All We Know About Maximizing CPU Related Performance

[Tutor]

Updates

I. Sometimes, propriety designs prevent system improvements. This portion of post 1099, above, is a no-go: "I'm also considering ... water-cooling their [ my two 32-core systems' ] 4xE5-4650s. I cannot find any water blocks to fit them securely because the secured seating for the heatsinks are a proprietary design size. Thus, there's no way to attach the water blocks securely.

II. When your pigs get too fat, you get much less lean meat. Amazon notified me that those $800 new Titan Zs mentioned in post 1098, above, have shipped. So far - so good. The seller increased the price of them about a day ago to $900 ea. No matter if I do promptly receive them in perfect working order, the very fact that the sales prices of the Titan Zs are significantly dropping everywhere raises my concern a little and gives rise to my question, "What does Nvidia have for the next horizon?" The Z's began to dribble out in June of this year (not even 6 months ago). Even EVGA's site, which has often shown (and consistently for the last month or so shows) them as out of stock, and now shows a price drop of %50. None of the other GTXs have experienced price drops as vast so quickly. And then again, almost everyone of who I'm aware (including me) firmly believes that the Titan Zs were grossly overpriced originally. Maybe, the fat pigs are being slaughtered.

III. My OctaneRender V1.2 Benchmark rendering goal for Six GTX Titan Z = TE (original Titan Equivalency) of 15.83.

1) OctaneRender time goal for 1 GTX Titan Z: 72 sec. (render time for one of my GTX 780 Ti) / 2 = 36 Sec.

2) OctaneRender time goal for 2 GTX Titan Z: 36 sec. / 2 = 18 sec.

3) OctaneRender time goal for 4 GTX Titan Z: 18 sec. / 2 = 9 sec.

4) OctaneRender time goal for 6 GTX Titan Z: 9 sec. / 1.5*/ = 6 sec.

-
Yielding a TE of 15.83 ( 95 sec [ see http://www.barefeats.com/gputitan.html ] / 6 sec = 15.83333333333333 ).


*/ You have to keep doubling the number of GPUs to cut the render time progressively in half, but I’ll be installing only six of them in a Tyan, which breaks the doubling progression by not going from 4 to 8 GPUs. That’s what accounts for that 1.5 multiple ( 6 / 4 = 1.5 ). The Tyans will also be tasked with video/film production chores. I’ll be using three of the four (non-GTX occupied) slots to assist in accomplishing those chores. In the fourth of those four (non-GTX occupied) slot will be my GT 640 4g. So in sum, my goal is to make this system render about sixteen times faster than the Titan system tested by Barefeats (using the original reference design Titan) and have more than twice the compute ability of the Google brain system aggregation [ http://www.youtube.com/watch?v=2JjxgJcXVE0 ].

 

[Tutor]

Titan Z Update

The Amazon Titan Z seller referenced in my post 1098, above, and in my last post isn't serious. Thus, I'd recommend avoiding him.

 

[lowendlinux]

Big Maxwell is around the corner and with it will come a new Titan supposedly at 12GB. There was a sisoft bench that right now is either a new Titan or a new Quadro the internet hasn't figured out yet which.

 

[Tutor]

Big Maxwell is around the corner and with it will come a new Titan supposedly at 12GB. There was a sisoft bench that right now is either a new Titan or a new Quadro the internet hasn't figured out yet which.

Thanks for the update. There have been many rumors of new Titans in the last few months:

http://wccftech.com/alleged-gtx-titan-x-specs-benchmarks-leak-sisoft-sandra/
http://www.techspot.com/news/58793-nvidia-geforce-gtx-titan-2-gtx-960-detailed.html
http://linustechtips.com/main/topic...ged-specs-leaked-–-gtx-980-ti-or-gtx-titan-x/
http://wccftech.com/nvidia-geforce-gtx-980-ti-gtx-titan-x-coming/
http://videocardz.com/51032/nvidia-maxwell-gm200-spotted-new-gtx-titan-ii-core
http://www.redgamingtech.com/nvidia-releasing-geforce-titan-x-980-ti-says-rumors/
http://egmr.net/2014/10/another-nvidia-titan-z-price-drop-preparing-another-maxwell-card/ .


