The probability that GPUGRID becomes a calculation error is high though the project of SETI@HOME and GPUGRID is executed with BOINC. Moreover, the message that BOINC stopped is frequently generated in the calculation process, too.

It has not driven recklessly when only SETI@HOME is executing it.

Please the translation tool is used, and understand and acknowledge [zurai] sentences.


My Compter Spec/ DELL DIMENSION 9200
(01) Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz [x86 Family 6 Model 15 Stepping 6]
(2 processors)
(02) NVIDIA GeForce 8600 GTS (256MB) driver: 19107
(03) Microsoft Windows 7
Ultimate x86 Edition, (06.01.7600.00)

The probability that GPUGRID becomes a calculation error is high though the project of SETI@HOME and GPUGRID is executed with BOINC. Moreover, the message that BOINC stopped is frequently generated in the calculation process, too.

It has not driven recklessly when only SETI@HOME is executing it.

Please the translation tool is used, and understand and acknowledge [zurai] sentences.


My Compter Spec/ DELL DIMENSION 9200
(01) Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz [x86 Family 6 Model 15 Stepping 6]
(2 processors)
(02) NVIDIA GeForce 8600 GTS (256MB) driver: 19107
(03) Microsoft Windows 7
Ultimate x86 Edition, (06.01.7600.00)

Just so I understand correctly you are getting a lot of computation errors.

This is not suprising as the card that you mentioned doesn't have much memory (256Mb isn't much) and its not based on the G200 chip. There are a number of cuda issues with the GPUgrid application and older cards, which is why the project recommends one of the following cards:

GTX275
GTX280
GTX285
GTX295
Tesla 10

---

BOINC blog

Your 8600 GTS is now too slow to be able to finish in time. The tasks will be terminated by the project.

MarkJ,
I was under the impression that the amount of RAM used by GPUGRID is about 50MB.
Had this changed?

GTX275
GTX280
GTX285
GTX295
Tesla 10

Hope this includes the newer but less powerful GT 240's with 1GIG of DDR3 VRAM because I am building 2 computers with these cards primarily for video viewing rather than gaming and was hoping to press them into service after xmas with the owners permission of course.

Has anyone any real world results on these cards?

EDIT to add: they are Gigabyte factory OCed versions.

---

Radio Caroline, the world's most famous offshore pirate radio station.
Great music since April 1964. Support Radio Caroline Team -
Radio Caroline

No real world experience, but going by the 40nm GT215 core you should be fine. With 96 shaders and DDR3/DDR5 it will only be about half as fast as a GT260, but then you have 2 of them! You should get a descent return. These are mid range cards, but use slightly less electric and with the reduced core size should generate slightly less heat. That said, the fan looks a bit work shy, and blowing the heat from the GPU over the capacitors would not be my choice of design.

Perhaps you could post the specs and any comments at the page below, when you get them,
http://www.gpugrid.net/forum_thread.php?id=1150

MarkJ,
I was under the impression that the amount of RAM used by GPUGRID is about 50MB.
Had this changed?

I'm not too sure how much it actually uses, however most projects won't run in anything less than 256Mb. I recall Seti had to set theirs at 249Mb or some weird number to allow lesser cards to be used.

Add the Win 7 eye candy and you don't have much vram left for anything. Disabling Aero does free up a bit of vram.

---

BOINC blog

No real world experience, but going by the 40nm GT215 core you should be fine. With 96 shaders and DDR3/DDR5 it will only be about half as fast as a GT260, but then you have 2 of them! You should get a descent return. These are mid range cards, but use slightly less electric and with the reduced core size should generate slightly less heat. That said, the fan looks a bit work shy, and blowing the heat from the GPU over the capacitors would not be my choice of design.

Perhaps you could post the specs and any comments at the page below, when you get them,
http://www.gpugrid.net/forum_thread.php?id=1150

I will of course and happy christmas to all

---

Radio Caroline, the world's most famous offshore pirate radio station.
Great music since April 1964. Support Radio Caroline Team -
Radio Caroline

MarkJ,
I was under the impression that the amount of RAM used by GPUGRID is about 50MB.
Had this changed?

I'm not too sure how much it actually uses, however most projects won't run in anything less than 256Mb. I recall Seti had to set theirs at 249Mb or some weird number to allow lesser cards to be used.

Add the Win 7 eye candy and you don't have much vram left for anything. Disabling Aero does free up a bit of vram.

The latest version of GPU-Z tells you how much video RAM is being used.
On my GT240 (being tested), when running an m3-IBUCH_min_TRYP task it stated that it was using 171MB. At the time the GPU load was 69% and the Memory Controler was 43%.
However on my GTX260 397MB is being used, despite the memory controller load being 11%. Obviously the GTX260 has more bus width and bandwidth, so it is able to allocate and use more RAM.
So the amount of RAM used by the task depends on the cards capabilities.
Its worth noting though, that even with the GTX260 it is below 512MB, so it is still not worth having the extra 512MB in a 1GB card, if it costs more!

The latest version of GPU-Z tells you how much video RAM is being used.
On my GT240 (being tested), when running an m3-IBUCH_min_TRYP task it stated that it was using 171MB. At the time the GPU load was 69% and the Memory Controler was 43%.
However on my GTX260 397MB is being used, despite the memory controller load being 11%. Obviously the GTX260 has more bus width and bandwidth, so it is able to allocate and use more RAM.
So the amount of RAM used by the task depends on the cards capabilities.

Based on my knowledge of how CUDA programing works (which admittedly is hardly at the expert level), I'd say that the difference in memory usage has to do with the number of processors (aka "shaders") on the video card. The more processors available, the more threads that can physically run simultaneously, and therefore more data needs to be loaded into RAM simultaneously. Hence, with the faster cards with more processors, memory usage aught to go up proportionally. I'm sure it's more complicated than that, but I think it would be a safe assumption that the number of shaders has a lot to do with the amount of ram being used.

I think explanation was better.
The more shaders and CUDA cores you have the more RAM you need to use.
That said, to use more RAM you need a wider bus/more bandwidth.

If you got a GT240 with 512MB RAM you would be fine, but a similar card with 2GB RAM would be a waste of money, for here anyway.

I expect the GT300 core cards will require more RAM, as they will have many more shaders. 2GB might be the norm, but perhaps 4GB cards will be common too.

Now, one of your lists of recommended cards includes some GTX260 cards and one does not. HP has recently started offering the GTX260 cards as an option on some of their computers, but nothing higher yet.

If you've uncovered more information on how to tell which GTX260 cards are suitable and which are not, could you mention more on that?

For example, 216 shaders vs. 192 shaders?

I've seen a few other computer companies that offer computers with higher GTX cards, but without the unpacking and initial setup service I'll need to actually use them.

Also, you might want to check the Subscribe button for this thread; it doesn't seem to work fast enough to tell that it worked before moving on to another thread.

The 55nm GT200b cards work fine - all sp216.
The 65nm GT200 cards do not!