8-core cruncher?

Message boards : Number crunching : 8-core cruncher?

To post messages, you must log in.

1 · 2 · 3 · 4 · Next

AuthorMessage
Profile Paydirt
Avatar

Send message
Joined: 10 Aug 06
Posts: 127
Credit: 960,607
RAC: 0
Message 35387 - Posted: 23 Jan 2007, 14:41:44 UTC
Last modified: 23 Jan 2007, 14:42:16 UTC

OK, I've done a decent amount of research on this. This build would be more expensive PER MACHINE, but less expensive per credit (or per processor power).

You could build a 8-Core, dual four-core Xeon 1.6 GHz machine for $1300 +case +HD +PSU. My guess is that it would be around 2.66 the power of a E6600 build at less than 2.66 the cost (if you bought 2.66 E6600 builds to each E5310 build).

($696) 2 Intel Xeon Quad-Core E5310 1066 MHz 8MB L2 cache CPU
($298) Gigabyte GA-7VCSV Dual Xeon Server Motherboard
($288) 2 GBs Kingston FB-DIMM RAM. (must get FB-DIMM RAM with this mobo)
($1282) total, plus case & HD & PSU.

Even with something like Rosetta@Home which uses around 120k RAM per instance, you would still have enough.

As far as CPU crunching, does anyone think there would be something cheaper for processing power/dollar spent (less than $2000)? This build would have eight 1.6 GHz processors.
ID: 35387 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 35388 - Posted: 23 Jan 2007, 14:51:44 UTC - in response to Message 35387.  
Last modified: 23 Jan 2007, 14:53:51 UTC

Am gonna luv the day when an 8 core unit is considered "standard".

Not a bad price to build something like this, just don't forget the Windoze tax! Two cpu's, two o/s licenses?

I understand that Linux is free, but I believe that for purposes of Rosetta crunching, Windoze performs better than Linux.

($696) 2 Intel Xeon Quad-Core E5310 1066 MHz 8MB L2 cache CPU
($298) Gigabyte GA-7VCSV Dual Xeon Server Motherboard
($288) 2 GBs Kingston FB-DIMM RAM. (must get FB-DIMM RAM with this mobo)
($1282) total, plus case & HD & PSU.

ID: 35388 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Paydirt
Avatar

Send message
Joined: 10 Aug 06
Posts: 127
Credit: 960,607
RAC: 0
Message 35389 - Posted: 23 Jan 2007, 14:59:13 UTC

Can you point me to anything that benchmarks Windows vs Linux BOINC crunching?

Is it that the linux client simply is not as optimized?
ID: 35389 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 35390 - Posted: 23 Jan 2007, 15:02:38 UTC - in response to Message 35389.  

Just anecdotal from reading the message boards

Can you point me to anything that benchmarks Windows vs Linux BOINC crunching?


Maybe this is it

Is it that the linux client simply is not as optimized?

ID: 35390 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Paydirt
Avatar

Send message
Joined: 10 Aug 06
Posts: 127
Credit: 960,607
RAC: 0
Message 35391 - Posted: 23 Jan 2007, 15:03:32 UTC

It should be noted that the 4-core Xeon E5310s are a little hard to get a hold of. They range in price from $350 to $450.

I wonder what their availability will be like in three months.
ID: 35391 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1829
Credit: 114,335,292
RAC: 51,548
Message 35393 - Posted: 23 Jan 2007, 15:11:34 UTC - in response to Message 35388.  

Not a bad price to build something like this, just don't forget the Windoze tax! Two cpu's, two o/s licenses?


XP Pro or Win2K/2K3 will all run dual-socket with one licence.

I don't there's much difference between the linux and windows performance, but i might be wrong. I know the benchmarks are much lower in Linux, but that's pretty meaningless.

I think the cheapest way would be to get the E6300s and overclock them - they can usually run at over 3.2GHz so two of these would match your 8 core monster for much less money. They don't have as much cache as the xeon i believe (2mb vs 4mb), but i think once you've got 1MB per core then the increase makes little difference to the (current) Rosetta models.

And for a cruncher a case is an unnecessary luxury! ;)
ID: 35393 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 11 Feb 06
Posts: 316
Credit: 6,589,590
RAC: 2,906
Message 35398 - Posted: 23 Jan 2007, 15:50:33 UTC - in response to Message 35387.  

($288) 2 GBs Kingston FB-DIMM RAM. (must get FB-DIMM RAM with this mobo)


You will need 4x DIMMs to get the best performance out of it. There was a thread over on SETI (I think), where someone only had two DIMMs, and the performance was lousy. He added two more to take advantage of dual channel (I think I't called) bandwidth, and his performance shot up.

