PDA

View Full Version : Looking at AMD Dual Socket C32 Motherboards, worried though.



rgathright
03-09-11, 09:31 AM
I'm getting tired of watching kevint and Spankinmonkee kick my butt in Androinc and intend on hitting #1 there by November 2011! :mad:

My solution is to build a new Dual Socket C32 system but worry that these motherboards will not take the new Bulldozer processors.

Anyway, here they are... can you tell me which one I should get?

Cost is not the problem since I want to get it right the first time.

ASUS KCMA-D8 Dual Socket C32 AMD SR5670 ATX Dual AMD Opteron 4100 Series 4/6 Core Processor Server Motherboard
http://www.newegg.com/Product/Product.aspx?Item=N82E16813131670
http://www.mwave.com/mwave/specHR/images/AA98998a.jpg

TYAN S8225AGM4NRF Dual Socket C32 AMD SR5690 Extended ATX Dual AMD 45nm 4-Core/6-Core Opteron 4100 Series Processors (Lisbon) Server Motherboard
http://www.newegg.com/Product/Product.aspx?Item=N82E16813151226
http://www.coolcomputing.com/images/tyan-S8225WAGM4NRF.jpg

Fire$torm
03-09-11, 12:28 PM
WoW!!! Sweet boards but I'm clueless :(

Looks like a question for the "J-Crew" \m/

Dandasarge
03-09-11, 12:58 PM
I'm not 100% on this but all the ways I've worked it, it’s been more cost effective to build 2 cpu’s vs one duel. Only thing that brings the singles is the 2 PSU’s. But I’ve never seen a cost effective motherboard. Also the downside is if one component fails you could lose all production vs ½. I’m not trying to talk you out of this build I just want the best value going into our crunchers.

John P. Myers
03-09-11, 03:51 PM
No those motherboards will not be compatible with Bulldozer. If money isn't an issue and you want to do some serious CPU spanking, go with this quad CPU 48 core badboy. (http://www.newegg.com/Product/Product.aspx?Item=N82E16813151219)

Or one of these dual CPU 24 cores (http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007629%20600008002&IsNodeId=1&name=Dual%20Socket%20G34) might be a bit more reasonable.

Bulldozer is socket G34, same as these mobos. It should only require a BIOS update to get them working.

rgathright
03-09-11, 04:16 PM
No those motherboards will not be compatible with Bulldozer. If money isn't an issue and you want to do some serious CPU spanking, go with this quad CPU 48 core badboy. (http://www.newegg.com/Product/Product.aspx?Item=N82E16813151219)

Or one of these dual CPU 24 cores (http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007629%20600008002&IsNodeId=1&name=Dual%20Socket%20G34) might be a bit more reasonable.

Bulldozer is socket G34, same as these mobos. It should only require a BIOS update to get them working.

So what is the desktop version (IE: AM3 and C32) of the Bulldozer really called? The news sites have me all confused over here.

Already played with the G34 while building a VMWare box, they are real nice but the price per CPU is insane.

John P. Myers
03-09-11, 04:50 PM
So what is the desktop version (IE: AM3 and C32) of the Bulldozer really called? The news sites have me all confused over here.

Already played with the G34 while building a VMWare box, they are real nice but the price per CPU is insane.

Bulldozer based desktop will be socket AM3+
Server socket G34

YoDude9999
03-21-11, 10:46 PM
[Dude shakes his head] What are you guys? Die hard CPU cruncher fans?...
The PG&E people are gonna do something to your A$$ that you won't like when you flip the switch on!

I suppose, in reality though, a pile of GPUs sucking down the wattage is just as bad in the long run.

I just have to wonder, is CPU crunching really worth it?

joker
03-21-11, 10:57 PM
I just have to wonder, is CPU crunching really worth it?

Depends on what you are trying to achieve. Only a few projects (at this time) use GPU power. Not a good thing for all the MM's out there. The other part of it is (IMHO) not everyone has a crunching GPU.

Maxwell
03-21-11, 11:23 PM
Depends on what you are trying to achieve. Only a few projects (at this time) use GPU power. Not a good thing for all the MM's out there.
+1

I'm much prouder of my 1M in the CPU projects than the GPU projects. Leiden was my most recent 1M project, and that took me over a month of going at it HARD. 1M in DNETC? 14 hours. If I were a "max credit" cruncher, I'd have just a good enough CPU to feed the GPU, and probably not crunch on it. I would also be crunching fewer projects.

joker
03-21-11, 11:34 PM
Wait a minute!! Maxy and I agree on something???.....Dear diary..... :p

Maxwell
03-21-11, 11:36 PM
Wait a minute!! Maxy and I agree on something???.....Dear diary..... :p
Dear Diary,

joker met up with my mom last night...:((:p

YoDude9999
03-21-11, 11:42 PM
@ Joker,

I suppose that's actually correct. Not as many GPU crunchers around and a short supply of GPU projects to choose from.

