Building a 32 thread xeon system doesn't need to cost a lot

Message boards : Number crunching : Building a 32 thread xeon system doesn't need to cost a lot
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 12 · Next

AuthorMessage
Al Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Avatar

Send message
Joined: 3 Apr 99
Posts: 1682
Credit: 477,343,364
RAC: 482
United States
Message 1793523 - Posted: 4 Jun 2016, 23:01:26 UTC - in response to Message 1793509.  

Hmm, that looks somewhat familiar, I think I need to do a little digging around here, I might just luck out and have something similar. If not, thanks for the link, I'll hit Microcenter tomorrow worst case, I'm sure that they have something like it, sadly they are pretty much the only real computer store left standing around here today. Amazing how many have fallen by the wayside over the years. But then, I'd dating myself, I mean who uses _computers_ these days, right? That is so 1990's!?! lol

ID: 1793523 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13727
Credit: 208,696,464
RAC: 304
Australia
Message 1793537 - Posted: 4 Jun 2016, 23:39:12 UTC - in response to Message 1793424.  

It is 13c now and temperatures have dropped to 70-80c. It can get as hot as 25c in the afternoon. It will get warmer as summer hits.

If the ambient temperature was 35°c or higher then i'd expect CPU temps around 80°c or more, but to my way of thinking, even with the standard 1U server case heatsink, with the ambient that low I would have expected the CPU temperatures to be lower, mid 60°s at most.


The air coming out the back is toasty warm.

Only toasty warm?
If the CPU is at more than 75°c i'd expect that air to be hot, very hot. Not too hot to hold your hand there, but still damn hot.


Removing the air baffel causes the fans to ramp up as the temperatures rise.

With the lid off, how hot is the CPU heatsink to touch- be careful. If the heatsink is fitted correctly and the CPU is over 70°c or more there's a good chance you could burn yourself.
If the CPU is at that temperature, but the heatsink only feels warm, it's not fitted correctly (heat isn't transferring from the CPU to the heatsink very well).


I was going to replace the 1u heat sinks with full size LGA1366 heat sinks except the mounting holes are different.

Better heatsinks & fans would certainly make a big difference.
Grant
Darwin NT
ID: 1793537 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1793543 - Posted: 4 Jun 2016, 23:51:10 UTC - in response to Message 1793523.  

Al, many hardware stores sell multi bit sets with an assortment of sizes, I just checked my $5 32 piece set and it goes down to T6.

Or tool stores as well.

And as a plus, you can put the 1/4 bit in your DeWALT to make the job go faster :))) Sorry just could resist that one :D
ID: 1793543 · Report as offensive
Profile Zalster Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 27 May 99
Posts: 5517
Credit: 528,817,460
RAC: 242
United States
Message 1793546 - Posted: 4 Jun 2016, 23:55:52 UTC - in response to Message 1793543.  
Last modified: 4 Jun 2016, 23:58:52 UTC

I wasn't going to do it but since you mentioned it....

https://youtu.be/SOSDbKaOksA

Don't be this guy!!

The wrong tools for the wrong job...
ID: 1793546 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13727
Credit: 208,696,464
RAC: 304
Australia
Message 1793554 - Posted: 5 Jun 2016, 0:15:33 UTC - in response to Message 1793546.  

I wasn't going to do it but since you mentioned it....

https://youtu.be/SOSDbKaOksA

Don't be this guy!!

The wrong tools for the wrong job...

It used to be necessary to do an apprenticeship to be a butcher. Now it looks like any one can do it, no training required.
That was painful to watch.
Grant
Darwin NT
ID: 1793554 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1793569 - Posted: 5 Jun 2016, 1:12:23 UTC - in response to Message 1793504.  
Last modified: 5 Jun 2016, 1:15:33 UTC

EXCELLENT! I will give that a shot, when looking at it, it looked sort of like a type of rivet or pressed stamping(which I thought was quite weird for something as delicate as a motherboard, but it is HP, so who knows?), as the holes didn't appear symmetrical, at least it seemed that way when I looked at it, so they must be pretty small torx bits. Any idea as to the size they take? I'd love to get that thing torn down and assembled tonite. Have to start digging for my torx set, though from what I remember, I may have to get some smaller bits, I think the smallest I have is T10 if I remember correctly, and I'd guess that that would be something between a T4 to T6? Thanks again for the great news.