Also, you can OC these fairly easily (although I haven't done that myself yet). It requires covering a couple pins with very small pieces of electrical tape.
Reno, NV
Team: SETI.USA
ID: 35398 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Paydirt
Avatar

Send message
Joined: 10 Aug 06
Posts: 127
Credit: 960,607
RAC: 0
Message 35403 - Posted: 23 Jan 2007, 17:19:47 UTC

Zombie, you are correct. I found the thread on the SETI boards. It DOES significantly utilize the 4 memory banks.

2 FB-DIMMs vs 4 FB-DIMMs

There's a nice reply with a picture toward the bottom of the thread.


ID: 35403 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Who?

Send message
Joined: 2 Apr 06
Posts: 213
Credit: 1,366,981
RAC: 0
Message 35411 - Posted: 23 Jan 2007, 19:45:31 UTC - in response to Message 35403.  
Last modified: 23 Jan 2007, 19:55:19 UTC

Zombie, you are correct. I found the thread on the SETI boards. It DOES significantly utilize the 4 memory banks.

2 FB-DIMMs vs 4 FB-DIMMs

There's a nice reply with a picture toward the bottom of the thread.



If i was you, i ll buy a faster Quad core CPU and wait 6 months to buy the 2nd one, after your wallet recover. You do not need to buy the 2 CPUs the same day.
Check my V8 machine, it is what you are trying to build :)


who?
ID: 35411 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Paydirt
Avatar

Send message
Joined: 10 Aug 06
Posts: 127
Credit: 960,607
RAC: 0
Message 35448 - Posted: 24 Jan 2007, 15:08:24 UTC

Maybe newegg.com doesn't have them, but are there dual socket 775 motherboards?
ID: 35448 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
MikeMarsUK

Send message
Joined: 15 Jan 06
Posts: 121
Credit: 2,637,872
RAC: 0
Message 35449 - Posted: 24 Jan 2007, 15:14:15 UTC
Last modified: 24 Jan 2007, 15:47:38 UTC

How do you get these things to overclock? Something I saw on the web seemed to imply you need a 5000x chipset (workstation), but it wasn't very clear. Bios support?

PS I looked at the UK prices for a system with 4 x 1GB FB-Dimms, and 5320 (couldn't find 5310), and it was much more expensive, nearly double the price at current exchange rates.

ID: 35449 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 11 Feb 06
Posts: 316
Credit: 6,589,590
RAC: 2,906
Message 35454 - Posted: 24 Jan 2007, 17:25:02 UTC - in response to Message 35448.  

Maybe newegg.com doesn't have them, but are there dual socket 775 motherboards?

As far as I know, 775 chips do not work in tandem like that. You must use 771 socket chips for multi-chip applications.
Reno, NV
Team: SETI.USA
ID: 35454 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
zombie67 [MM]
Avatar

Send message
Joined: 11 Feb 06
Posts: 316
Credit: 6,589,590
RAC: 2,906
Message 35455 - Posted: 24 Jan 2007, 17:29:32 UTC - in response to Message 35449.  

How do you get these things to overclock? Something I saw on the web seemed to imply you need a 5000x chipset (workstation), but it wasn't very clear. Bios support?


As far as I know, there are no 5000* chip-set boards that support OC via BIOS. You have to do a fairly simple HW hack, involving covering a pin or two on the socket with a very small piece of electrical tape.

The process is detailed over on XtremeSystems.

No, I haven't tried it myself...yet.
Reno, NV
Team: SETI.USA
ID: 35455 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 35463 - Posted: 24 Jan 2007, 19:44:32 UTC - in response to Message 35455.  
Last modified: 24 Jan 2007, 19:46:58 UTC

@Who?: Quad-core chips are a present day reality. Some mfgrs are talking about 80-core chips in the next 5-7 yrs.

In your own personal opinion, not representing that of your employer, what can we expect in the next 1-2 years? In the next generation on the drawing board, will density double to bring us to octo-core (ala Cell BE), or will it more than double, to put us on the path to 80-core by 2011?
ID: 35463 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Who?

Send message
Joined: 2 Apr 06
Posts: 213
Credit: 1,366,981
RAC: 0
Message 35471 - Posted: 25 Jan 2007, 0:23:52 UTC - in response to Message 35463.  

@Who?: Quad-core chips are a present day reality. Some mfgrs are talking about 80-core chips in the next 5-7 yrs.

