2x7990 or 4x7970
#21
agree blandy. and that's why we started putting a max of three cards per chassis, with one pci slot in between each card. that way the cards are not directly side-by-side and have adequate airflow. four cards is doable if you can put some kind of spacer in between them, but still not ideal.
#22
Before putting Arctic Accelero Coolers on:
http://home.btconnect.com/md5decrypter/g...280004.JPG
http://home.btconnect.com/md5decrypter/g...280009.JPG

After, (The cardboard underneath stops the heat being blown down and cooking mobo / CPU / PSU etc and it was a problem!):
http://home.btconnect.com/md5decrypter/g...200029.jpg

Examples of dust on existing reference cards, (they were second-hand obviously):
http://home.btconnect.com/md5decrypter/g...40001_.jpg
http://home.btconnect.com/md5decrypter/g...40004_.jpg
http://home.btconnect.com/md5decrypter/g...40005_.jpg
Owner of: www.md5decrypter.co.uk
#23
I'm really surprised that you needed to put Accelero coolers on 5870s, especially when you had that much spacing in between them. I like your setup there though, that's pretty slick!
#24
I could not overclock the cards at all with reference coolers on, cracking hard-ass algos like WPA/WPA2 or mscache2, they would still hit 90oC on stock clocks easy. With Accelero's on, they dont go over 66oC when clocked @ 900MHz core and 1300MHz mem. I can clock higher but keep becomes unstable after 950MHz.

That said, the bloody dust build-up on the reference coolers caused most of the over-heating issues lol Smile but cleaning the reference coolers is not easy, have 2 take them totally apart!
Owner of: www.md5decrypter.co.uk
#25
Some time passed and I got a PC! Big Grin
But the speed test shows less than expected:

Quote:Hashes: 1 total, 1 unique salts, 1 unique digests
Bitmaps: 8 bits, 256 entries, 0x000000ff mask, 1024 bytes
Workload: 256 loops, 80 accel
Watchdog: Temperature abort trigger set to 90c
Watchdog: Temperature retain trigger set to 80c
Device #1: Tahiti, 2048MB, 1100Mhz, 32MCU
Device #2: Tahiti, 2048MB, 1100Mhz, 32MCU
Device #3: Tahiti, 2048MB, 1100Mhz, 32MCU
Device #4: Tahiti, 2048MB, 1100Mhz, 32MCU
Device #1: Kernel ./kernels/4098/m0000_a1.Tahiti_1084.4_1084.4 (VM).kernel (322816 bytes)
Device #2: Kernel ./kernels/4098/m0000_a1.Tahiti_1084.4_1084.4 (VM).kernel (322816 bytes)
Device #3: Kernel ./kernels/4098/m0000_a1.Tahiti_1084.4_1084.4 (VM).kernel (322816 bytes)
Device #4: Kernel ./kernels/4098/m0000_a1.Tahiti_1084.4_1084.4 (VM).kernel (322816 bytes)
...
Speed.GPU.#1...: 5313.1M/s
Speed.GPU.#2...: 5329.8M/s
Speed.GPU.#3...: 5318.6M/s
Speed.GPU.#4...: 5312.8M/s
Speed.GPU.#*...: 21274.4M/s

I saw HD7970 test on another post - the speed was about 8000 M/s. Why I got 5300 on the same HD7970? Maybe I need something to adjust? Thank you.
#26
Quote:Workload: 256 loops, 80 accel

that's what you need to adjust. if you're using lite, use "-m 0 -b --benchmark-mode 1" to test your speed. if you're using plus, use "-n 800 -u 1024" with a large keyspace.
#27
Thank you. Tryed "-n 800 -u 1024" with another parameters (as in theme '8 x AMD HD7970 @ 1050MHz'):
-n 800 --gpu-loops 1024 -1 ?l?d?s?u --outfile-watch 0 --restore-timer 0 --runtime 10 --pw-min 8 -m 0 ffffffffffffffffffffffffffffffff ?1?1?1?1?1?1?1?1?1

Now I got:
Quote:Speed.GPU.#1...: 9733.2M/s
Speed.GPU.#2...: 9733.4M/s
Speed.GPU.#3...: 9733.5M/s
Speed.GPU.#4...: 9733.5M/s
Good result!
But now I see that I need to install additional 4 fans, because cards temperature reaches 90 degrees. I'll work on that.
#28
I need to make something like this:
http://home.btconnect.com/md5decrypter/g...280004.JPG
http://home.btconnect.com/md5decrypter/g...280009.JPG

I should use "Adapter Extender Cable": http://cablesaurus.com
If I will use this "x1 -> x16" - how it affect the speed? Also, If I will use two of them to get cable with length*2, how it affect the speed?
#29
that's blandyuk's rig.

the speed will not be affected using a 1x slot.

i've daisy-chained extenders and not noticed a difference in speed.
#30
x1 will not affect speed for hashcat.

;D I'm too slow