benchmarking fah units

[RETARD]

in the interest of hardware comparisons - is there a way the same unit(s) can be used to compare data across different types of hardware?
if not, perhaps someone could share a successful way to do so
TIA

[p0wer]

I guess Stanford could gather such data, but given the amount of people folding this wuld be hell lot of information and they don't want to/don't have resources to share it. All we have is stats on folding web page (actually, only this)

There are some web sites around that tried to gather stats, it looks like not many people know about them though.

[HendricksSA]

Not sure which work units you are wanting to compare. There are lots of topics on the bigadv size work units. One good one was provided by Grandpa_01 at viewtopic.php?f=55&t=14757#p145568

In the past, what folks have done is pick a work unit they want to compare and then create a new topic (such as: viewtopic.php?f=55&t=17730 and ask for inputs.

[Evil Penguin]

As far as nVidia vs ATi goes...
p11293 (ATi) and p6800 (nVidia) could possibly be compared.

[toTOW]

For my reviews on FAH-Addict, I backup a bunch of significant WUs : one of each type (number of atoms typically) that a client can run, and I run offline with a v6 client set to ignore deadline informations.

[patonb]

Theres a certain forum that a member made a test setup using a few units, that can be run.

Just don't try to upload the results as theyre way outa date.

I'd post it but last time I got bitched at for recuiting, yet I've never folded forum... Theyre just one of the top teams that are extremely hardware knowledable

[gwildperson]

We used to contribute our information to http://fahinfo.org/ and it did an excellent job of comparing performance on different hardware when we were talking about the differences in CPU hardware. Now everybody wants to compare performance of GPUs and that really means [GPU with SMP] performance.

Some 3rd party needs to design a new database that allows for the additional complexity of multi-client/multi-slot configurations. We need somebody to volunteer, or the author of fahinfo needs to upgrade the design of his site. Finding things in a long list of forum posts can be a real drag.

[k1wi]

The biggest problem is that benchmarking is such a fraught issue. All it takes is for a user to be running an application in the background and you could end up with a benchmark that is passively undervalued.

k1wi