Intel has announced, that some of the chip-sets for sandy-bridge are faulty. So one should wait some time before buying these MB's.
I have only a german explanation, but I'm shure, you can find it also in your language.

Um was geht es?

Die Intel 6 Series Express Chipsätze für Sandy Bridge und Intel Xeon C200 Series, namentlich H67, P67 und die Mobile-Chipsätze HM65, HM67, QM67, QS67, UM67 der ersten Generation können über die Zeit im Gebrauch einen Defekt am SATA-II-Controller aufweisen. Alle Mainboards mit Cougar-Point-Chipsätzen, die seit dem 9. Januar ausgeliefert wurden, werden nach derzeitigem Kenntnisstand umgetauscht werden müssen.

Das Problem

Die SATA-2-Ports – genauer die Ports 2 bis 5 mit 3 GBit/s (3G) – können mit der Zeit instabil arbeiten und im schlimmsten Fall vollständig ausfallen. Daran angeschlossene Festplatten oder DVD-Brenner können unzuverlässig arbeiten und bei einem Defekt des SATA-2-Ports nicht mehr angesprochen werden. Die SATA-6G-Ports (6 GBit/s) sind von diesem Defekt laut Intel nicht betroffen.

---

As per: http://boinc.berkeley.edu/wiki/SourceCode
Sources were obtained via

svn co http://boinc.berkeley.edu/svn/trunk/boinc

BTW, this link will give you BOINC source code, not SETI sources. I'm quite sure that BOINC performance will not benefit much from any SIMD instructions used, better try to aquire SETI scientific application code.

The problem with the Sandy Bridge (p67 and h67) is the intel SATA 3G ports. If you have ports from Marvell or some other manufacturer, it's wise to move your hard drive (and DVD drive, if applicable) to that SATA controller. Or you could use the Intel SATA 6G ports, which are unaffected by the issue. If you own a gigabyte board, they've already stepped up and set up a replacement program, I am sure others are to follow. Intel put aside $700 million to fix the problem, so I assume that everyone who has a faulty board will get a replacement. If we're lucky, they'll upgrade us to the (unreleased as of this post) Z67 board, which has additional features that didn't make it onto the p/h 67 boards.

My motherboard/cpu was delayed in Iowa due to the massive snow/ice storm that slammed into the Midwest yesterday. Hopefully it will be here tomorrow so I can also help with the testing.

---

-baron_iv
Proud member of:
GPU Users Group

As per: http://boinc.berkeley.edu/wiki/SourceCode
Sources were obtained via

svn co http://boinc.berkeley.edu/svn/trunk/boinc

BTW, this link will give you BOINC source code, not SETI sources. I'm quite sure that BOINC performance will not benefit much from any SIMD instructions used, better try to aquire SETI scientific application code.

That's true, but outlaw will need the BOINC sources to build libraries.

The AK_v8 sources in the download section at Lunatics are from the first version. What outlaw needs is an export of AK_v8\trunk at revision 76 to match what our AK_v8b SSE2+ CPU builds are, and there haven't been any changes in trunk since then. Although I could do it, my dial-up would make it inconvenient.

                                                                Joe

As per: http://boinc.berkeley.edu/wiki/SourceCode
Sources were obtained via

svn co http://boinc.berkeley.edu/svn/trunk/boinc

BTW, this link will give you BOINC source code, not SETI sources....

That's true, but outlaw will need the BOINC sources to build libraries.
...

Are you guys sure ? Both that checkout, and the following,

http://setiweb.ssl.berkeley.edu/sah/seti_source/nightly/

get me a "seti_boinc" project, aka a "setiathome_6.11_windows_intelx86.exe"

Thanks!

---

You will need both BOINC sources, as described in the BOINC Wiki, and SETI sources, as described on Porting and optimizing SETI@home.

The

svn checkout https://setisvn.ssl.berkeley.edu/svn/seti_boinc

line is current, as far as I know, but note

If you are looking for the SETI@home CUDA source, it is currently located at https://setisvn.ssl.berkeley.edu/svn/branches/seti_cuda. It will eventually be merged with the main source branch.

I'm not sure whether 'eventually' has arrived yet.

