Everyone - please post OGR benches from 460 client here - for speeds page :)

[amok]

aberant, I can't remember if you were around when that change was made, but it was mostly concerned with DES, not RC5 . We didn't see any improvement in RC5 with the MMX extensions added to the client. They introed that just before the last DES contest.

 

[Leo]

[Jun 24 22:14:56 UTC] Benchmark for RC5 core #2 (RG class 6)
0.00:00:16.16 [2,660,227.95 keys/sec]
[Jun 24 22:15:15 UTC] Benchmark for OGR core #0 (GARSP 5.13)
0.00:00:16.09 [5,295,524.07 nodes/sec]


P3-700 oced to 933 mhz on a tyan s1854 mobo.

 

[Leo]

Just wanted to add... Awesome page you have there Mika. I've been interested in what a G4-500 would get since a friend(irrational machead) just recently showed me a benchmark showing a G4-500 beat a p3-1000 in photoshop on some filters...

Now I can rest assured that my p3-933 isn't p1mped by a 500 mhz g4 in Seti or Rc5

 

[strates]

Amok, The guys at d.net did recently release a client that was MMX enhanced in RC5. Although the DES calculations did benifit greatly from MMX, the clients were MMX enhanced when they were released.

 

[Kwatt]

P223 MMX
[Jul 15 11:19:47 UTC] Benchmark for OGR core #0 (GARSP 5.13)
0.00:00:16.15 [1,049,497.33 nodes/sec]

 

[Mad Pierre]

I;ve Got [ 4463341.13 nodes/sec ] out of a PIII 650 MHz , and i've got 2485500.00 out of a PIII450 . It look like this OGR likes a fast L2 Cache. Update the PIII450 is now geting 3066500.33 nodes/sec , I had two differant clients on the PC's> so OGR is no L2 Cahce depantant

 

[Eug]

[Jul 15 17:30:24 UTC] Completed OGR stub 24/8-3-42-2-13 (3,238,183,055 nodes)
0.00:08:45.46 - [6,162,521.22 nodes/sec]

463 client. System below. BX 112 MHz Cas 2. Boots at 897 so that makes 6870 nodes/s/MHz

 

[wildhagen]

dnetc v2.8010-463-CTR-00071214 for Win32

[Jun 15 17:44:05 UTC] Benchmark for OGR core #0 (GARSP 5.13)
0.00:00:16.35 [2,260,483.02 nodes/sec]

Pentium-2, 333 Mhz @ 343 Mhz.

 

[Joe O]

I've summarized the OGR benchmarks from the above posts, and added 3 of mine as well. Here are the results in descending order of Nodes/sec/MHz:

Description @ MHz ========' Nodes/sec ========' Nodes/sec/Mhz
Celly 533A@897 @ 897 ========' 6162521.22 ========' 6870.15
PIII 650 MHz @ 650 ========' 4463341.13 ========' 6866.68
Pentium-2, 333 Mhz @ 343 Mhz. @ 343 ========' 2260483.02 ========' 6590.33
Athlon 600 @ 650: @ 650 ========' 4184365.13 ========' 6437.48
Celeron 300a/450, @ 450 ========' 2557661.16 ========' 5683.69
P3-700 oced to 933 mhz @ 933 ========' 5295524.07 ========' 5675.80
Celeron 300a/450, @ 450 ========' 2520000 ========' 5600.00
PIII 700 @ 700 ========' 3911675.13 ========' 5588.11
Celeron 366@550 @ 550 ========' 3046955.00 ========' 5539.92
PIII 700 @ 700 ========' 3875661.60 ========' 5536.66
PIII450 @ 450 ========' 2485500.00 ========' 5523.33
Celeron 433@468 Avg : @ 468 ========' 2575154.08 ========' 5502.47
PIII 620 (Katmai) @ 620 ========' 3380000.00 ========' 5451.61
PIII 450@556.OGR-long ==> @ 556 ========' 2969378.31 ========' 5340.61
PIII 450@556.OGR-short ==> @ 556 ========' 2946536.44 ========' 5299.53
K6-III @400 @ 400 ========' 2000000 ========' 5000.00
Cyrix 6x86 P150+@120MHz @ 120 ========' 594663.36 ========' 4955.53
K6 233 @ 233 ========' 1143099.00 ========' 4906.00
k6-3 450 @ 450 ========' 2176156.75 ========' 4835.90
P223 MMX @ 223 ========' 1049497.33 ========' 4706.27
K6-2 450 @ 450 ========' 2082565.00 ========' 4627.92
Pentium MMX @200 @ 200 ========' 840000 ========' 4200.00
P166@200 MMX:Long ==> @ 200 ========' 824979.65 ========' 4124.90
p233mmx: @ 233 ========' 959500.64 ========' 4118.03
p200-mmx @ 200 ========' 820954.87 ========' 4104.77
P166@200 MMX:short ==> @ 200 ========' 799082.39 ========' 3995.41
Pentium Classic 166 @ 166 ========' 602380.00 ========' 3628.80
Winchip C6 at 225 @ 225 ========' 736995.00 ========' 3275.53



I have these in an Excel Spreadsheet if anyone is interested. I can also give you a tab delimited txt file if you desire. E-Mail me at osieckis@nji.com

 

[Eug]

Heheh. Celeron II's rule.

