AlphaServer Running Time

[xman]

This is running on AlphaServer with EV6.8 processor@1GHz (8MB cache)

------------------------------------------------------------
Distributed Folding Benchmark Script V1.0

Sample Size: 1015 structures over 1020 seconds.

Structures Per Second: 1.00
Structures Per Minute: 59.71
Structures Per Hour: 3582.35
Structures Per Day: 85976.40

OS - Running Kernel Version V5.1
------------------------------------------------------------

Much faster than my poor AMD MP 1.2GHz

[pointwood]

Is this with the new protein just released today? And are you doing the comparison with the same protein?

So far, it looks like the new protein is faster than the previous one.

Looks like those Alpha's are pretty fast :thumbs:

[xman]

The benchmark is running protein released today. The foldit file is using
./foldtrajlite -f protein -n native -qt -df -i f

I was doing comparison with the same protein with P3 1.26GHz and my poor AMD MP 1.2GHz. But found that Alpha faster than P3 faster than AMD.

[collin]

I concur with the observation that the current protein target is a fair bit faster than the older one - roughly 20% faster for all my machines.

I believe the df client is indeed cache-hungry given the benchmark of a 8mb-cache-monster ;) Is the P3 1.26 is a Tualatin with 512k cache? With a bigger cache, and prefetch capabilities, it would certainly be faster than MP1.2.

Having said that, the df client would certainly benefit from an increase in fsb, ram type (sdr<ddr<rdram) and chip's mmu/l2 capabilties.

For comparison, my P3-866(256k L2) with PC133 is doing 28 structures a minute and P3-450 (512k L2) is doing 17.7 structures a minute.

[xman]

Yes , you are right. The p3 is tualatin with 512K of cache, and with 1GB ram. Its result is competable with p4. However, it is still slower than alpha chip. And the alpha server i think is running DDR.