Ok, I have an initial VS2010 optimized with AVX 32-bit build using

svn co http://boinc.berkeley.edu/svn/trunk/boinc

here:

http://superconfigure.wordpress.com/2011/02/04/advanced-vector-extensions/

I don't have to tell you guys how to setup the app_info.xml file obviously.

I'll keep working on the 64-bit version.

Of interest is the +- WU changes on the same Win7Sp1 Sandy Bridge system using the stock seti build vs this AVX enabled one....

Normally I'd be the 1st line tester but in this case only a select few *cough Todd* have the hardware...

---

Right now all I have is a system running the x64 version of Win7, but next week I should have my other cpu in hand. Should have been here today but got delayed by the storm I think. Haven't been back to the office yet-it may be there.

Todd

---

Is there any interest in a linux version? I think I can handle doing that one, because I'm FAR more competent on linux than windows. My sandy bridge system came in this evening, installing all the components now. Should have it up and running in a little while.

Linux has supported AVX since Kernel 2.6.30 (June 2009) and it's fully supported in GCC 4.5 (optional support in 4.4 too). Figures that linux would be light years ahead of Windows. ;)

I am not aware of GPU support on Linux, so the linux port may not even be useful for those who use their CPUs in combination with GPUs. If I had to guess, NVidia will be supported, ATI will not, but that's pure speculation on my part. Either way, that has no bearing on the CPU AVX support, so if there's interest, I'll fire up linux and make a build sometime this weekend.

Outlaw, I will test your build ASAP and let you know the result. Thank you for the hard work on the build.

---

-baron_iv
Proud member of:
GPU Users Group

Be prepared with better cooling, when using AVX.
With non-AVX Linpack libraries my new 2500K system reaches max CPU cores temperature of 66 degress, overclocked to 4,7 GHz(1,29 V) on air(Noctua U12 + Noiseblocker XLP). When I ran AVX version of Linpack, temperature increased to 79 degrees.

---

Is there any interest in a linux version? I think I can handle doing that one, because I'm FAR more competent on linux than windows. My sandy bridge system came in this evening, installing all the components now. Should have it up and running in a little while.

Linux has supported AVX since Kernel 2.6.30 (June 2009) and it's fully supported in GCC 4.5 (optional support in 4.4 too). Figures that linux would be light years ahead of Windows. ;)

I am not aware of GPU support on Linux, so the linux port may not even be useful for those who use their CPUs in combination with GPUs. If I had to guess, NVidia will be supported, ATI will not, but that's pure speculation on my part. Either way, that has no bearing on the CPU AVX support, so if there's interest, I'll fire up linux and make a build sometime this weekend.

Outlaw, I will test your build ASAP and let you know the result. Thank you for the hard work on the build.

The Linux Fermi build is getting up to speed over at Lunatics.

---

Be prepared with better cooling, when using AVX.
With non-AVX Linpack libraries my new 2500K system reaches max CPU cores temperature of 66 degress, overclocked to 4,7 GHz(1,29 V) on air(Noctua U12 + Noiseblocker XLP). When I ran AVX version of Linpack, temperature increased to 79 degrees.

Dang, those new SB processors overclock like crazy! 1.29V for 4.7ghz is absolutely incredible. Sounds like you should move up to water-cooling though. Personally, I don't run my CPUs over 70C for any length of time, but I know others who have ran them at higher temps for long periods of time without any issues. 4.7ghz on 1.29v...AMAZING.

I really like the fact that it increases temps like that, it means that the app is really stressing the CPU. So that's great news.

---

-baron_iv
Proud member of:
GPU Users Group

2500K @ 4.7 ghz is impressive, indeed. And for scientific advancement, of course.
Were we but Gamers, to what height would we reach? Found this for your (my) titilation:

NRG-TORNADO-i7S Intel Core i7-2600k Quad Core Overclocked up to 5.00GHz, Liquid Cooled.

Excepting it should read six cores.

Cheers

Excepting it should read six cores.

Why?
The i7 26xx series are Quad Core (with HyperThreading).

---

Grant
Darwin NT

I've been modest with overclock. My specimen can do 5 GHz with same cooling, LinX stable(then SETI stable too) using 1.4V vCore voltage. But it needs something better than old 4-pipe Noctua cooler, temperature reaches 85 degrees C. Something like Megahalems, Archon, Armageddon, Silver Arrow etc., will do the job, definitely.
Here it is(tested with non-AVX Linpack)
I just dont need those 300 MHz, they would come at additional price in consumption and cooling which I don't like to pay. This processor is not meant for crunching, it just feeds CUDA devices :)
2600K are similar, they don't overheat more bcs of hyper-treading, in difference with old i7 s.1366 CPUs, where turning HT off leads to temperature decrease with 10-15 degrees(when overclocked).

