Review - AMD Radeon R9 290X

[Milzo]

any ideas as to why sha256crypt on twin 290x's is way down, all the other crypt hashes have increased.

https://ow.ly/tF7TZ

[magnum]

any ideas as to why sha256crypt on twin 290x's is way down, all the other crypt hashes have increased.

Did you set memory clocks the same too?

[Milzo]

I haven't got the 7990 anymore but the mem clock on that was set at 1500Mhz.

Nevertheless i pushed the mem clocks on the on the 290x's to 1500 from 1250, though still the same.

Code:

#od6config --adapter 0,1 --set core=1050,mem=1500,fan=100,power=+50

Adapter 0: AMD Radeon R9 290 Series
    New clocks          : 1050 Mhz core, 1500 Mhz memory
    New fan speed       : 100%
    New powertune value : 50

Adapter 1: AMD Radeon R9 290 Series
    New clocks          : 1050 Mhz core, 1500 Mhz memory
    New fan speed       : 100%
    New powertune value : 50


#./oclHashcat32.bin -b -m 7400
oclHashcat v1.01 starting in benchmark-mode...

Device #1: Hawaii, 3072MB, 1050Mhz, 44MCU
Device #2: Hawaii, 3072MB, 1050Mhz, 44MCU

Hashtype: sha256crypt, SHA256(Unix)
Workload: 64 loops, 8 accel

Speed.GPU.#1.:    23001 H/s                  
Speed.GPU.#2.:    23023 H/s
Speed.GPU.#*.:    46023 H/s

[ecos]

Having done that you can also OC the card to 1100/1350 or so, with no problems.

--> Before I go for the overclocking option, I would very much like to see what I did wrong before. Also, this goes into a production box, OC is not yet an option, it might become an option after heavy stress testing.

Edit: I see now you did use the od6config tool, so my bet is you did not flick that switch on the card. It needs to be set "away from the rear".

--> The switch had already been set "away from the rear".

Edit2: If the card still throttles despite everything done right, it really gets too hot (>94C) and you should be thankful it saves itself by throttling. Fix cooling if this happens.

--> Temp without OC is somewhere over 75, but not higher than 85.

Have you had the chance to test some NTLM hashes? Just for reference, I'd like to find out if my measurements are really in the woods or not ... if not NTLM stats are available, any alternative stats (not necessary MD4) are fine to compare.

Edit - plus I am quite confused about the results of the benchmarking mode (-b), which are fine, as expected ... the trouble starts when I try to crack real data.

Cheers,
ecos

[epixoip]

benchmark mode is single hash brute force. accel/loops setting, attack mode, word length, keyspace, number of hashes, etc all impact the real-world performance.

[ecos]

benchmark mode is single hash brute force. accel/loops setting, attack mode, word length, keyspace, number of hashes, etc all impact the real-world performance.

I'm aware of this, that's why I have asked if my real-world measurements are correct, similar to others, or I am doing something wrong. I was also surprised by effect of the number of hashes on the performance on 290X (and not on 6990), as the NTLM/MD4 is not salted. If no measurements are available, no prob, I'll keep digging

Cheers
ecos

[magnum]

I get real-life speeds (for a single hash) on par with the benchmarks, and disregarding the exact figure it's not fluctuating much. Are you not using the -n option? When you use -b you'll get hints for what figures to use for -n and -u in a real run.

[ati6990]

hmm to much hype of those 290x (500 uro) i think....
next time i try to beat them with my msi 7970 be and my gigabyte windforce 5x , maybee some mhz can kill streams.

[epixoip]

i can achieve real-life single hash brute force speeds on par with the benchmarks as well. but of course multi-hash and other attack modes, performance drops.

[epixoip]

hmm to much hype of those 290x (500 uro) i think....
next time i try to beat them with my msi 7970 be and my gigabyte windforce 5x , maybee some mhz can kill streams.

clock-for-clock the 290X is almost consistently 36% faster than the 7970. so no.