As if luck would have it, a workunit put two cards together that are virtually the same, but GTS250 should have the edge due to the 55nm board. However, looking at the results below suggests I am wrong:

From GTS250 Task ID 432911

core_client_version>6.6.15</core_client_version>
<![CDATA[
<stderr_txt>
# Using CUDA device 0
# Device 0: "GeForce GTS 250"
# Clock rate: 1890000 kilohertz
# Total amount of global memory: 1073414144 bytes
# Number of multiprocessors: 16
# Number of cores: 128
# Time per step: 141.641 ms
# Approximate elapsed time for entire WU: 88525.497 s
called boinc_finish

</stderr_txt>
]]>

From GTX9800+ Task ID 433779

<core_client_version>6.4.5</core_client_version>
<![CDATA[
<stderr_txt>
# Using CUDA device 1
# Device 0: "GeForce 9800 GTX/9800 GTX+"
# Clock rate: 1890000 kilohertz
# Total amount of global memory: 536608768 bytes
# Number of multiprocessors: 16
# Number of cores: 128
# Time per step: 60.070 ms
# Approximate elapsed time for entire WU: 37543.545 s
called boinc_finish
</stderr_txt>
]]>

Things that are the same:

* Number of multiprocessors
* Clock Rate

Things that are different:

* Boinc Version
* Global Memory
* Time Per Step
* Elapsed Time

Anyone have any idea why the GTX9800+ would perform over twice as fast?

Note: I am running two other projects at the same time (one which is CPDN), but I have 3Gb of memory so I didn't think it is the factor in this instance?

Rob

The cards should perform identically and the 55 nm chip (which should also be on the 9800 GTX+, anyway) doesn't make any difference in performance. In your list there's also a task with 90 ms / step and one with 75, which would be much more in line with expectations.

Do you have any other 3D software running (game) while running GPU-Grid? Was the wall clock time really ~24h?

MrS

---

Scanning for our furry friends since Jan 2002

I think that these cards are identical, GTS250, and 9800GTX+.

MrS wrote:

Do you have any other 3D software running (game) while running GPU-Grid?

I only had a quick 30 minute game in the whole period it was running and that was on the not very intensive Quake Live

MrS wrote:

Was the wall clock time really ~24h?

It was about 21 hours (75600 seconds).

I've just installed the new Nvidia beta drivers that came out today, so I will see if they improve things at all. My anti-virus is told to ignore the Boinc folders. Is there anything else I can check?

Rob

I think that these cards are identical, GTS250, and 9800GTX+.

Yep. Only difference is the name and pricing.

---

Yep. Only difference is the name and pricing.

And the PCB layout, idle clocks and idle power consumption. but none of these affect performance.

MrS

---

Scanning for our furry friends since Jan 2002