Nope, not HP's socket, it's an Intel standard socket. With HP non-standard backplate.

You will have to figure out the size for the Torx bit; the ones in the sockets on the Z800 are larger than Torx 8, I remember, because I had a T8 driver that was too small...my Z800 is squirreled away somewhere as I "graduated" to a dual 2011 MB with 2 x E5-2670 for my new cruncher Big32. With 2 x GTX980 (ref) and Lunatics 0.44 looks like it will do around 30K RAC or better with 3 vlars on each of the GPUs.

But the "rivets" on top of the socket are the gizmos that hold the backplate; there are threaded sockets at the other end that take the threaded portions of the backplate (Z800 OR 1366 desktop).
ID: 1793569 · Report as offensive
Cruncher-American Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor

Send message
Joined: 25 Mar 02
Posts: 1513
Credit: 370,893,186
RAC: 340
United States
Message 1793570 - Posted: 5 Jun 2016, 1:19:31 UTC - in response to Message 1776402.  
Last modified: 5 Jun 2016, 1:20:05 UTC

I've got to say it's a tempting build if you don't mind the power bills.


My version of this with 2 x E5-2670 (v1) and 2 x ref GTX 980 draws 500-550 watts at the wall. My previous version (I7-4820K + 2 x GTX 980) drew 500 watts.
ID: 1793570 · Report as offensive
AMDave
Volunteer tester

Send message
Joined: 9 Mar 01
Posts: 234
Credit: 11,671,730
RAC: 0
United States
Message 1797797 - Posted: 21 Jun 2016, 22:34:20 UTC

Here is an AMD option for a 32c/64t box, available in ’17.

But wait one more year, and you can build a 48c/96t (shrunk to 7nm) box.

Or, you can opt for a lighter 22c/44t box.

But wait, there’s more . . . backed by Big Blue, it’s the

“KiloCore”

> 1000-core processor

> energy-efficient 621 million transistor

> can execute 115 billion instructions per second while dissipating only 0.7 Watts which means it can be powered by a single AA battery. The KiloCore chip executes instructions more than 100 times more efficiently than a modern laptop processor.

> it could manage 1.78 trillion instructions per second

Unfortunately, there is no release date as of yet.
ID: 1797797 · Report as offensive
Al Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Avatar

Send message
Joined: 3 Apr 99
Posts: 1682
Credit: 477,343,364
RAC: 482
United States
Message 1797967 - Posted: 22 Jun 2016, 16:11:12 UTC - in response to Message 1797797.  
Last modified: 22 Jun 2016, 16:12:47 UTC

Funny guy... :-p

Wait, that article wasn't published on April 1? Hmmm

ID: 1797967 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1798008 - Posted: 22 Jun 2016, 18:35:00 UTC - in response to Message 1797967.  

Funny guy... :-p

Wait, that article wasn't published on April 1? Hmmm


I have a few issues with their statement.
"The 1,000 processors can execute 115 billion instructions per second while dissipating only 0.7 Watts which mean it can be powered by a single AA battery."

Power dissipation is not the same as power consumption. A device could have a power input of 10w & dissipate 0.7w. Which would make it ~93% efficient.

With all of the power wires going into the device I imagine it might consume a fair bit more juice at full oomph.



I would guess that the press release information may have an error. It it likely the chip can execute 115 billion instructions per second & it may also be able to run with 0.7w of power input. I imagine it is not both at the same time unless it is some kind of RISC architecture running very specific instructions. Like those coin miners running hashes.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1798008 · Report as offensive
Al Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Avatar

Send message
Joined: 3 Apr 99
Posts: 1682
Credit: 477,343,364
RAC: 482
United States
Message 1798015 - Posted: 22 Jun 2016, 19:42:31 UTC - in response to Message 1798008.  