The Titan Z's 50% price drop helps to fuel those rumors.

 

[Tutor]

Applied oTitan Equivalency - making render timing/pricing estimates.

I now have an oTitan Equivalency of between 83 to 84. I can now project the time it will take for me to render a scene or project by multiplying how long it would take to render the scene with one reference design (RD) original (o) Titan by .0123. E.g., if it would take one year to render a movie project with one RD oTitan, it would take me, applying all of my GPUs, without relying - significantly- on my CPUs, 365 x .0123 or about 4.4895 days or 4 days and about twelve hours. If it would take 30 minutes to render a scene with one RD oTitan, it would take me, applying all of my GPUs, without relying - significantly- on my CPUs, 30 x .0123 or about 0.369 minutes or about about 22 seconds (0.369 * 60 sec. = 22.14 sec).

To cut render times in half in Octane works like this:
1) If one GTX oTitan takes 30 minutes to render a scene.
2) It'll take two oTitans to cut the last render time (30 minutes) to 15 minutes.
3) It'll take four oTitans to cut that last render time to 7.5 minutes, or in half.
4) It'll take eight oTitans to cut that last render time to 3.75 minutes, or in half.
5) It'll take sixteen oTitans to cut that last render time to 1.875 minutes, or in half.
6) It'll take thirty-two oTitans to cut that last render time to 0.9375 minutes, or in half.
7) It'll take sixty-four oTitans to cut that last render time to 0.46875 minutes, or in half.
8) So, I had to use interpolation because my TE is between 83 and 84 and is also between 64 and 128.

Information such as this can also help you to make better pricing estimates for your render jobs for the benefit of others or those render jobs done for you by others, using judicious sampling.

 

[Tutor]

Liquid-cooling GPUs allows their capabilities to shine and promotes their longevity.

I've reached the maximum no. of systems that I care to babysit - 24. Using (a $70 Corsair Hydro H55 CPU cooler [ http://www.corsair.com/en-us/hydro-s...iet-cpu-cooler ] (it has the least height of the Corsair Hydros) plus a $30 Kraken G10 CPU to GPU converter [ http://www.nzxt.com/product/detail/1...u-bracket.htm ) that I mentioned in earlier posts or full bodied water-coolers, I'm now going to be water-cooling all of my GPUs. That includes those in my Mac Pros. I'll also be examining the feasibility of water-cooling all of their CPUs using Corair or other water-cooling systems. Since my 2007/2009->2010 Mac Pros don't have room internally for the radiators/coolers, I'm looking at ways to install the radiators on the outside front grill. Obviously, this will change their esthetics, but since I looking forward to draining from them all of the useful life that they can offer, that's a compromise that I willing to make to extract from them the last bit of performance in aid to my render farm.

My water-cooled CPUs and GPUs have temperatures of 24 to about 30 degrees centigrade at idle and average temperatures of about 45 degrees centigrade under full load. With these modifications, I anticipate that I'll be able to get my oTitan RD TE closer to 90, without purchasing newer GPUs. However, by upgrading my GPUs overtime (and selling the older ones that I choose not to keep as backups) to the latest and greatest (something it is impossible to do with my CPUs), I anticipate that I'll be able to reach a TE of greater than 100 by mid 2016.

Because this is the path that I've chosen to follow to maximize my CPU-like performance, the now new Mac Pro and, probably newer iterations of it, just don't aid me in satisfying my needs or accomplishing my goals.

 

[Tutor]

Actions speak louder than any words.

