About the new gromacs core

Yes everyone there is a new gromacs core out right now
version 1.47 made it's debut yesterday.

with this new version you will see a 33% production decrease.
yes i said decrease.

They seemed to be having stability issures with the sse instructions with the amd cpu's thus this new core disables the sse support but enables the 3dnow support by default for a amd cpu.

now for a intel cpu it keeps the sse instructions on by default

and yes there is a fix for us that weren't having stability issues to be able to switch back to sse by default...that will be released soon

as always may the fold be with you

scoobiedoo

*edit* if you want to read about the new core and the stability issues please goto this article:
http://forum.folding-community.org/viewtopic.php?t=4409

Well,at least they found the trouble.I was one who had some trouble with the other core.It would lock just 2 out of the eleven I run right now even at default speeds.I wish I would have seen the post in the folding forums.The ones that gave me so much trouble where the 542's.I tried everything to get those to run.They would run prime and memtest (ran 48hrs each)but not those pesky gromacs.Standford shows that I managed to do 328 of them.I just thought it was my mach's,glad to see it wasnt.Thanks for the heads ups Scoobiedoo.

i'm running an amd with the forceasm flag and i never had trouble with the 542's. i love those things.

oof! that hurts.

Well,at least they found the trouble.

Click to expand...

I'm not so sure they found the trouble so much as a way to make the trouble stop (Patient: Dr., it hurts when I do this. Dr: Then don't do that!)

I understand this is just an interim step, but on the whole, a solution as extreme as this one can hopefully be refined so that we can have SSE back even on the problem XP's

I must admit, I've got one AMD rig with stability problems @stock. I'll have to see how it does on the new core. As for the rest of 'em, I'm keeping an eye peeled for the switch to turn SSE back on

I think that's pretty dumb.. There should be an option to enable 3dnow instead, if you're having stability problems. All of my rigs are rock solid with gromacs, and now I have to pay the price for (mostly) people who wouldn't lower their OCs a little bit to make their system stable.

If you ask me, the production they lose from crashing clients is probably far less than a 33% drop in production from every Athlon XP folding out there. Its good they try to solve the problem, but this seems a bit drastic and hurts the project

/edit I finished reading the thread you linked and I see at the end they are calling it a temporary thing. Sounds like they will add some sort of force SSE command later on for those of us who are stable.

-Rav

Well there is something strange to report: With the new core and my 2525Mhz AMD, a P540 is now taking 8:10 a frame, with the older core, I got like 5:10 a frame on the same protein. Here is the weird part. On my dually running 2 instances as usual, also with the new core and a new protein, a p909 worth 38 points, the CPU's at 1950 Mhz, I get 7:23 a frame. Maybe these newer proteins work well with the new core? Who knows but the production loss sux0rs big time...

Okay listen up everyone here are a couple of screenies comparing an intel cpu with the new gromac core and a amd cpu with the new gromac core:
what you need to notice is the circles lines

mspaint RULES!