it's place...

Hopefully this rma replacement doesn't end up with the same fate as others!

Card #1 and #3 were rma'd for a refund way back... so single card no SLI heh

Wish me luck guys :)

[Highland3r]

Ouch, so thats all 3 cards that have died!

[eva2000]

yup first time ever in over 9+ yrs of pc use, that i had a video card die.... never had one die before so to have 3 in a row heh

[rob[GL]]

I bought TWO 7900GT KO SUPERCLOCK cards, and I returned them the very next day... Bought myself an ATi RADEON X1900CF 512... The single ATi card performs much better than both 7900's put together in SLI...

Maybe I got two of the problem 7900's?

I don't care, it was the first time, and last time I'm going to try Nvidia...

[Reinvented]

yup first time ever in over 9+ yrs of pc use, that i had a video card die.... never had one die before so to have 3 in a row heh

eva, that you posting on evga forums? ;)

6 cards here are dead also. RMA after RMA...they offered me 7800GTX's, but i declined. I ended up having to call the supervisor because I cant wait! I want to bench too. :(

However, my GT KO's, ONLY lasted 3 days tops. Well, not even a full 3 days...more like 2 days and some odd hours.

Are you disappointed? How many cards do you have left?

[eva2000]

heh yeah me over there.... bought 3x eVGA 7900GT CO SC... all 3 are faulty.. 2x rma for refund and 1x rma for replacement with eVGA 7900GT KO SC due to arrive this week or even today if i'm lucky... just hoping it ain't about dud..

so you're the one with 6x eVGA 7900GT KO SC rmas ?

[Reinvented]

yeahhh :( Sorry, just GT KO's.. SC has a premium...i don't feel like paying it