Today Nvidia released the Tesla K80 [ http://videocardz.com/54068/nvidia-announces-tesla-k80-first-graphics-card-with-24gb-ram ]. What powers this dual GPU 24G (12G per GPU), video abhorrent beast? Like most Teslas are, this card has no video outputs. It's solely for compute functions. Why does it use two GK 210s - offpring of the GK110a (first gen - GTX Titan) and, most recently, the 2nd gen GK110b (latest GTX 780 6G, GTX 780 Ti, GTX Titan Black and GTX Titan Z) and not some variant of Maxwell? Or, more directly, why doesn’t Nvidia’s latest GPU card with the highest CUDA based compute ability have Maxwell GPUs? Because actions speak loader than words, “Maxwells are for gaming/video cards.” Don't believe that the card maker doesn't know what's best for CUDA compute ability. Moreover, that Videocardz's article said don't hold your breath for new Nvidia cards being released this month or in next. So if you're having problems running CUDA on your GTX 970s and 980s, know that Nvidia feels your pain. That's why it has decided to stick with Keplers when it comes to CUDA compute cards.

 

[lowendlinux]

I don't if the assumption you made is correct. This Tesla is passively cooled and intended for the server and since Teslas are really intended to custom code I don't imagine people are ready for new gen and new code quite yet. We also haven't seen big Maxwell yet so it may not be ready for Tesla. It could also be something as simple as a refresh like they've done the past few generations. I'm not ready to slide down that slide with you yet.

 

[Gav Mack]

Showing my age here somewhat tutor but that Lian-li case is reminiscent of the DEC Vax 11 systems in my computing youth although much more sleek and modern

Must be a pleasure to have a case which isn't a tight fit for practically anything!

 

[fhenry]

I've reached the maximum no. of systems that I care to babysit - 24. Using (a $70 Corsair Hydro H55 CPU cooler [ http://www.corsair.com/en-us/hydro-s...iet-cpu-cooler ] (it has the least height of the Corsair Hydros) plus a $30 Kraken G10 CPU to GPU converter [ http://www.nzxt.com/product/detail/1...u-bracket.htm ) that I mentioned in earlier posts or full bodied water-coolers, I'm now going to be water-cooling all of my GPUs. That includes those in my Mac Pros. I'll also be examining the feasibility of water-cooling all of their CPUs using Corair or other water-cooling systems. Since my 2007/2009->2010 Mac Pros don't have room internally for the radiators/coolers, I'm looking at ways to install the radiators on the outside front grill. Obviously, this will change their esthetics, but since I looking forward to draining from them all of the useful life that they can offer, that's a compromise that I willing to make to extract from them the last bit of performance in aid to my render farm.

My water-cooled CPUs and GPUs have temperatures of 24 to about 30 degrees centigrade at idle and average temperatures of about 45 degrees centigrade under full load. With these modifications, I anticipate that I'll be able to get my oTitan RD TE closer to 90, without purchasing newer GPUs. However, by upgrading my GPUs overtime (and selling the older ones that I choose not to keep as backups) to the latest and greatest (something it is impossible to do with my CPUs), I anticipate that I'll be able to reach a TE of greater than 100 by mid 2016.

Because this is the path that I've chosen to follow to maximize my CPU-like performance, the now new Mac Pro and, probably newer iterations of it, just don't aid me in satisfying my needs or accomplishing my goals.

Can't Wait for a 2007 mp water cooled, at xlryourmac there is one exemple with mp 2006 3,33ghz with besl mod with x5350 es

 

[Tutor]

I don't if the assumption you made is correct. This Tesla is passively cooled and intended for the server and since Teslas are really intended to custom code I don't imagine people are ready for new gen and new code quite yet. We also haven't seen big Maxwell yet so it may not be ready for Tesla. It could also be something as simple as a refresh like they've done the past few generations. I'm not ready to slide down that slide with you yet.

