I crunched two numbers and found out that my GeFarce 210 GPU is a V16 running on only one cylinder.

my SETI task was very roughly 40-50 quadrillion FLOP(s) and completed in some 8800 seconds etc.

But the total GFLOPS of my GPU came out to be only 4.7 GFLOPS. This is interesting because my card was once rated at 76 GFLOPS.
It is a Geforce 210 with 16 Cuda Cores.

4.7 GFLOPS x 16 cuda cores = 75.2 GFLOPS, which is very close to the well known and bench-marked speed of this particular video card!

I am using the same driver and service pack and everything, only a faster more reliable motherboard.

Anyone have a clue or has this terrible affliction happened to you?!

the 210 is not a very fast card. it's basically renamed 9XXX series card.

You can expect to achieve about 10% of the actual top performance on seti regardless of which GPU you use.

2 hours seems like a long run. Perhaps checking your results vs. your wingmen may help id if the speed is incorrect

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

And what do you expect from GeFarce 210 ??!

Only Fermi class GPUs (GF 4XX) can run more than one task at a time,
and even for them this is not the default behavior - the user have to do special efforts for this to happen.

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

well, plain and simple I expected the benchmarked speed in GFLOPS to match that of the SETI project results, but the SETI effort only announced a rate of about 4.7 GFLOPS

and the deduction process says that if my GiForza 210 card is clocked anywhere between 67-76 GFLOPS using all 16 shader cores in parallel, and 4.7 x 16 is ~75

Then doesnt it seem like SETI is only using 1 of the cores for its particular programs?

This is more plausible than e.t. life less intelligent than humans any day in my opinion...

The GFLOPs given by BOINC is the theoretical number insisted by the marketing departments of GPGPU manufacturers and should not be used as a barometer for actual performance.

In fact, the CUDA app at SETI was written by nVidia themselves, and was designed from the ground up to be massively parallel. Keeping in mind that different mathmatical calculations can have varying affects on the performance of a given piece of silicon and its chosen design, any indication of a GPGPU failing to meet its expected marketing number should be seen as its real world performance.

Get GPU-Z
http://www.techpowerup.com/gpuz/

Check the GPU load (one SETI CUDA task uses all the GPU cores it can/need)

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

i have so many mixed messages and such about the GFLOP power of the YiForce 210, now the MESSAGES tab in BOINC reports my GPU as having 45 GFLOPS Peak power. This number has been 67 GLFOPS and also 76 GFLOPS. This leads me think that something is funky and if Nvidia is publishing this statistic, then something is even more fishy, but I tried GPU-Z and it is very nice.

The only problem is now that SETI is down, I completed all my CUDA tasks before you recommended GPU-Z!!!!

thanks for the recommedation anyway! crysis puts my system at 100% all around non-stop.

thanks asus for their square capacitors that reach 202F degrees instantly.