Have the scientists/developers/programmers been able to make any progress on updating the apps for GTX Titan compatibility?

I've got mine busy generating rainbow tables at present and would like to put it back to work here at GPUGrid.

Operator
____________

Thought I'd add something I noticed today. The details page for my system with the GTX Titan says it has 4GB of ram.

"Coprocessors NVIDIA GeForce GTX TITAN (4095MB) driver: 314.9"

The problem is that the GTX Titan actually has 6GB of ram.

Could that account for why tasks are failing? A memory size mismatch between what the system thinks the card has and what it actually does have?

Has anybody else with a GTX Titan noticed this discrepancy?

Operator[/b][/quote]
____________

This is from the BOINC log at startup:

3/11/2013 5:56:31 PM | | NVIDIA GPU 0: GeForce GTX TITAN (driver version 314.9, CUDA version 5.0, compute capability 3.5, 4096MB, 8384512MB available, 4989 GFLOPS peak)

3/11/2013 5:56:31 PM | | OpenCL: NVIDIA GPU 0: GeForce GTX TITAN (driver version 314.09, device version OpenCL 1.1 CUDA, 6144MB, 8384512MB available)

Strange. Only the OpenCL section is reporting the correct memory amount.

Operator
____________

The NVIDIA driver API seems only to return a 32-bit value.

It's been reported.

I saw over at the SETI forums that the Titan memory error was reported to the BOINC alpha list.

So I guess the BOINC devs are working that.

But are you indicating that would be a Nvidia driver problem instead?

Operator
____________

Well. In the first place, those who run a 32-bit operating system will only ever see 4GB - that's the limit.

For 64-bit OSs: you'll have seen Claggy's reports of the 'wacky memory' bugs, when BOINC was reporting millions of millions of megabytes for some cards. The BOINC devs can't afford to buy a new card to test every time they come out - certainly not at Kepler and Titan prices - so Rom remoted in to my GTX 670 to work out what was going on.

Apparently, the 64-bit drivers require the programmer to pass a 64-bit memory buffer to return the memory size: but they only populate the lower 32 bits. Any initialisation garbage in the upper 32 bits remains unchanged - that's what led to the wacky memory sizes. I think Rom cured that by zeroing all 64 bits before use: if the drivers are still only filling in 32 of them, there's not much more to be done.

I wonder what the Tesla drivers do? Anyone got one they can lend to BOINC?

I did fail to mention that I was running Win7x64. Sorry.

I have seen several event log postings over on the SETI board showing the same 4096mb memory amount (you were on that conversation as well).

I have a support ticket in with Nvidia and am in the process of a dialog with them now.

Operator


____________

Hi,

I have a titan and also I do work for other projects.

We found having the default memory speed caused too much errors.

When we clock down the memory from 3000MHz down to 2500MHz, the apps were a _lot_ more stable.

At 3000MHz, some apps would die within mins (4hrs was the best stability). Drop the speed down, stable as a rock. Ran for 48hrs+ no dramas, all results double checked aok.

Hope this helps.

-- Craig

Hi,

I have a titan and also I do work for other projects.

We found having the default memory speed caused too much errors.

When we clock down the memory from 3000MHz down to 2500MHz, the apps were a _lot_ more stable.

At 3000MHz, some apps would die within mins (4hrs was the best stability). Drop the speed down, stable as a rock. Ran for 48hrs+ no dramas, all results double checked aok.

Hope this helps.

-- Craig

Mem clocky are ok at 3000Mhz no worries.

You can check milkyway (with DP enabled) and run 5 tasks in parallel. Primegrid sieve can be ok also.

Mem clocky are ok at 3000Mhz no worries.

You can check milkyway (with DP enabled) and run 5 tasks in parallel. Primegrid sieve can be ok also.

I now have two Titans (as does nucleon I think)and am having no difficulty running Einstein@home (GPU only tasks), as well as the DSTRGEN project.