I hope that you're 100% right and that I'm 100% wrong because I don't want anyone who has purchased those GTX 970 and 980 Maxwells for CUDA computing to end up having wasted their money. I too considered purchasing 980 Classifieds. However, time will tell whether I made the right decision to forgo them, and to get more tried and true Keplers such as Titan Zs, 780 6Gs and 780 Tis, which haven't had any CUDA related issues. Those Maxwells do seem to excel at OpenCL; but unfortunately for me - not many GPU assisted applications that I use, rely on OpenCL or excel with it.

 

[lowendlinux]

I have to learn hold off on typing in the morning my last post was clear as mud, there were whole words missing

 

[Tutor]

To all of my cousins - no matter how distant - In Lumine Tuo Videbimus Lumen

Can't Wait for a 2007 mp water cooled, at xlryourmac there is one exemple with mp 2006 3,33ghz with besl mod with x5350 es

I'd water-cool my many older systems [Power (PC/Mac/Macintosh/Computing) 180s, 7500/100s, 8100/80s, 8500/120s and 9600/200s, and my Atari TT030 (now an 040), Amigas 500, A1200, 2000s (with Video Toasters), 3000 (with Video Toaster) and 4000 (with Video Toaster) (now all with the fastest 040s or other Motorola upgraded chips) and (all still working greater than ever with the latest CPUs that I could find to upgrade them and later overclock them)] if I could/can find coolers for them at reasonable prices. Of course, my water-cooling schedule places my most recent acquisitions first. So it'll be a while before I get to my 2007 MPs and even much later before I get to my Ataris, Commodores, and Power /PC/Macintosh/Computing/Macs. Obviously, I keep my stuff and keep it running for as long as I'm able to fix them when needed (except that I will occasionally give them away to youth who show a strong interest in computing technology). Thanks for reminding me of the BESL mod. I'll consider whether it offers anything of significance to my 2007s before I water-cool them.

----------

I have to learn hold off on typing in the morning my last post was clear a mud, there were whole words missing

My friend,

You're not alone - sometimes I'm more guilty than are you. And yet, I fully understood you. No harm, no foul. [ http://en.wikipedia.org/wiki/Chick_Hearn#Chick-isms ] Moreover, I especially love those who disagree with me and help to enlighten me. Thanks - may your light continue to shine brightly.

----------

Showing my age here somewhat tutor but that Lian-li case is reminiscent of the DEC Vax 11 systems in my computing youth although much more sleek and modern

Must be a pleasure to have a case which isn't a tight fit for practically anything!

The similarities between those two cases are close. They are the most pleasurable cases to own and work with that I've ever had. Like Jello, "there's always room for more." Within the next few months, Ill be using their expansiveness to re-invigorate my EVGA SR-2 builds to make them more relevant.

BTW - I show my 61 years young every chance I get.

 

[DJenkins]

... I'm now going to be water-cooling all of my GPUs. That includes those in my Mac Pros. I'll also be examining the feasibility of water-cooling all of their CPUs using Corair or other water-cooling systems.

Hi Tutor, good to see you're still on the quest for coolness

Are you going to stick to the all-in-one coolers?

Better results could be achieved with a custom loop but then there's the added mess of maintenance and draining it every time you want to swap something out. That's the thing that scared me off the most, and I don't change my components anywhere near as frequently as you!!

 

[fhenry]

I'd water-cool my many older systems [Power (PC/Mac/Macintosh/Computing) 180s, 7500/100s, 8100/80s, 8500/120s and 9600/200s, and my Atari TT030 (now an 040), Amigas 500, A1200, 2000s (with Video Toasters), 3000 (with Video Toaster) and 4000 (with Video Toaster) (now all with the fastest 040s or other Motorola upgraded chips) and (all still working greater than ever with the latest CPUs that I could find to upgrade them and later overclock them)] if I could/can find coolers for them at reasonable prices. Of course, my water-cooling schedule places my most recent acquisitions first. So it'll be a while before I get to my 2007 MPs and even much later before I get to my Ataris, Commodores, and Power /PC/Macintosh/Computing/Macs. Obviously, I keep my stuff and keep it running for as long as I'm able to fix them when needed (except that I will occasionally give them away to youth who show a strong interest in computing technology). Thanks for reminding me of the BESL mod. I'll consider whether it offers anything of significance to my 2007s before I water-cool them.