From what I've personally seen however, is that CPU crunching seems to be a failing means to a worthy cause. The credits are so low for CPU crunching that it seems to me to be almost worthless, though I do see those Aqua credits as something to take note of. Now if the rest of the projects threw out credit as well as the GPU projects did, well guess what? There would be an even playing field throughout the land and it wouldn't matter if you were CPU or GPU crunching, it would all be the same crunching.

I have to admit I'm a credit whore and therefore is the reason I joined up to TopGun. Not for any project in particular, but for the most credit a project is offering. And if the project a person chooses to run is what suits them, it is not my call to say anything about it, they will do what they like and that is fine, it all adds up as long as we're all on the same TEAM!

Yo-

joker
03-22-11, 12:17 AM
@Yodude. I am totally with you about the credits thing (I too am a credit whore). I have several GPUs crunching to get max credits (about 900K a day). On the other side, I have built my systems with the "max" CPU power as well so that I can get some MM's as well (say nothing Max!). As you said, far less credits but (as you also said) all those credits go to the team so we all win!

@Max....Since we are talking about moms.... http://www.hulu.com/watch/72434/saturday-night-live-snl-digital-short-motherlover-censored

Maxwell
03-22-11, 01:37 AM
@Max....Since we are talking about moms.... http://www.hulu.com/watch/72434/saturday-night-live-snl-digital-short-motherlover-censored
That was awesome. And I must say, Susan Sarandon totally made that video...

joker
03-22-11, 01:47 AM
Happy Mothers Day. ;) Sorry to hijack this thread.

rgathright
03-22-11, 05:52 AM
@ Joker,

I suppose that's actually correct. Not as many GPU crunchers around and a short supply of GPU projects to choose from.

From what I've personally seen however, is that CPU crunching seems to be a failing means to a worthy cause. The credits are so low for CPU crunching that it seems to me to be almost worthless, though I do see those Aqua credits as something to take note of. Now if the rest of the projects threw out credit as well as the GPU projects did, well guess what? There would be an even playing field throughout the land and it wouldn't matter if you were CPU or GPU crunching, it would all be the same crunching.

I have to admit I'm a credit whore and therefore is the reason I joined up to TopGun. Not for any project in particular, but for the most credit a project is offering. And if the project a person chooses to run is what suits them, it is not my call to say anything about it, they will do what they like and that is fine, it all adds up as long as we're all on the same TEAM!

Yo-

I know that my strategy with CPU power is futile in the face of better credits from GPU power.

However, I am a SQL Server/.Net Coder/Dynamics AX programming whore that is always looking for ways to increase the performance of my company's data systems. GPU's cannot crunch SQL Server stored procedures. No, LINQ will not do that or drive a car. :p

Dandasarge
03-22-11, 10:49 AM
I know that my strategy with CPU power is futile in the face of better credits from GPU power.

However, I am a SQL Server/.Net Coder/Dynamics AX programming whore that is always looking for ways to increase the performance of my company's data systems. GPU's cannot crunch SQL Server stored procedures. No, LINQ will not do that or drive a car. :p

Well why didn't you start off with that :)

YoDude9999
03-22-11, 09:44 PM
@ OP, Suddenly, things make more sense. Go for the most you can afford. I would look into AMD however as they are usually much less expensive in the long run. Stick with name brands and look for lifetime warranties.

@ Joker, I hate RAP, but for once I found THAT video to be a hit! That makes two I actually like.

zombie67
03-22-11, 10:58 PM
From what I've personally seen however, is that CPU crunching seems to be a failing means to a worthy cause. The credits are so low for CPU crunching that it seems to me to be almost worthless, though I do see those Aqua credits as something to take note of. Now if the rest of the projects threw out credit as well as the GPU projects did, well guess what? There would be an even playing field throughout the land and it wouldn't matter if you were CPU or GPU crunching, it would all be the same crunching.

There is no such thing as cross-project parity. Not only does it not exist, it is impossible for it to exist at all. I won't bore everyone with the reasons why (again).