Haven't tried PrimeGrid but good to know it works.

I have a ticket in with Nvidia on the memory reporting issue (4096mb reported vs 6144mb actual). Nothing back from that yet. The rep said that he was unaware of any issues in BOINC or GPUGrid so I pointed them toward the forums where the problems were mentioned and sent in some screenshots and they are presently "researching" the problem.

So since MW works as well, I may go back over there for a while then.. until we get the longs sorted here.

I don't think I'll be downclocking the memory just yet, unless somebody can conclusively state that doing so fixes the long work units here at GPUGrid.

Operator
____________

Milky is open_cl so the same power as top ati is a little miracle. I have not expected that.

DistrRtGen can crunch one task in 530s with titan, overclocked titan in 480s. This is a little better then top ati.

But I'm looking forward to gpugrid. I have buy this card for gpugrid.

But I'm looking forward to gpugrid. I have buy this card for gpugrid.

There is mention of a specific OpenCL fix for TITAN in the 314.21 beta driver release notes:

[GeForce GTX TITAN]: Application crashes or a TDR occurs when running
applications that use OpenCL. [1237211]

I don't know what progress is being made on GTX Titan compatibility with GPUGrid apps, but as has been stated here and in other areas the Titan seems to be happily crunching away on Einstein, Milky Way, Prime Grid, Distributed Rainbow Table generator, etc.

Comments made by Mr. Haselgrove regarding the cause of the misreporting of the total available memory (4096mb vs 6144mb) due to the driver API leaves me wondering if this is in Nvidia's lane, BOINC's lane, or both, and whether anybody is working to resolve this issue or if this issue even matters when crunching CUDA apps with a Titan.

Does this memory reporting mismatch affect how the apps run (or fail to run) on a Titan? Or is it a 'red herring'? I don't know. But obviously some projects are not having any issues at all running on the new card.

Does anyone have a timeline, idea, clue, or rumor regarding when the GTX Titan might be able to successfully support GPUGrid project apps?

Thanks,

Operator
____________

Great news!

Thank you Sir!

Operator
____________

If you have a GeForce Titan, pending GPUGRID, the following projects are compatible,

Boinc

Collatz, Math
DistrRTgen, Math
WCG HCC GPU, Bio
Moo!, Math
Einstein, Astro
MilkyWay, Astro
Seti Beta, Astro

Folding @ Home, Bio

@+
*_*
____________

Encryption is usually done with primes, and the calculation of prime numbers require a lot of mathematical calculations especially if one wants to calculate prime numbers very quickly.

L'encryptage est généralement effectué avec des nombres premiers, et le calcul de nombres premiers demandent beaucoup de calculs en mathématique surtout si on veut calculer des nombres premiers très rapidement.

@+
*_*
____________

The problem with the GeForce TITAN and GpuGrid a chance to be corrected?
I personally have tried almost all projects. But mostly I run projects that are devoted to biology and rotate with a GPU, GPUGRID, WCG HCC and Folding @ Home.
It's a shame not to be able to rotate GPUGRID.

Le problème de GpuGrid avec la GeForce TITAN à une chance d''être corrigé ?
Personnellement j'ai essayé pratiquement tous les projets. Mais je fais tourner principalement les projets qui sont consacrés à la Biologie et qui tournent avec un GPU, GPUGRID, WCG HCC et Folding@Home.
C'est dommage de ne plus pouvoir faire tourner GPUGRID.

@+
*_*
____________

Mmmm and they dare to say that ignorance does not breed stupidity.

All research uses maths, but using mathematics doesn't mean you're performing research!
Bottom line, it's a meaningless waste of electric and resources - a needless contribution to global warming.
It shouldn't be mentioned along side science, medicine or research.

Moo! is as batty as The Batty Bat

Touching base, I think we're now past the deadlines mentioned.

Any update?

-- Craig

Any update?

