Search results

I would go with the Supermicro below. Good layout for CPUs and VRMs, and the Supermicro boards have more "control" than Tyan boards. Personally, I've always gone for the H8QGi-F variant of the board, more PCIe slots, and may have higher resale value (at nominal additional cost)...

For the 2u cases, one approach is to route the hoses out the back of each 2u case -- can probably drill some properly sized holes to make this very clean. Have the rad(s) on the outside of the cases. Given the load, you could even experiment with multiple systems connected in series to one...

I used these with my EK blocks & brackets: http://www.koolance.com/water-cooling/product_info.php?product_id=1039 http://www.koolance.com/water-cooling/product_info.php?product_id=1056 (I also had to get the G34 bolts for koolance CPU-370 blocks on another system).

Jeanjean, Did your EK blocks come with the small extension blocks/tabs that effectively extend each water block for G34 mounting? That's new if so, and is good information to share. I had to make brackets for my system which uses the Acetal version of those blocks.

Coming from an SR-2 at ~4.3ghz ddr3-1870 7-8-7, I saw no reduction in gaming performance going to a 4x 6180 system running at 2.7. Most of the games I play are GPU bound, not CPU bound, particularly at high resolutions. (2560x1600 middle monitor only everything maxed "ultra" or 7680x1600...

One is running at 2.8, 1.2125v, 7-7-7 memory, 15:40 TPF on 6904, 845W. One is running at 2.7, 1.175v, 6-7-5 memory, 16:00 TPF on 6904, ~750W. Tested, but don't really run at this since the power curve gets silly beyond 2.7: 2.9, 1.2625v, 7-7-7 memory, 15:13 TPF on 6904, 945W. At 2.8...

At the moment I'm not selling any of the block mounts that were pictured in the ATCS840 based system w/EK waterblocks. However, I have raw materials for 8 that are a new design that will support the VRM waterblock in addition to the CPU block. Once those are built I'm open to selling the old...

One of my 4p boxes is gaming/video/photo/development system part time, and folding the rest of the time. It's got a radeon 6990, Asus Xonar soundcard, and LSI 9265-8i raid controller. I don't use the 6990 for GPU folding...

The source for FAH is not needed -- it's possible to build the core engine from the published gromacs source base. The gromacs piece that is relevant from a testing and performance perspective is mdrun. (I built a 64bit Windows mdrun.exe about a year ago with some optimizations which were...

Separate instances is what we are running today -- eg: 4 separate 4p machines, each running their own work unit. Numbers for that were included above, compared to a hypothetical perfect scaling set of linked machines that would all be processing the same work unit.

Can you post details on the HT-link card and other details of how this is supported (BIOS supported needed, software, etc)? I've tested fah up to -smp 160 and scaling is linear relative to number of cores. The currently shipped windows binary has a hard limit of 64 threads due to it being a...

Many of the intel 4p motherboards have a large number of PCIe slots. 11 PCIe slots in several systems I've seen, and these tend to be EFI based bios systems, so the usual limitations around number of adapters tends not to be an issue. Not sure about the systems built around the chips you've...

I don't think that was a single WU. In fact, this box appears to have been assigned standard bigadv for about a week at this point, whereas a slow 2p system has no trouble getting 6904 units... I think the 933k drop was a bunch of results submitted at once, as a result of part of the home...

I don't have a mill at this point, but somebody I know does -- looking at that as the first choice. Have been considering getting something small, eg: like the small CNC machines offered by littlemachineshop.com, but don't have alot of time to invest in this currently. checkout some of the...

Here's a progress update on the VRM cooling. I changed the approach from building copper blocks that mate to other blocks to just fabricating solid copper one-piece waterblocks. These will be attached to an aluminum channel mated to the case; one different approach to this would be to mount...

Server side is really the best way to do this. Server already has all necessary information, at least for two permutations: 1. Effective work-unit turn-around time. This is already used to calculate bonus, etc. 2. TPF and effective gflops of system. This is included in the log file sent...

6903 and 6904 frame times below... If this is ultimately for Stanford, perhaps Stanford should just mine the logs for this? When the work unit results are uploaded, log files are included in that, which has the TPF and various performance statistics. Stanford should be able to see this data...

I'm thinking a structural aluminum angle or channel segment, mounted to the case, which the copper VRM blocks are in turn mounted to -- would lock in place from the side, or typical downward pressure mount respectively. Given the weight of the copper, and the lack of support close to...

You can tell if throttling has taken place via the handy -htc feature of TurionPowerControl (make sure to use the most recent version of this that is multi-node aware). TurionPowerControl -htc | grep -i past HTC features has been active (means overheated in past): false HTC features has...