HAL, I'd have to agree with you after looking at that pic closer. It has four 24 pin power connectors going into it, and just one of those powers the most power hungry desktop CPUs made, though sometimes with a helper 4 or dual 4 pin for addl oomph. But still, the pic of the board it is attached to kind of flys in the face of what the article is preaching. Is it a multi 1000 watt system or isn't it? :-O

ID: 1798015 · Report as offensive
Claggy
Volunteer tester

Send message
Joined: 5 Jul 99
Posts: 4654
Credit: 47,537,079
RAC: 4
United Kingdom
Message 1798017 - Posted: 22 Jun 2016, 19:49:53 UTC - in response to Message 1797797.  

“KiloCore”

> 1000-core processor

> energy-efficient 621 million transistor

> can execute 115 billion instructions per second while dissipating only 0.7 Watts which means it can be powered by a single AA battery. The KiloCore chip executes instructions more than 100 times more efficiently than a modern laptop processor.

> it could manage 1.78 trillion instructions per second

Unfortunately, there is no release date as of yet.

http://vcl.ece.ucdavis.edu/pubs/theses/2015-1/AStillmakerDissertation.pdf

It's been mentioned on the parallella.org forum's:

Jar:
I've noticed a couple articles pop up recently. At the risk of being labelled a pedant, the "KiloCore" chip isn't the first one-thousand core chip. At least PEZY-SC beat them to that, I believe. Maybe others.

Regardless of the title, it looks like interesting work. It's low power for sorting data. It's hard to draw conclusions about other applications.

For those of you that think Epiphany is too easy to program, this may be the architecture for you! It has 128x35-bit (or is it 40-bit?) instructions. And 256x16-bit (512 bytes) data scratch memory. Plus some on-chip 64 kB banks of SRAM outside of the cores.


Claggy
ID: 1798017 · Report as offensive
Profile Jimbocous Project Donor
Volunteer tester
Avatar

Send message
Joined: 1 Apr 13
Posts: 1853
Credit: 268,616,081
RAC: 1,349
United States
Message 1802639 - Posted: 15 Jul 2016, 3:28:32 UTC

So, here's another way to get into Xeon on the cheap, if you have a bit of luck or some tech skills.
The 3ghz Core2Quad chip, Q9650 in LGA775, is still way over-priced due to demand for folks wanting to do an upgrade on Core2Duos on the cheap.
Turns out it's possible to modify a Xeon LGA771, e.g. the E5450 3Ghz Quad core, to run in an LGA775 motherboard, with some very specific caveats that must be met.
Rather than do the mod myself, I bought a modded Xeon for $41, and if that all works out nicely it will revive that box for less than half what it cost me to change out the Duo for a Core2Quad on my HP.
The wild thing is, unmodded E5450s are going on eBay for $12-15, the hardware for the mod costs under $5, so with a bit of work and luck it can be done for around $20. Not sure that can be beat. Better yet, where the Q9650 uses 95w, the E5450 is only 80, as long as you get the right one (e.g. SLBBM). Get the wrong one (e.g. SLANQ) and it sucks 120w (not good).
This only works on certain chipsets and motherboards, but there's an extensive list out there. Good info at http://www.delidded.com/. But rather than busting the tabs off the CPU socket, it's better to use a drill and make new notches on the CPU.
If anyone's interested, I'll keep you posted once I have the CPU installed.
ID: 1802639 · Report as offensive
Al Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Avatar

Send message
Joined: 3 Apr 99
Posts: 1682
Credit: 477,343,364
RAC: 482
United States
Message 1802655 - Posted: 15 Jul 2016, 5:43:33 UTC - in response to Message 1802639.  

Jim, I have a number of rigs I have put together with Intel socket 775 motherboards, and I went with the Quad-Core Xeon X3370 SLB8Z. I'm wondering how those compare to what you're using? The power is close - 80w vs. 90w, but there is no modification needed on them, I've just dropped them into the socket, and powered them up. Here are Intel's specs of the E5450 and the X3370, one thing that caught my eye when reviewing them both was the Recommended Customer Prices. When they were released, the X3370 was $339.00, and the E5450 was $969.00. The X3370 was released 9 months later, but that seems to be a pretty short period of time for that significant of a price drop, so I wonder what the reason of the price differences was? The only other difference I could see was the E5450 has FSB Parity. Not sure if this is that important to those of us crunching or not, I guess it would be nice to have the error correction included on the proc, but it sure seemed to add significantly to the costs.