Not forgotten, just preëmpted.

MJH

I have both Titans running 5 concurrent Einstein tasks each and it's running like a freight train.

Interesting, it looks like the Titan is running about 35-40% slower than an HD 7970 with 5x WUs and around the same speed as an HD 7950, according to this chart at Einstein:

http://www.dskag.at/images/Research/EinsteinGPUperformancelist.pdf

Interesting, it looks like the Titan is running about 35-40% slower than an HD 7970 with 5x WUs and around the same speed as an HD 7950, according to this chart at Einstein:

http://www.dskag.at/images/Research/EinsteinGPUperformancelist.pdf

I don't even see a Titan on that list...

It's not, but here's the 5x Titan results for you to peruse:

http://einstein.phys.uwm.edu/results.php?hostid=6889477&offset=0&show_names=1&state=3&appid=0

This is two EVGA Titan SC (air cooled) in a Dell T3500 WS with a Xeon W3565, 12GB, Win7Pro x64, and nothing else installed that isn't necessary to run BOINC/Apps.

Admittedly, BOINC still only sees 4096mb of the 6144mb of memory on the cards. I don't know if that has anything to do with it or not.

Operator
____________

They use CUDA 3, because their app fails with CUDA4-5. Apparently it should be a CUDA problem, that was acknowledged by nVidia, but wasn't fixed yet.

BRP4cuda32nv301 - they probably need to update their apps to better utilize the titan. Their apps were built for earlier GPUs, earlier cuda versions and compilers, so they are certainly not designed for Titans.

Thank you all for the recent information and observations regarding the GTX TITAN and CUDA crunching. I chose the TITAN SC for a single-gpu solution for a double-duty gaming and distributed computing box that would use less power than my previous pair of GTX 580 SCs. So far, it hasn't disappointed. If it crunches work units faster at some future date, so much the better ;)

Interesting, it looks like the Titan is running about 35-40% slower than an HD 7970 with 5x WUs and around the same speed as an HD 7950, according to this chart at Einstein:

http://www.dskag.at/images/Research/EinsteinGPUperformancelist.pdf

I don't even see a Titan on that list...

Edit: BTW, thanks to dskagcommunity at Einstein for this fine table of GPU performance!

I don't think the reported memory is the problem with Titan. The apps only use as much memory as necessary anyway, and 4 GB is still plenty (be it for 1 or 2 GPU-Grids or 5 Einsteins).

And regarding Titan and CUDA 3: each SMX is pretty much the same as for the other Keplers, just with some more features (more FP64 units etc.) and more registers. If these features are not being used, it should perform just as well as any other Kepler with this amount of shaders and clock speed.

MrS

I have done a little performance tuning but I think there is more work to do yet.

I have reduced the workload to 4 WUs per Titan now because my CPU has only 8 threads.

I am now averaging 2600 sec. per Einstein GPU WU with an average daily of 87,453.02.

Just waiting for GPUGrid apps I can run instead.....

Operator

____________

I have reduced the workload to 4 WUs per Titan now because my CPU has only 8 threads.
I am now averaging 2600 sec. per Einstein GPU WU with an average daily of 87,453.02.

Already done, i visit the most forums from my mainprojects every day to look whats going on AND if all of my unattended machines running fine ;)

BUUUTTTT, 87k credits? Do i something wrong? My chart calculates ~66,400k

((60secs*60mins=3600secs*24h=86400secs/average calc time 2600secs)*4WUs)*500Credits per WU=~66400
____________
DSKAG Austria Research Team: http://www.research.dskag.at

I didn't do any calculation I just copied and pasted the RAC on the account page.

In fact today's RAC value for ID: 6889477 is: "Recent average credit 88,240.40".

After reflection, I now realize that the RAC and the daily may not be the same.