In your own personal opinion, not representing that of your employer, what can we expect in the next 1-2 years? In the next generation on the drawing board, will density double to bring us to octo-core (ala Cell BE), or will it more than double, to put us on the path to 80-core by 2011?


sorry, i can't answer this, I am involve in the many core stuff ;)

who?
ID: 35471 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mats Petersson

Send message
Joined: 29 Sep 05
Posts: 225
Credit: 951,788
RAC: 0
Message 35486 - Posted: 25 Jan 2007, 12:01:47 UTC

I'm probably also too close to the work of these things to actually answer the question to Who?, but if we just look at it realistically, it takes roughly two-three years to go from "architecture specification" to "finished chip" in any chip manufacturer.

Today, there are dual and quad-core chips in existance from AMD and Intel.

To make a 80-core chip, you'd have to completely re-architect the chip architecture (or use roughly twenty times the current size die, which is unreasonable even if we ignore the large caches that take up some 50-70% of the die-size of current dual/quad-core chips).

If we re-architect the chip to such an extent that the core-size is SIGNIFICANTLY smaller, it would most likely impact software model that the processor can use, so it would not be compatible with the current architecture, and thus wouldn't allow you to use current software.

2011 is only 4 years away - that's not so long that the software of today (and the next couple of years) will be completely useless.

Intel already tried to push a new architecture to the market (Itanium), and I'd say that this has not been the success that Intel wished for. To achieve a new architecture's success, it needs to be sufficiently better for new applications, and at the same time comparable for current/older applications. If it's not compatible (enough), it won't succeed. Many manufacturers have come out with "better" products, but unless they are also compatible with the major markets of the business, they are not successfull.

The 80-core architecture is probably a fictive device. My mum has a book from the late 50's/early 60's that I used to read when I was little. It showed the future of our lives, including space-travel and personal aircrafts, in the next couple of decades. If you look out the windows (now more than 4 decades later), you'll notice the personal aircrafts are not flying around in the towns or countryside outside... ;-) It just goes to show how hard it is to foresee how fast the development is going to be in the decades ahead....

--
Mats
ID: 35486 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 35504 - Posted: 25 Jan 2007, 16:40:38 UTC - in response to Message 35486.  

Ok. Which "rumor mills" or blogs (The Inquirer ?) have a half decent track record? I can't find much info as to what is currently on the drawing boards. Just curious if the magic number is likely to be 8, or will it likely be some other number. Just want to be pointed in the correct direction as to where to look for this info.


I'm probably also too close to the work of these things to actually answer the question to Who?, but if we just look at it realistically, it takes roughly two-three years to go from "architecture specification" to "finished chip" in any chip manufacturer.

Today, there are dual and quad-core chips in existance from AMD and Intel.

ID: 35504 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
MikeMarsUK

Send message
Joined: 15 Jan 06
Posts: 121
Credit: 2,637,872
RAC: 0
Message 35507 - Posted: 25 Jan 2007, 17:20:02 UTC


Inquirer's good, but they don't have a magic mirror either! :-)

Powers of two is a good bet, the main factor is surface area of a core versus how much silicon it's economically practical to put into a single package.

So each time the process shrinks, you'll get another maximum number of cores per package.

Considering x86 cores only - 4 in 65nm (now), 8 in 45nm (volume in 2008), possibly 12 or 16 in 32nm (volume in 2010??), ... etc etc.

Sun has already managed to get a huge number of cores onto a single chip, but it's not x86.


ID: 35507 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Paydirt
Avatar

Send message
Joined: 10 Aug 06
Posts: 127
Credit: 960,607
RAC: 0
Message 35509 - Posted: 25 Jan 2007, 20:30:43 UTC

You're definitely asking the right people, they just can't answer the question.

I read something on dailytech.com about how they tested a wafer and it could be a reality (80 cores) by 2011. That would be awesome, especially for science! I think the tech blog concluded that it may be possible to have mass production teraflop-on-a-chip by 2011.


ID: 35509 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
The_Bad_Penguin
Avatar

Send message
Joined: 5 Jun 06
Posts: 2751
Credit: 4,271,025
RAC: 0
Message 35514 - Posted: 25 Jan 2007, 21:37:30 UTC - in response to Message 35507.  
Last modified: 25 Jan 2007, 21:42:43 UTC

Timing is everything! Just found this on The Inquirer.

Inquirer's good, but they don't have a magic mirror either! :-)


Talking about AMD:

"Thereafter, two and four-way systems will pop up like magic mushrooms on a damp autumnal field as the focus will shift to 16-core systems."

So it may be that we are increasing by squaring, instead of powers of two... (ok, very wishful thinking, lol !)

Powers of two is a good bet, the main factor is surface area of a core versus how much silicon it's economically practical to put into a single package.



ID: 35514 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · 3 · 4 · Next

Message boards : Number crunching : 8-core cruncher?



©2024 University of Washington
https://www.bakerlab.org