Benchmarks for hashcat v4.00 on AMD and NVIDIA cards

[Nikos]

Hello.

I'm looking for benchmark results of the latest hashcat v4.00 running on AMD and NVIDIA cards.

A Polaris vs Pascal vs VEGA excel sheet, like those posted for earlier versions of hashcat.

Does anyone have a link ?

Is it true that VEGA is a lot faster than GTX 1080 ?

And how about a VEGA liquid vs 1080 Ti comparison ?

TIA!

[Flomac]

Someone posted a VEGA benchmark some months ago and it's not very impressive. No comparison to Pascal. And no idea if it changed over time with better drivers. But I guess there will still be a big gap.

https://hashcat.net/forum/thread-6739-page-2.html

[atom]

Most relevant, maybe not perfect: https://docs.google.com/spreadsheets/d/1...=785768233

[Nikos]

So, according to atom's relevant reply - benchmarks leveraging hashcat v4.00 - it is true.

VEGA 64 Air is A LOT FASTER than GTX 1080.

Waiting for the VEGA 64 Liquid results vs 1080 Ti

[Chick3nman]

So, according to atom's relevant reply - benchmarks leveraging hashcat v4.00 - it is true.

VEGA 64 Air is A LOT FASTER than GTX 1080.

Waiting for the VEGA 64 Liquid results vs 1080 Ti

I wouldn't call it a "lot" faster than a 1080, in most of the common algorithms it's only a difference of a few percent, but there are definitely some algorithms where VEGA shines. Of course, all of this is trumped by the fact that the VEGA is using roughly 2x as much power as the 1080(180w vs 350w). In performance per WATT comparison, it's not even close. Also, the liquid VEGA cards will not be much faster than the air cooled since they will still be saturated by the heat under full load. The 1080Ti will also still beat the VEGA64 in most, if not all algorithms. And all of this is moot anyways because Pascal is already out of production. Volta pretty handily creams Pascal judging from the V100 benchmarks.

[atom]

Benchmarks are from Vega64 Air, which has 290W TDP (and from my experience it never pulls more than 250W). It's the liquid version which draws 350W.

Anyway the V100 is the killer if it wouldn't be so damn expensive. Maybe next years Volta consumer GPU's will change things.

[cpu_pirate]

I am interested if anyone has actually tested these results, when I first got my 2 vega 64's I posted a benchmark that looked really good, just like the one atom posted, however, in practice the card was actually about 40% slower than the benchmark speed.

[Chick3nman]

@atom I wasn't aware there was that big of a difference between the AIR and Liquid VEGA cards. Still doesn't bring it down to the 1080's perf/watt level but it's good to note. Volta GeForce cards are definitely going to be top dog soon.

[Nikos]

My impression is that you are talking about a chip that is simply not existent for the 99,99999% percent of the people - and I'm referring of course to Volta.

Only the very expensive chip GV100 is available for data centers and there are no consumer cards of Volta.

Actually, just don't expect any consumer cards soon, till at least Q2 of 2018.

But around Q3 of 2018, Navi will be available too, so let's keep the discussion around real cards please.

It's Polaris vs Pascal vs Vega comparison only.

[Flomac]

Volta will be very likely launched/available for the comsumer market in Q1 2018, early Q2 2018 the latest. And the coming up card will be again around 70% faster as its predecessor, the GTX1080.

Navi being available in Q3 2018 is highly doubtable. AMD wants to use 7nm and since a new and smaller process usually works fine for small chips like DRAM or smartphone-SOCs, huge GPUs are completely different beasts to handle. They will surely try to catch the christmas market in 2018, but we'll see if yield rates are going to be high enough for more than a bunch of show cards. Or if they sell it first exclusivly to the automotive market like NVidia is doing it with Volta now.