Thanks,

Operator
____________

He ran 5 WUs before, so the RAC is higher than actual until the next two weeks or so i think. Then it should be ok. We will see ^^
____________
DSKAG Austria Research Team: http://www.research.dskag.at

RAC and "average credits per day" are the same. Did you include RAC from other cards or machines or CPU work?

MrS

And today's RAC for that machine/project is: 89,855.65 probably due to a little more tweaking on my part.

Just bumping this thread.

Any update to Titan compatibility with gpugrid?

-- Craig

Has there been any analysis of any type performed to determine why no GPUGrid apps will run on the GTX Titan?

Anything?

I was under the impression that the developers/programmers now had a GTX Titan (or two) to use in their development process.

Yet we have heard nothing at all as to what the issue(s) are.

Is there anybody successfully running a GPUGrid app using a Tesla K20 or do they have the same problem?


Will there be a date that we GTX Titan owners can expect to be able to support the GPUGrid project with our hardware?

Thank you.

Operator
____________

In programing you cant give a date when you fix some problems, thats why so much games are such a shit today when the developer want to bring it out @ a special date ;)
____________
DSKAG Austria Research Team: http://www.research.dskag.at

Folding is compatible with GeForce TITAN,

For the Folding @ home

http://fah-web.stanford.edu/cgi-bin/main.py?qtype=osstats2

project and have already tried to Gromacs GPU, and they abandoned in favor of OpenMM,

https://simtk.org/home/openmm

GPUGRID uses what?

@+
*_*

Folding and Titan in action ->

https://www.dropbox.com/s/t5fs17aq678uahq/Folding.jpg
____________

Zarck;

Thanks, I am now on F@H crunching away with both Titans.

Again, thanks for the tip.

Operator
____________

You use amber ?

http://ambermd.org

@ +
*_*

You use amber ?

http://ambermd.org

@ +
*_*

From what I gather from discussions around here Titans do work for us locally, except for the fact that a big portion of them apparently contain some kind of fault, for which reason we don't buy any more of them.

It might be a CUDA problem as well.
Einstein@Home for example have issues running their tasks on CUDA 4 and 5. All their attempts to run BRP fail there, so they stick with CUDA 3. Apparently nVidia has acknowledged this problem, but haven't provided a fix since a few years... CUDA 3 might be the reason why Titan runs there.
So the problem here might be a mix of acemd + CUDA 4 + GK110 hardware.

In another distributed app, once we decreased the ram speed (3GHz -> 2.5GHz) everything was fine. (mersenne LL testing, FFT/memory intensive)

Problem being Titan is one of the few vid cards with ram modules on the alternate side of the board where there's no fan/cooling. Also Titan's thermal limit seems to be a lot lower than other nvidia cards. Need to keep it under 80degC (as measured via GPUz). I currently have my machine open with a floor fan pointed at it :)

I'm currently running Milkyway on mine. My dual Titan machine is currently 6th:

http://milkyway.cs.rpi.edu/milkyway/top_hosts.php?sort_by=expavg_credit&offset=0

It might make 5th.

-- Craig

I'm currently running Milkyway on mine. My dual Titan machine is currently 6th:

http://milkyway.cs.rpi.edu/milkyway/top_hosts.php?sort_by=expavg_credit&offset=0

It might make 5th.

-- Craig

How many simultaneous tasks do you do?

I do 8 on each Titan to keep GPU usage at 100%

-- Craig

I'm running only 1 WU per GPU to maintain reasonable temperatures (summer is coming here). I tried to run 2 which was sufficient to load the GPU at 100%, but it was running too hot, so I need to improve cooling.
How long does a single WU take on Titan?

Funny enough, 1WU takes the same amount of time as 8. @1WU GPU usage was 12.5% approx.

-- Craig

Funny enough, 1WU takes the same amount of time as 8. @1WU GPU usage was 12.5% approx.