ID: 1802655 · Report as offensive
Profile Jimbocous Project Donor
Volunteer tester
Avatar

Send message
Joined: 1 Apr 13
Posts: 1853
Credit: 268,616,081
RAC: 1,349
United States
Message 1802666 - Posted: 15 Jul 2016, 8:32:18 UTC - in response to Message 1802655.  

Jim, I have a number of rigs I have put together with Intel socket 775 motherboards, and I went with the Quad-Core Xeon X3370 SLB8Z. I'm wondering how those compare to what you're using? The power is close - 80w vs. 90w, but there is no modification needed on them, I've just dropped them into the socket, and powered them up. Here are Intel's specs of the E5450 and the X3370, one thing that caught my eye when reviewing them both was the Recommended Customer Prices. When they were released, the X3370 was $339.00, and the E5450 was $969.00. The X3370 was released 9 months later, but that seems to be a pretty short period of time for that significant of a price drop, so I wonder what the reason of the price differences was? The only other difference I could see was the E5450 has FSB Parity. Not sure if this is that important to those of us crunching or not, I guess it would be nice to have the error correction included on the proc, but it sure seemed to add significantly to the costs.

Good info, didn't realize that there were native LGA775 Xeons, or had forgotten. Guess I should have checked. The parity could actually be an issue, dunno how I missed that. Guess we'll see.
Other significant difference, besides 95 v 80w, is that the 55xx series support hyperthreading where the 33xx do not. That, to my mind, would justify the extra hassle, as I have noted gains in productivity crunching with HT enabled. Most siginificantly, that with HT on GPUs get supported better while still keeping the CPUs crunching.
There's also an issue of which chipsets will support the swapout, and in some cases which BIOS, as not all have Xeon microcode. Case in point, the Intel P45 seems to do fine supporting the 55xx, but the Q45 is a train wreck.
Finally, the 55xx are a bunch cheaper than the LGA775 33xx, primarily because of the upgrade path, I suspect.
The delidded site has a pretty decent list of mobos and the processors that have been tested with them.
ID: 1802666 · Report as offensive
Al Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Avatar

Send message
Joined: 3 Apr 99
Posts: 1682
Credit: 477,343,364
RAC: 482
United States
Message 1802677 - Posted: 15 Jul 2016, 11:55:54 UTC - in response to Message 1802666.  

Regarding the E5450 HT, I didn't know that because that would have been a huge miss for me when I was evaluating which procs to put into those boards, so I took a quick look at the Intel site again, and according to it:

Advanced Technologies
Intel® Turbo Boost Technology ‡ 		No
Intel® Hyper-Threading Technology ‡		No
Intel® Virtualization Technology (VT-x) ‡ 	Yes
Intel® VT-x with Extended Page Tables (EPT) ‡	No
Intel® 64 ‡					Yes
Idle States 					Yes
Enhanced Intel SpeedStep® Technology		Yes
Intel® Demand Based Switching			Yes
Thermal Monitoring Technologies 		Yes


looks like it doesn't support it either? Or am I not looking at the right thing?

ID: 1802677 · Report as offensive
Profile HAL9000
Volunteer tester
Avatar

Send message
Joined: 11 Sep 99
Posts: 6534
Credit: 196,805,888
RAC: 57
United States
Message 1802708 - Posted: 15 Jul 2016, 15:27:28 UTC - in response to Message 1802677.  