----------

No problem, not Easy to find but 10% increase cpu is nice. I also see to possibility to increase the ram of those 2007 mp up to 56 go seen on the netkas forum.

I will soon create a thread regrouping all post about the ultimate upgrade beyon the most common upgrade : besl + 8 go ram stick and your Nice testing with 2 titans

 

[Tutor]

One must be virtuos when building a custom cooling system - much patience is required

Hi Tutor, good to see you're still on the quest for coolness

Are you going to stick to the all-in-one coolers?

Not for the Tyans. Each Tyan will have 6 Titan Zs and each Z has two GPUs. It's simply not doable with all-in-ones, plus each Tyan has two CPUs that I'd like to keep cooler with water, although I did give these a peek: http://www.ldcooling.com/shop/43-ld-pc-v10-reverse .

Better results could be achieved with a custom loop but then there's the added mess of maintenance and draining it every time you want to swap something out. That's the thing that scared me off the most, and I don't change my components anywhere near as frequently as you!!

I'm building a custom loop for 6 Titan Zs and the dual 5680s, with most of the important water-cooling appliances for each Tyan (and another PSU) housed in one of my old, unused chassis which me and Mr. Dremel will modify as needed. The back of each Tyan case will also have to be modded slightly to allow for external cords, cables and tubing with the top of the case fully installed [see that cMPish grill at the upper part of the rear of the Tyan's case - well you won't be seeing much of it shortly]. I'll be using a four pump/one reservoir combo - Alphacool Repack [found at Frozen PC], and two industrial use radiators - the Phobya G-Changer Triple 120mm HPCs (on each of which I'll put 6 quiet fans and each Phobya will be dedicated to cooling three Zs and one CPU). When all is said and done the projected average total cost of each of the six Titan Zs (and the cost of water cooling them and the CPUs and the cost of the additional PSU, but of course excluding what was already in the Tyan) will be under $1.9k per Z - which is more than $600 (US$) less than that of having purchased 12 Titan Blacks alone (and not the water-cooled versions of them which each are a couple of hundred US$ more) [ 1.9 * 6 = 11.4; 1.0 * 12 = 12; leaving a difference .6 ] . Remember also that consolidating more GPUs in each system saves money when if comes to 3d rendering software licensing costs, which is system (not GPU) based. Thus, I'll have the rendering equivalent of 12 Titan Blacks in one system, exceptionally cooler and cooler CPUs all within that price package. Additionally in each Tyan is a GT 640/4G that I'll use for interactivity when doing 3d projects and a 4 port SATA external connection card that I'll reconnect to an 8T external raid 0 of 4 Western Digital 2T drives. Within the next week I be installing in my first Tyan revamp the sixth Titan Z and dual 960 G Revo PCIe cards [ each with up to 1800MB/s reads, 1700MB/s writes, and 140,000 random write IOPS ]. I'll run the two Revo drives in raid 0 to double those speeds. Next comes watering the GPUs and CPUs. In sum, I have a lot of work planned for each Tyan. Moreover, installing a sixth Titan Zs to occupy the last of the dual width only slots and two single slot Revo drives will fill up the four currently empty slots in the pic, below and thus all ten of its x16 size slots will be filled. Two of those x16 size slots are electrically x4 - those two slots now and will hold the GT 640/4g and the Sata X4 (x4) card.

Oh yes! To avoid the added mess of maintenance and draining it every time I want/need to swap something out, I'll be installing quick disconnects throughout the circulatory system to avoid leakage and waste of coolant [ http://www.maximumpc.com/article/ho..._techniques_detailed_and_constructed?page=0,1 ]. Why can't we have those installed internally within us?