One thing I've noticed though. With the client active, both Unreal Tournament and Quake III Arena slow slightly. Only 1 FPS each (which is pretty meaningless I know), but it's definitely there and reproducible.

 

[Eug]

By the way, just out of interest's sake, what is that PIII 700/933 running with? VIA mobo? CAS 3 memory? The scores seem a bit odd given the other PIII Cumine numbers in the list.

What chip features does OGR use?

 

[Jimbo]

[Jul 15 21:42:50 UTC] Benchmark for OGR core #0 (GARSP 5.13)
0.00:00:16.02 [3,743,923.79 nodes/sec]

[Jul 15 21:46:51 UTC] Benchmark for RC5 core #2 (RG class 6)
0.00:00:16.19 [1,877,442.48 keys/sec]

This is a Pentium III 667 on an 810e board (my daughter's machine)

Jimbo

 

[Eug]

[Jul 15 22:26:25 UTC] Benchmark for OGR core #0 (GARSP 5.13)
0.00:00:16.14 [6,199,089.91 nodes/sec]

[Jul 15 22:26:51 UTC] Benchmark for RC5 core #2 (RG class 6)
0.00:00:16.14 [2,533,753.15 keys/sec]


OK. I only just figured out how to do the benchmarking. The previous numbers I supplied above were from real jobs, not the benchmark, if that makes any difference.

[Edit] OK, Joe. The numbers for long and short OGR (run multiple times) ranged from 6.181 Mnodes/s to 6.203 Mnodes/s, which pushes the Nodes/s/MHz to about 6900. The previous real job was 6.163 Mnodes/s. [/Edit]

 

[Joe O]

Benchmarks run at "normal" priority, while the client runs at "idle" priority unless you overide it.

On an empty machine it should make no difference. On a fully loaded machine it will.
Empty==no other tasks running
Fully Loaded==Lots of other tasks running

 

[crYnOid]

well i would like to update my benchmark i am now getting
[Jul 16 04:56:16 UTC] Benchmark for OGR core #0 (GARSP 5.13)
0.00:00:16.15 [2,735,630.50 nodes/sec]
in the benchmark but i have got
[Jul 15 22:09:42 UTC] Completed OGR stub 24/5-9-4-20-1 (60,856,708,052 nodes)
0.05:37:07.83 - [3,008,562.60 nodes/sec]

this is with a k6-3 450 + latest client

 

[mechBgon]

For my Celerons at 552MHz:

[Jul 18 06:00:18 UTC] Benchmark for OGR core #0 (GARSP 5.13)
0.00:00:16.25 [3,108,542.03 nodes/sec]

 

[Joe O]

crYnOid,
Are you overclocking your machine? Have you run any program to confirm your processor speed? What motherboard are you using?

 

[NT4Mike]

PPro200 w/256k cache

DNETC V2.8010-463-CTR-00071214

OGR LONG BENCH CORE #0 (GARSP 5.13)

run 1-----1,356,362.10
run 2-----1,358,867.76
run 3-----1,358,031.57
---------------------------------------
1,357,753.81 nodes/sec avg.


-Mike

 

[Eug]

OK, second computer:

Celeron 366@457 on Soyo 6VBA+133.

3,120,062.26 nodes/s = 6827 nodes/s/MHz.

Not using it though, because I dunno how to properly sneakernet it without running around all the time.

 

[Joe O]

I've put a description of an approach to sneakernetting OGR in http://forums.anandtech.com/messageview.cfm?catid=39&threadid=199389

 

[Eug]

Hmmm... Joe O may be right. It seems that the people with the 463 are doing quite well. When you post your speed, please also post the client version.

 

[crYnOid]

Joe O: I am using the 463 client. The cpu is not overclocked, and i have since got score of

[Jul 16 22:58:28 UTC] Completed OGR stub 24/5-9-4-20-7 (55,282,379,688 nodes)
0.05:06:20.02 - [3,007,743.17 nodes/sec]

and

[Jul 17 02:07:57 UTC] Completed OGR stub 24/5-9-4-21-17 (34,208,508,226 nodes)
0.03:09:30.59 - [3,008,506.69 nodes/sec]

these units have been done overnight when my computer has not been in use so i think they are valid

 

[Viztech]

Please see this thread. http://forums.anandtech.com/messageview.cfm?catid=39&threadid=199643

viz

 

[TOOCOOL]

For the 2.8009-460-CTR-00060723 client for Win32,
my k6-2 550 is 2.9 m/nodes/sec

 

[Joe O]

crYnOid,
Thank you for the new numbers. Your results with the 463 client were so much better than others with 460 and 462 clients that I tried it on my machine. Viz tried it on his machines. We found anywhere from a 20 to 30% improvement with 463b over previous clients. I don't know when I would have gotten around to the new client if not for you. Thank You!

ps We've go a new thread going at http://forums.anandtech.com/messageview.cfm?catid=39&threadid=199643 So if you could post RC5 benchmarks for you processors here or there we would appreciate it.
pps You don't need to be running RC5 to do the benchmarks.