It's offtopic, but I'd like to share here my research over Sandy Bridge CPUs last month. It's not very complete, 2500K and appropriate motherboards were hard to find from local reseller here in Bulgaria.
First of all, there is very responsive and cheap motherboards from AsRock. I picked AsRock P67 Extreme4, bcs it features 3x PCIe x16 slots(2 x8 and 1 x4, all 2.0). It costs me 125 EURO(converted from our local currency. We are EU members, but not using EURO yet), which I consider very good price. Any better MoBo, as CUDA host, cannot be found - all three PCIe x16 MoBos from other manufacturers are far more expensive.
Here are some interesting numbers. I have set of 3 MAR units and one VLAR unit(no difference in crunch times), which I use as comparison base.
On one of my 920 systems, I turned HT off just to compare it to 2500K CPU, using Lunatics SSE4.1 client.
920 is clocked at 4040 MHz and average completion time is 75 minutes for a WU. With 2500K I got very interesting results. It turns ot memory makes HUGE difference. 2500K@4040 did does units in 64 minutes with memory clocked at 1866 9-9-9-27, while 2500K@4500 GHz did them in 69 minutes, with memory at 1066 7-7-7-20.
Those memory dependency would apply with more power on 2600K systems, where HT is present.
Another interesting thing to know is, that 2600K CPUs, as rule, overclock better and uses just a tiny bit more electricity and got just a 2-4 degrees more warm than their comparative 2500K brothers.
Judging by the numbers, 2500K@4.5 can do a little worse than GTX 460@800/1600/2000 for just about same power consumption(approx 100 watts over idle, but depends on particular specimen). 460 can do 5-6 units per hour, 2500K can do 4, roughly said, which is not bad. With AVX it may do twice faster, IMHO.
Quantity of GFlops done in with Linpack with AVX is twice bigger than a non-AVX Linpack.

---

Here http://lunatics.kwsn.net/1-discussion-forum/akv8b-rev76-sources.msg35571.html;topicseen#msg35571 link to most current AKv8b sources (fastest CPU SETI MB app so far).

Here http://lunatics.kwsn.net/1-discussion-forum/akv8b-rev76-sources.msg35571.html;topicseen#msg35571 link to most current AKv8b sources (fastest CPU SETI MB app so far).

Is this version newer than what is included in Lunatics unified installer v0.37? I tried downloading the file, however, it takes me back to the 10 second waiting period in a never ending cycle. Is there any other location that I can download from?
______________

Here http://lunatics.kwsn.net/1-discussion-forum/akv8b-rev76-sources.msg35571.html;topicseen#msg35571 link to most current AKv8b sources (fastest CPU SETI MB app so far).

Is this version newer than what is included in Lunatics unified installer v0.37? I tried downloading the file, however, it takes me back to the 10 second waiting period in a never ending cycle. Is there any other location that I can download from?

It's not application but sources to build last released Lunatics SETI MB app.
I provided sources for someone who would like to re-build app with AVX-enabled compiler and has no access to Lunatics internal SVN server.

Here http://lunatics.kwsn.net/1-discussion-forum/akv8b-rev76-sources.msg35571.html;topicseen#msg35571 link to most current AKv8b sources (fastest CPU SETI MB app so far).

Is this version newer than what is included in Lunatics unified installer v0.37? I tried downloading the file, however, it takes me back to the 10 second waiting period in a never ending cycle. Is there any other location that I can download from?

It's not application but sources to build last released Lunatics SETI MB app.
I provided sources for someone who would like to re-build app with AVX-enabled compiler and has no access to Lunatics internal SVN server.

And it downloaded absolutely fine - no 10-second wait here.

Is there any interest in a linux version?

Definitely! I'm getting a new SB soon to replace my aging colo q6600 webserver, which also uses 3 cores for crunching. As the elec is built into the rental, I would give it a couple of gpus too if there were any decent linux gpu apps.

I got a big i7 to replace it last year, but it became too useful at home!