Compare pic, below, to pic in post #1099 (on prior page in this thread) to see why this waiting for parts and deals has gotten to be a p___ i_ t__ ___. Is patience a virtue or a test? You decide.

 

[Tutor]

Changing the world - one at a time.

No problem, not Easy to find but 10% increase cpu is nice. I also see to possibility to increase the ram of those 2007 mp up to 56 go seen on the netkas forum.

I will soon create a thread regrouping all post about the ultimate upgrade beyon the most common upgrade : besl + 8 go ram stick and your Nice testing with 2 titans

All wonderful information. Let me know when and where to find your thread after it goes live. Keep up that great work. Our old stuff only becomes irrelevant when we choose to no longer set our minds in motion to keep that stuff relevant to our perceived needs. Keeping that stuff relevant is a very rewarding challenge.

My challenge to all of us: Immediately after having decided that an old system is no longer worth the effort to keep it relevant to what you do, find an inquisitive/interested child, and freely give it to him or her and commit the time to help him/her to grow it further, i.e, make it relevant to their interest(s). Your reward will be astounding, for you'll clearly see what's the most valuable thing in life we have and can do - use our time and resources to help others, particularly the youth, to grow.

 

[nigelbb]

Showing my age here somewhat tutor but that Lian-li case is reminiscent of the DEC Vax 11 systems in my computing youth although much more sleek and modern

Must be a pleasure to have a case which isn't a tight fit for practically anything!

I worked for DEC (& the successor companies). That case looks like a Micro-VAX II which I can recall being launched. http://vaxine.bitcon.no/vaxine.html

The previous VAXen (the correct form of the plural were much bigger. With the original VAX 11/780 you could open the door & duck inside under the 8" floppy disk drive. It was the size of two 6' rack mount cabinets.

The Micro-VAX II revolutionised the computer industry as for the first time a proper time sharing multiuser computer system could be bought out of the departmental budget & they were no longer beholden to monolithic mainframes supported by the IT department.

 

[Tutor]

Size does matter.

I worked for DEC (& the successor companies). That case looks like a Micro-VAX II which I can recall being launched. http://vaxine.bitcon.no/vaxine.html

The previous VAXen (the correct form of the plural were much bigger. With the original VAX 11/780 you could open the door & duck inside under the 8" floppy disk drive. It was the size of two 6' rack mount cabinets.

The Micro-VAX II revolutionised the computer industry as for the first time a proper time sharing multiuser computer system could be bought out of the departmental budget & they were no longer beholden to monolithic mainframes supported by the IT department.

The Micro-VAX II does have a closer resemblance.

 

[Tutor]

Doubling your pleasure (or pain).

How are getting two Intel Xeon E5-2699 V3 18-core CPUs for a little less than the price of one elsewhere [ see http://item.taobao.com/item.htm?spm=a230r.1.14.16.s5UvB8&id=39844514063&ns=1#detail and http://cny.fx-exchange.com/usd/12000.00-exchange-rates.html ]?

 

[Tutor]

Freezing Your GPU and CPU from FrozenPC

Santa,
Please bring me 4 of the following and a case appropriately designed to hold all of them and a very large SuperMicro or Tyan tweakable motherboard:

http://www.frozencpu.com/products/1...40_AM2_Xeon_CE-48-D-1C-1G.html?tl=g30c367s679

for the ultimate in extreme CPU/GPU cooling.

 

[Gav Mack]

I worked for DEC (& the successor companies). That case looks like a Micro-VAX II which I can recall being launched. http://vaxine.bitcon.no/vaxine.html

The previous VAXen (the correct form of the plural were much bigger. With the original VAX 11/780 you could open the door & duck inside under the 8" floppy disk drive. It was the size of two 6' rack mount cabinets.

The Micro-VAX II revolutionised the computer industry as for the first time a proper time sharing multiuser computer system could be bought out of the departmental budget & they were no longer beholden to monolithic mainframes supported by the IT department.

