Is there a performance database?

[Lord_Humungous]

For those of you who are pretty much forced to run a wndows box because their favorite games won't run with linux(or are too lazy to make them work), but don't want to install OfficeBloat97/2000/XP -

http://office.microsoft.com/download.../xlviewer.aspxoddslot

[Complex33]

Hey Xyzzy can you put up the htm version I sent you then there will be no worries :)

Hopefully I can find some time to clean it up a little and make it a bit more presentible and useful but alas I'm in the middle of moving and getting back into the swing of school (read meager engineering graduate life)

So these two atoms were walking down the street...

Complex

[Xyzzy]

Originally Posted by Complex33

Hey Xyzzy can you put up the htm version I sent you then there will be no worries :)

http://www.teamprimerib.com/htm/Prime%20Processor%20Relations.htm

[Angular]

Originally Posted by ebx

CPU clock isnt the only factor. L2 cache size/speed, memory type,
speed will all make a difference. A 166DDR AMD XP 1800+ beats
a 133DDR 2100+ by a big margin.

Do you have those numbers handy?

On the benchmark page the following numbers are listed for the 256 FFT:

AthlonXP 2100+ 1733 133DDR 256 Full 0.026
AthlonXP 1800+ 1533 166DDR 256 Full 0.021

I do not believe that running Async FSB and Memory buses are possible on today's Athlon chipsets. Correct me if I am mistaken. If the 1800+ was not specially multiplier unlocked from 11.5x then the processor would be running at 1909 MHz. Perhaps this computer was running at 10.5x or 11x for 1743 to 1826 MHz. If is was unlocked to run at 1533 MHz it would need a 9.23x multiplier :(

Something looks fishy here right? Since the overclocked benchmark page has:
AthlonXP 1800+ 1533 1742 218DDR 256 Full 0.022

Is seams like an Athlon at 1533 MHz w/DDR333 is 10% faster than a 1733Mhz w/DDR266:

AthlonXP 1500+ 1333 1625 171DDR 256 Full 0.023
AthlonXP 1500+ 1333 1525 152DDR 256 Full 0.024


It would be nice to get the Chipset listed in the benchmark tables... Memory timing settings also may have a noticeable effect. Could we add more detail on the benchmark pages?

[Tasuke]

Most people when they OC an AXP, try to set the multiplier as low as possible, in order to get the highest FSB. Same goes for a ddr P4.

I currently have a DDR P4 running at 152QDR with 210DDR ram, and WOW is it fast. I also have a P4@133QDR 133DDR, I will try to see if I have time to bench them tonight. They are both running near 2.4Ghz.
One is a 1.6A the other is 1.8A.

But you have to realize that the clock speeds need to be Sync-ed in order to isolate Bus/ram as a factor, and they would have to be on a similar platform.

Since both my computers are on the same Motherboard, I can have a like/like Benchmarking score, and see if the 80MhzDDR memory tweek was worth it. I normally would have by now, but I have been swamped at work of late.

[ebx]

Originally Posted by Angular

Do you have those numbers handy?

On the benchmark page the following numbers are listed for the 256 FFT:

AthlonXP 2100+ 1733 133DDR 256 Full 0.026
AthlonXP 1800+ 1533 166DDR 256 Full 0.021

No this is a run-to-spec box. No OC or anything. CL2 Corsair 512M. GByte 7VXRP.
I never got around to OC it. FreeBSD 4.6.

Last night, I ran the benchmark a few time with mprime running in the background. I was
still getting 0.022. I will stop mprime tonight to get a full set of data and post here.

By the way, not a bug but the benchmark results state table size 892k instead of 896k.
Should be an easy fix.

[ebx]

This is a M$ free box so no CPUID. Here are the raw data.

Your choice: 18


Your timings will be written to the results.txt file.
Compare your results to other computers at http://www.mersenne.org/bench.htm
That web page also contains instructions on how your results can be included.

Timing 90 iterations at 256K FFT length. Best time: 21.154 ms.
Timing 72 iterations at 320K FFT length. Best time: 27.275 ms.
Timing 60 iterations at 384K FFT length. Best time: 34.632 ms.
Timing 51 iterations at 448K FFT length. Best time: 39.483 ms.
Timing 45 iterations at 512K FFT length. Best time: 43.787 ms.
Timing 36 iterations at 640K FFT length. Best time: 56.475 ms.
Timing 30 iterations at 768K FFT length. Best time: 68.735 ms.
Timing 25 iterations at 892K FFT length. Best time: 80.682 ms.
Timing 22 iterations at 1024K FFT length. Best time: 92.099 ms.
Timing 18 iterations at 1280K FFT length. Best time: 123.366 ms.
Timing 15 iterations at 1536K FFT length. Best time: 146.933 ms.
Timing 12 iterations at 1792K FFT length. Best time: 177.399 ms.


Hit enter to continue:

[xtreme2k]

I do not believe that running Async FSB and Memory buses are possible on today's Athlon chipsets. Correct me if I am mistaken. If the 1800+ was not specially multiplier unlocked from 11.5x then the processor would be running at 1909 MHz. Perhaps this computer was running at 10.5x or 11x for 1743 to 1826 MHz. If is was unlocked to run at 1533 MHz it would need a 9.23x multiplier

Memory can be run at Async mode on the KT333 based motherboards. You can have a 266MHz FSB while running the memory at 333MHz. You can ALSO unlock your AthlonXP, reducing the multiplier and run EVERYTHING at 333MHz (this will give best performance as the FSB itself is a bottleneck).

[Angular]

Originally Posted by ebx

No this is a run-to-spec box. No OC or anything. CL2 Corsair 512M. GByte 7VXRP.

So your FSB is the standard 266 DDR and only the Memory is running at DDR333?

I believe you, its just suprizing that the CPU can use Memory bandwidth faster than it recieves it from the Chipset (DDR controller).

[ebx]

Originally Posted by Angular

So your FSB is the standard 266 DDR and only the Memory is running at DDR333?

The truth is, my vendor put together the board and cpu for me. I didnt change a thing. They better not to OC anything for their own good.

The system controller (KT333) sits between the CPU and memory. Given enough circuits, I dont see why we could not run seperate rates at each side. As to how beneficial it is, I dont know.

I will see how far I can raise the fsb without upping vcore. Dont want to touch that since I am using the fan came with cpu. Had AMD not released the 2 new cpu today, I might do some serious OC on this 1800. But now I want one of the 2400 for under $200!

[Angular]

Perhaps the faster memory to chipset link reduces the latency that the CPU sees. Does the chipset have a memory buffer? It certainly does not increase the memory bandwidth. If an AGP or PCI card was needing lots of memory access, then it DDR333 w/266FSB may be more benifical.

Few big vendors offer Athlon systems... May I ask whom your vendor is?