Good to see more G34 builds! re: temps at 68C. Try flipping fans over -- vent straight up (pull). Separately, try tilting fans at say a 15 degree angle. Curious if that makes a difference with temps.

The SM chassis tend to have several (very loud) intake fans near the front of the case -- would be mounted near the right of the pictured CPU4 (and CPU2), and would blow across the VRMs and into the CPU cooler. CPU1, 3 typically exhaust out the back of the case, so there would be some...

Readings taken using a Fluke IR thermometer while running sustained folding@home load @ 2.7ghz. No fan pointed at VRMs during readings. With a (loud) panaflo 120mm fan pointed at the VRMs, temperatures are 15-20C lower. These readings were from CPU4. Other CPUs have similiar readings...

With the w83795 device loaded, what's the output of the sensors command? On the supermicro board, it seems that some of the sensor alarms may be set too low. At least the SR5690 chip regularly exceeds the configured critical temperature (70C). You can over-ride this with a line like this in...

You need the w83795 sensor device -- modprobe w83795 If not present on your system or part of included lm-sensors device support, need to pickup source here: http://khali.linux-fr.org/devel/misc/w83795/ make make install modprobe w83795 Unfortunately, if you have IPMI enabled via...

Here's a brief video. Try the "FullHD" version -- try downloading the file as it's very high bitrate. http://sfield.smugmug.com/Computers/Operation-Black-Quad-G34/18178480_SGvrxc#1401255653_r8WxsF7-A-LB Memory in this new/retrofit system is still running at 9-9-9, cpus @ 2.7ghz -- still...

Thanks, wasn't aware of that -- tried the main for sale area first. Items posted.

I've been consolidating and updating hardware, and have and SR-2 system for sale. Great working condition, typically clocked at 4.255ghz. EVGA SR-2 motherboard. I'm original buyer. Includes box, all accessories and original air cooling supplies. (2) Intel Xeon x5680 retail chips. (2)...

[SOLD] I've been consolidating and updating hardware, and have (4) Opteron 6180 2.5ghz 12-core "magny cores" chips for sale. These work in single, dual, and quad socket G34 motherboards. Recommendation is running (4) in a quad socket motherboard for 48 cores total. These are about...

I only played the one game for about 15 minutes last night, since I'm still setting up software. Seemed very playable, as before. I'll install FRAPS and check framerates tonight. Most games will likely only use one physical CPU, so it may be worth pushing that CPU to an even higher clock...

Thanks! One usage is gaming -- added a few more photos to original post.

Under load, temps are up about 5C by the time the last CPU in the loop is hit. Noctua 120mm radiator fans are currently running at 1000RPM. This is similiar to the first WC 4p machine I built -- temps are up 6C at the last CPU in the loop.

Note the case is the SR-2 version of the double wide 21.

Thanks! System is operational. More photos and details later. Initial post updated with more photos and additional system specs. System is used for programming, photo and video processing, gaming, folding. Yes, those chipset fans are louder than all other fans combined. In fact, they...

This is a "retrofit" around an existing workstation system. Danger Den Double Wide 21 tower. Supermicro Quad G34 Motherboard. Silverstone ST-1500 power supply. Koolance CPU-370 waterblocks. Dual water loops, (3) 3x120mm radiators. Radeon 6990 with EK 6990 water block. LSI 9265-8i...

16:00 TPF with 6180s running at 2.7ghz, 6-7-5-20-1T memory running at DDR3-1440. NB @ 1944. ~760W power draw 16:12 TPF with different chips running at 2.7ghz, 6-7-5-20-1T memory running at DDR3-1333. NB @ 1800. 788W folding power draw (1.175v cpu0, 1.15v cpu1-3) 15:45 TPF with different...

I do this with two cables with no issues. Haven't tried it in other (eg: shared) configurations -- not sure this is even supported with the quad SM board. One nice thing about using two cables is you can pull the eth0/eth1 plugs, yet still interact with the system (as if you were local) via...

Thank you, and sorry for the delay replying on this -- vacation took over (rare/good). Here's more detail: http://www.evga.com/forums/tm.aspx?m=1103828

A few things to try - Verify HT=enabled, NUMA=enabled, memory channels populated and configured optimally. The supermicro board picks up the custom SPDs on all nodes, but it's not yet understood how to select banks other than CPU4 from a SPD programming perspective. So the DIMMs need to be...

For the EK AMD block, I used a metal file to file down 1-2 mm of the acetal around the left and right motherboard mounting screws, which allows the block to be properly aligned with the chip. Probably doesn't make a big difference in block performance -- mostly asethetics and stability in your...

6904 finished without incident. Results file is basically same size as 6903. 6904 conclusion: Parallel run - timing based on wallclock. NODE (s) Real (s) (%) Time: 96416.794 96416.794 100.0 1d02h46:56...