That's the exact Vax I was thinking of - I remember the huge model Vax systems too and the PDP-11 monsters also. Huge cast ally casings for 5-10 megabyte winchesters which must have been 10-12U in terms of rack height. Before the day of my first Macintosh!

 

[Tutor]

Update - GTX 980 on Octane

This is the latest on Octane's coders' attempts to optimize Octane, through CUDA, for the GTX 980 (and also the GTX 970):

A user (pBarrelas} wrote: "The GTX980 is supposed to be faster than the GTX780 Ti, I just wish this will also happen in Octane in the near future... ."

This was the response from a member of Octane's development team: "But it isn't. Nothing we can do about it. So don't expect a 980 to be faster than a Titan Black / 780 Ti. (Just check my signature [ which says,” In theory there is no difference between theory and practice. In practice there is. - Yogi Berra. “] http://render.otoy.com/forum/viewtopic.php?f=33&t=43530&start=40

Current attempts to optimize Octane for the current TOL Maxwell (the GTX 980) has those cards now performing 3d rendering at about the same speed as reference design GTX 780s. This, however, represents a huge increase for the GTX 980s over that which they had in the last version of Octane. However, these mods also appear to have slightly negatively affected the performance of pre-Maxwell GPUs when running the latest version of Octane and CUDA.

I strongly suspect that Nvidia is drawing a higher lined price point for CUDA prowess than the GTX 970s and 980s fall and that line, while it most certainly won't (or shouldn't) be as high as Nvidia attempted to draw it last with the 3K Titan Zs (that now sell at about one-half of there intro prices), I still suspect that it'll be closer to 1K than is the price of the GTX 980s. More CUDA users can (and probably do) make more money from the applications taking advantage of CUDA than most gamers make from gaming. So, I don't begrudge Nvidia to that extent.

None of the above still, however, doesn't rob the 980s of those things at which they excel - they're fantastic gaming cards (OpenGL) and, dollar for dollar, they're vastly superior OpenCL performers than Nvidia has every produced.

 

[glenandsonja]

QBED's in a 2P ECC system

Gigabyte makes a dual CPU system - GA-7PESH3 [ http://b2b.gigabyte.com/products/product-page.aspx?pid=4468#ov ] if your applications justify use of a dual CPU system; and, of course, if your applications would benefit more from a 4 CPU system, then consider Supermicro [ http://www.superbiiz.com/detail.php?name=SY-847R7FP ] .

Hi Tutor, I've enjoyed your detailed and insightful posts, and noticed you're running QBED's. I'm looking for advice on which motherboard to use, for a dual CPU system using 2 QBED's that also supports ECC memory. I would also consider a quad CPU system, however the price of those motherboards is more than I want to spend (although I could probably trade something for one). You mentioned the GA-7PESH3 in the quote above, have you tested this with QBED's? I bought a Supermicro X9DRD-iF but couldn't get it to work with 2 QBED's (worked fine with 1), and I'd prefer not to get burned a second time . Thank you in advance for any information or help you can provide.

 

[Tutor]

Hi Tutor, I've enjoyed your detailed and insightful posts, and noticed you're running QBED's. I'm looking for advice on which motherboard to use, for a dual CPU system using 2 QBED's that also supports ECC memory. I would also consider a quad CPU system, however the price of those motherboards is more than I want to spend (although I could probably trade something for one). You mentioned the GA-7PESH3 in the quote above, have you tested this with QBED's? I bought a Supermicro X9DRD-iF but couldn't get it to work with 2 QBED's (worked fine with 1), and I'd prefer not to get burned a second time . Thank you in advance for any information or help you can provide.

For reasons having nothing to do with the GA-7PESH3 itself, I never purchased one and do not know personally anyone who did. I have no idea whether two DBED's would run in the GA-7PESH3. All of my QBEDs are either in my 4-CPU Supermicros or my 1-CPU Gigabyte GUP4 motherboards and they run fine in both of those. When you tried them on the X9DRD-iF, had you installed on that motherboard the latest bios?