But this is the very reason I use MegaMilestones to measure performance, instead of total credit. Projects can pay differently, GPU or CPU or both, optimized apps...none of it matters. MMs level the playing field equally. To get (say) the most 10k MMs, you have to crunch ALL projects to the same amount. The good paying AND the bad. It also has the side benefit of rewarding crunching for a variety of projects, not just the popular/famous ones (cough S@H).

joker
03-23-11, 12:53 AM
To bad they dont base the number one spot on MMs instead of total credits.

zombie67
03-23-11, 01:27 AM
To bad they dont base the number one spot on MMs instead of total credits.

I *do* base the number one spot on MMs. Who is "they"?

joker
03-23-11, 01:54 AM
As far as I know, with my limited knowledge, is that all stat sites gauge the #1 spot on max credits. That is who I would call "they". I am saying that it is too bad they "they" dont look at MMs to figure out the "top spot". Time to start your own stat site! And BTW, I am trying to be funny and not an a$$hole.

zombie67
03-23-11, 02:37 AM
As far as I know, with my limited knowledge, is that all stat sites gauge the #1 spot on max credits. That is who I would call "they". I am saying that it is too bad they "they" dont look at MMs to figure out the "top spot". Time to start your own stat site! And BTW, I am trying to be funny and not an a$$hole.

No need to start my own stat site. Bok already lists MM ranking, both personal and team. See, it's not up to the stats site to decide. It's up to *us* crunchers to decide which stats are meaningful. I choose MMs. Others choose total credits. Still others choose Formula BOINC. There is no "right" or "agreed" stat for what "#1" means.

Back when there was only one measuring stick, heck, even only one project, it was obvious. But ever since a second project was created, cross project totals are crap.

And of *course*, I know. Funny is the primary objective for any joker!

http://3.bp.blogspot.com/_Llxf0Z81zHA/SeRfDyNdXsI/AAAAAAAAFyA/ARMSzwkaMQI/s1600/gAlexRossJokerHarleyQuinn.jpg

DrPop
03-24-11, 12:30 PM
There *could* be a fairly simple way to do what Joker is proposing, though. *IF* the "credit ratio" of all projects were locked in.
It would simply be based off 1 project - say SETI or MW or something that's been around a long time.
Then each project's credits would be a (+) or (-) a certain % of that credit score. In other words, it would be a WEIGHTED total.
So 1MIL credits at AQUA might be worth 600K WEIGHTED, while 1MIL credits at Einstein could be worth 1.4MIL WEIGHTED.
IF the projects would simply supply the data on what it take to get a credit, and IF that could be a mathematical constant, then it's just a simple formula to work out which team's got the most computing power in the world.

zombie67
03-24-11, 08:48 PM
There *could* be a fairly simple way to do what Joker is proposing, though. *IF* the "credit ratio" of all projects were locked in.
It would simply be based off 1 project - say SETI or MW or something that's been around a long time.
Then each project's credits would be a (+) or (-) a certain % of that credit score. In other words, it would be a WEIGHTED total.
So 1MIL credits at AQUA might be worth 600K WEIGHTED, while 1MIL credits at Einstein could be worth 1.4MIL WEIGHTED.
IF the projects would simply supply the data on what it take to get a credit, and IF that could be a mathematical constant, then it's just a simple formula to work out which team's got the most computing power in the world.

And how do you account for optimized apps? Or when the project itself releases a faster app? And how do you account for projects that are disk/IO bound, or have other special requirements? How do you compare CPU to GPU credits when a GPU project has no CPU app?

Edit: Rather than me trying to explain why cross-project parity is impossible, and doing it *poorly*, here is POV doing it right:

http://www.boinc-wiki.info/User:Nicolas/Credit_scenarios

joker
03-24-11, 09:50 PM
This has gotten all to complicated for me so I am going to go get some ice cream.

Maxwell
03-24-11, 09:55 PM
This has gotten all to complicated for me so I am going to go get some ice cream.
Beer is more effective...

joker
03-24-11, 09:59 PM
Beer is more effective...

Ok Max man, you talked me into it! <:-P \:D/ :(( :-& ...a good night.

DrPop
03-27-11, 04:01 AM
Yeah...that's why I had a big IF in there. We can't seem to account for all those variables - heck, I don't even think the project admins know them half the time. So we're stuck with a lame way of doing things, with an overall BOINC score, so I'll be hitting some high-ish paying projects so SIC doesn't kick our hind end quite so fast. ;)