[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <73e61027b76522740e35385e25517ac3@smtp.hushmail.com>
Date: Mon, 7 Sep 2020 23:55:03 +0200
From: magnum <john.magnum@...hmail.com>
To: john-users@...ts.openwall.com
Subject: Re: NVIDIA GeForce RTX 3000 Series Performance
On 2020-09-07 10:26, Lukas Odzioba wrote:
> One more thing actually. I found this:
> https://en.wikipedia.org/wiki/FLOPS
>
> And in "FLOPS per cycle for various processors" table Ampere GeForce
> row we see a decrease in number of INT32 per cycle from 2 to just 1.
> So it makes even more sense to wait for benchmarks. Clocks are about
> the same and number of CUDA cores doubles so we technically could end
> up in a situation where 2080TI is about the same in hashing as 3080
> except some advantage in memory intensive hashes in favor of 3080.
Wow, that'd be a bummer :-(
However, they say things like "2 FP + 0 INT or 1 FP + 1 INT" so I'm
really hoping that could be appended with "or 0 FP + 2 INT" and that it
was simply left out bc they're only interested in FP.
But yes, only benchmarks will tell.
magnum
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
