On my AMD machine Result duration correction factor is 2.9. On an Intel machine it is 0.49 Why is there such a huge difference? Is the application just not optimized for AMD processors?

Also, the AMD computer claims more credit per work unit than it is granted. Why is that?

[Bump]

On my AMD machine Result duration correction factor is 2.9. On an Intel machine it is 0.49 Why is there such a huge difference? Is the application just not optimized for AMD processors?

The application isn't optimised for a specific platform over another. I believe the RDCF is affected by the amount of time the machine is not running the project and so isn't comparable between machines unless they're both running rosetta the same amount.

Also, the AMD computer claims more credit per work unit than it is granted. Why is that?

This is because the claimed credit is based on the benchmark which is too basic to give a realistic measure, whereas granted credit is based on work done. AMD CPUs have strong FPU performance which inflates the whetstone benchmark score artificially. The claimed credit can be ignored...

---

On my AMD machine Result duration correction factor is 2.9. On an Intel machine it is 0.49 Why is there such a huge difference? Is the application just not optimized for AMD processors?

The application isn't optimised for a specific platform over another. I believe the RDCF is affected by the amount of time the machine is not running the project and so isn't comparable between machines unless they're both running rosetta the same amount.

Also, the AMD computer claims more credit per work unit than it is granted. Why is that?

This is because the claimed credit is based on the benchmark which is too basic to give a realistic measure, whereas granted credit is based on work done. AMD CPUs have strong FPU performance which inflates the whetstone benchmark score artificially. The claimed credit can be ignored...

That still doesn't explain why a 2.0 Ghz Core2 is ~200% faster than a 1.6 Ghz Turion X2.

That still doesn't explain why a 2.0 Ghz Core2 is ~200% faster than a 1.6 Ghz Turion X2.

where are you getting your numbers for that from?

your Turion is getting around 25 credits per 10k-seconds and your Core2 is getting 40 credits per 21.7k seconds (both per-core). That's 9 credits per hour against 6.6 credits per hour (sample size of 1 on the core2...).

---

That still doesn't explain why a 2.0 Ghz Core2 is ~200% faster than a 1.6 Ghz Turion X2.

where are you getting your numbers for that from?

your Turion is getting around 25 credits per 10k-seconds and your Core2 is getting 40 credits per 21.7k seconds (both per-core). That's 9 credits per hour against 6.6 credits per hour (sample size of 1 on the core2...).

934880 Core 2 Duo @ 2.0 Ghz: 51.49 credits for 10,547.06 seconds
934895 Turion x2 @ 1.6 Ghz: 22.25 credits for 10,147.02 seconds.

WTF?

934880 Core 2 Duo @ 2.0 Ghz: 51.49 credits for 10,547.06 seconds
934895 Turion x2 @ 1.6 Ghz: 22.25 credits for 10,147.02 seconds.

WTF?

Those numbers look pretty close to normal to me. The core 2 architecture is one of the most efficient out there right now. Plus there is a substantial clock speed difference as well. I expect you will find similar differences at your other projects once you have some returned tasks, excluding SETI if you use optimised applications.

---

BOINC WIKI

BOINCing since 2002/12/8

there's natural variance in the time taken too - you need a much bigger sample size than 1 to test it.

---