Regarding the E5450 HT, I didn't know that because that would have been a huge miss for me when I was evaluating which procs to put into those boards, so I took a quick look at the Intel site again, and according to it:

Advanced Technologies
Intel® Turbo Boost Technology ‡ 		No
Intel® Hyper-Threading Technology ‡		No
Intel® Virtualization Technology (VT-x) ‡ 	Yes
Intel® VT-x with Extended Page Tables (EPT) ‡	No
Intel® 64 ‡					Yes
Idle States 					Yes
Enhanced Intel SpeedStep® Technology		Yes
Intel® Demand Based Switching			Yes
Thermal Monitoring Technologies 		Yes


looks like it doesn't support it either? Or am I not looking at the right thing?

CPUs using the Intel Core microarchitecture do not support HT.

There are small LGA 771 to 775 adapters that allow using the less expensive Xeon CPUs in 775 boards. Which is great if you happen to have a load of old 775 boards that support Xeon CPUs. Most recently it is easier to find 771 boards vs compatible 775 boards.
SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[
ID: 1802708 · Report as offensive
Al Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Avatar

Send message
Joined: 3 Apr 99
Posts: 1682
Credit: 477,343,364
RAC: 482
United States
Message 1802730 - Posted: 15 Jul 2016, 17:07:06 UTC - in response to Message 1802708.  

Which is great if you happen to have a load of old 775 boards that support Xeon CPUs. Most recently it is easier to find 771 boards vs compatible 775 boards.

Lucky for me, I just so happen to be in that situation. :-)

ID: 1802730 · Report as offensive
Profile Jimbocous Project Donor
Volunteer tester
Avatar

Send message
Joined: 1 Apr 13
Posts: 1853
Credit: 268,616,081
RAC: 1,349
United States
Message 1802774 - Posted: 15 Jul 2016, 22:00:13 UTC - in response to Message 1802677.  

Regarding the E5450 HT, I didn't know that because that would have been a huge miss for me when I was evaluating which procs to put into those boards, so I took a quick look at the Intel site again, and according to it:

Advanced Technologies
Intel® Turbo Boost Technology ‡ 		No
Intel® Hyper-Threading Technology ‡		No
Intel® Virtualization Technology (VT-x) ‡ 	Yes
Intel® VT-x with Extended Page Tables (EPT) ‡	No
Intel® 64 ‡					Yes
Idle States 					Yes
Enhanced Intel SpeedStep® Technology		Yes
Intel® Demand Based Switching			Yes
Thermal Monitoring Technologies 		Yes


looks like it doesn't support it either? Or am I not looking at the right thing?

You were, I wasn't ... ;)
ID: 1802774 · Report as offensive
Profile Jimbocous Project Donor
Volunteer tester
Avatar

Send message
Joined: 1 Apr 13
Posts: 1853
Credit: 268,616,081
RAC: 1,349
United States
Message 1803866 - Posted: 21 Jul 2016, 5:06:17 UTC
Last modified: 21 Jul 2016, 5:13:09 UTC

So, for anyone curious, the Core2Duo to Xeon upgrade here went off nicely. Can't use BOINC as a comparison benchmark as yet, and other differences noted below also matter, but here are a couple comparisons:
Motherboard         HP-8000     Gigabyte GA-EP45-UD3L
Socket              775         775
Memory              8gb DDR3    4gb DDR2
Family              Core2Quad   Xeon
CPU                 Q9650       X5450, Socket 771 modified for 775 use
CPU Speed           3.00 ghz    3.00 ghz
Power (TDP)         95W         85W
Cores               4           4
Threads             4           4
Cache               12 mb       12 mb
Floating pt.        3538        3645 million ops/sec (BOINC benchmark)
Integer             7634        8317  "       "       "
CPU Cost (used)     $85         $45

So it would seem it was worth the risk, and after looking I think I could mod the CPU to accommodate the change from socket 771 to 775.
Was it worth it? Probably not, given $20 for a new Arctic cooler for the CPU, $10 for a better case fan, and probably $20 yet to come for better RAM and an increase to 8gb. Otoh, it's also one less motherboard in the landfill ...
ID: 1803866 · Report as offensive
Previous · 1 . . . 4 · 5 · 6 · 7 · 8 · 9 · 10 . . . 12 · Next

Message boards : Number crunching : Building a 32 thread xeon system doesn't need to cost a lot


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.