-- Craig

In another distributed app, once we decreased the ram speed (3GHz -> 2.5GHz) everything was fine. (mersenne LL testing, FFT/memory intensive)

-- Craig

I do not think that will be a problem of heat or frequency, Titan automatically changes the speed according to the GPU load, if double precision mode is activated, the Titan running at 823 Mhz. Folding@home it in turns between 954 MHz and 967 MHz depending on the type of unit Folding calculated.
Under GPUGRID units go on miscalculation from the first second. I managed to calculate and validate units with my Titan for the following projects

Collatz, 
 DistrRTgen, WCG HCC GPU, 
 Moo!, 
 Einstein, 
 MilkyWay, Seti Beta,
Folding@home

Why miscalculations GPUGRID?

@+
*_*
____________

On GPUGRID there is not support for titan yet.

I'm guessing only:

1. gpugrid app is linked with cubin code, so version for CC35 cards is not included yet. In opposite to ditrrtgen which is linked with ptx code.

2. I'm not sure if CUDA4.2 supports CC35 cards (CUDA 5.0 yes but it's another story).

3. There may be other related problems with new CC (conditional compiling etc.)

Zdenek

I installed Boinc 7.1.1 Windows X64 for my TITAN,

http://boinc.berkeley.edu/dl/?C=M;O=D

And I received Gianni units, but there is always the same problem, GPUGRID displays error calculation from the first second calculation.

@+
*_*
____________

Again:

On GPUGRID there is not support for titan yet.

TITANS is the best four?



@+
*_*
____________

Wonder why the SLI isn't hooked up? We should find out the prices for the GTX780 and GTX770 tomorrow, if the 780 is less than $600 US, I might pick one up. I realize a lot of you guys still have kids at home and house/car payments but I don't, I guess I'm an old fart with no bills.

If I do upgrade, I'll just give away my GTX670's to anyone who wants them but only to folks here at GPUGRID.

No need for SLI, the person doing this card 4 Folding calculations in the same time.

That's very impressive. What application are we looking at here? Is that a custom skin for BOINC? With 4 Titans running full tilt in 1 machine must require a pretty beefy power supply, I wonder how many watts he's pulling at the wall?

That's very impressive. What application are we looking at here? Is that a custom skin for BOINC? With 4 Titans running full tilt in 1 machine must require a pretty beefy power supply, I wonder how many watts he's pulling at the wall?

I wonder about the cooling, don't see evidence of water cooling on the GPUs and there's no room for air flow.

80C is a little scary.

Looks like the GTX 780 is coming at about $650, according to this article. It seems like a better deal than the Titan in terms of performance offered for the price.


article:
http://videocardz.com/41960/first-geforce-gtx-780-review-leaks-out-inno3d-gtx-780-ichill-exposed

Yeah - cooling was the first idea that came to my mind.. With no spacing between the GPUs he cannot cool them down properly with air only. I'm wondering at what temperatures are those Titans running, but I believe they must be throttling.
Water cooling is nice, but when the ambient temperature is higher in summer, it's less efficient.

GeForce Titan an air inlet at the back of the card, the air is extracted from inside the PC outwardly.


____________

Nice. Nice.

____________

http://www.amazon.com/EVGA-GeForce-Dual-Link-Graphics-03G-P4-2781-KR/dp/B00CUIVTCS/ref=sr_1_17?s=electronics&ie=UTF8&qid=1369317160&sr=1-17&keywords=GTX+GeForce+780


____________

GeForce Titan an air inlet at the back of the card, the air is extracted from inside the PC outwardly.

Interesting link, looking around I saw:

> ACEMD is compatible with CUDA and OpenCL, the new standard framework for
> parallel and high-performance computing over different architectures.

Any new news coming, like OpenCL support?

Thought that might bring a smile :-)
Thanks for checking it out.

Il y a la solution de mettre une carte mère avec 7 PCI Express et de mettre une carte graphique un slot sur deux ?



There is a solution to a motherboard with seven PCI Express graphics card to a slot on both?


____________