DF and SMP

[pelligrini]

I'm not too sure about running it on an SMP machine. I've run into some strange and disapointing production numbers. Production on a hex and even a dual is crap. Yesterday I noticed that my hexes are doing poorly. Over 16-18 hrs my hexes produced around 8k per client, my dual P-Pros did about 12-13k, and my single P-Pro boards did 26-28k. The real pisser is that the regular P-Pros are runniing at 233Mhz, 3 of the Hexes have 333Mhz PIi overdrive chips, yet the 233 tripled their output? Even a P200 Overdrive did 11.6k. On a one of the hexes, I stopped 2 clients so it will run just 4 instances, Production wasn't that much better. One of my other Hexes crashed over night as well as a dual. I haven't got time to play around with the affinity and such.

[pelligrini]

I just started some of the clients over on several of the machines in different configurations. Some with a single instance of DF (only duals) some with DF and G@H, some otherwise. I really like the G@H client because of it's evvicient use of processor time, the DF client doesn't seem to be as efficient.

[ohms18k]

What about turning the priority up on the DF client?

[pelligrini]

I don't see how priority would affect a machine running exclusively DF. Besides, I'm really getting tired of jacking with this client.

[pelligrini]

Afer around 2 1/2 hours of crunching: (all P-Pros @233 Mhz)
3300 - Single P-Pro, win98 (I don't have a single CPU machine under W2K)
2200 - Dual P-Pro w/ One instance, W2k-Pro
2700 - Dual P-Pro w/ One instance & 1 G@H, W2k-Pro
1500ea - Dual P-Pro w/ Two instances, W2k-Pro
1700ea - Hex P-Pro w/ 3 DF and 3 G@H, W2k Adv-server

These three Hexes are running PII-Overdrives @333Mhz:
2400ea - 2 instances & 3 G@H
2x1600 & 1800 - 3 instances & 3 G@H
2000ea - 3 instances of DF only

I think all my SMP machines are going to go back on G@H. Too many wasted cycles. At least the hexes are going, until I ca figure out WTF is going on with them.

[LBaker]

Pell I don't know that it has anything to do with your problems, but the server is down at the moment. :rolleyes:

[Chris Wolfe]

Hey pell, your farm is all kvm'd together, right? Why not run the G@H client priority "normal" and the DF screensaver on each? Best of both worlds isn't it?

[BeeKs]

I am about to bring another Dual (ASUS A7M266-D/xp1800+) on line this week, and was planning to shut down one on my Xeon rigs. Did you want me to test anything out before I either start the G@H (for the Dual AMD's) or before I shut down the Dual Xeons?

Where do I go to get the client?


I would not be caught dead with a single processor workstation.

The current farm
3 Dual PIII Xeons
2x720/2m
2x800/1m (Yes I said 800/1m)
2x560/2m (2 B retired And Sold-ASUS XG-DLS)

2xA7M266-d's
2xXP1900 @1727 On-Line
2xXP1800+ in dry dock.

This will only be a limited experiment. If you want, say the word and I will configure one or both systems for testing. just tell me how you wnt them set up





Originally posted by pelligrini
I don't see how priority would affect a machine running exclusively DF. Besides, I'm really getting tired of jacking with this client.

[pelligrini]

Thanks for the offer Beaks, but I don't think I could put your duals to much use. I've got a whole bunch of dual P-Pros, some of the exactly the same. What's needed is some single processor systems of the same makeup to refrence the SMP machines against. You know of anyone with a quad or 8way with the same chips as yours?

[BeeKs]

The Xeon World is very small. I wish I knew folks with a 4 way 500. I doubt any of my pals could afford a 4-way 700/2m, and AMD will not have a 4-way ready till the claw hammer or sledgehammer.

[Yo_Mama]

I know it's not as efficient, because my CPU get 3-4C hotter when running genome...


Originally posted by pelligrini
I just started some of the clients over on several of the machines in different configurations. Some with a single instance of DF (only duals) some with DF and G@H, some otherwise. I really like the G@H client because of it's evvicient use of processor time, the DF client doesn't seem to be as efficient.

[pelligrini]

What I've noticed is that the production rate of the client running multiple instances on a multiple cpu system is not as good as one client on a single processor box. The production falloff gets much worse as the amount of processors goes up.

I have three P3-1000 one by itself and two on a VP6, the production per CPU (one client each chip) was about 30% less on the dual setup. That's a 60% hit in production for using dual instead of single chip machines. There are many variables in the comparison, one's Win98 the other's Win2k. The single CPU is on a 115Mhz FSB, the Dual is at 145, the dual has 256MB of ram and the other is 384.

After about 12 hours:
Single - 55006
Dual - 38680, 39900

I did some more tests with 4 different dual P-Pro ALR boards.
All w/ 2x 512k cache P-Pro200 @233, 2x 64mb SIMMS, Win2k-Pro
After about 12hrs of crunching:
ALR1 - 10953 (1xDF & 1G@H)
ALR2 - 7103 & 7000 (2xDF)
ALR3 - 11800 (1xDF)
ALR4 - 8128 & 8213 (2xDF, w/ affinity set)

That's about a 20-30% reduction in production per SMP client.

[phil]

Hmmm, that's interesting Pell....could this be a limitation of Intel's shared bus SMP implementation? The actual amount of data being moved over the memory bus is saturating it when running on multiple CPU's. A single client is not enough to saturate the bus.

How about a comparison using a dual vs single AMD machine? The AMD SMP implementation has a seperate memory bus per CPU I believe....this may help if data saturation is happening in Intel SMP systems.

[pelligrini]

Someone mentioned the Intel memory thing over on the DF forum. Here's the thread. (http://www.free-dc.org/forum/showthread.php3?threadid=502)

I still think it's more to do with the system overhead of the client. When I run a single instance of DF on a dual CPU usage according to the taskmanager is 65-70%. I've also noticed in Taskmanager that the "System" and "CSRSS.EXE" take up some total CPU time too.

I guess I'll have to build a single AMD XP-1800 to do some tests. :rolleyes: :D

[pelligrini]

ulv's post on the DF board seems to dis-prove the memory thing.

[quote]Tyan 2466MPX- 2 x XP1800 (1,53 Ghz): 165000 on each,
Abit KG7- Tbird 1,4Ghz: 206000.
Same memory (512Mb- same brand), same videocard, started within 1 minute.[/url]

That's about a 25% decrease, like my Intels.

[BeeKs]

DF, looks like a good program, but I think a 20% productivity hit before I even start is a bit much. I am watching the DF project close because it would be kinda dumb not to keep an open mind. However, a few of us are in a catch 22. As it is with your core members, the actions of some members require more thought towards the larger picture.

thx

Bob

[BeeKs]

Too Funny

404 ERROR:

Three things are certain:
Death, taxes, and lost data.
Guess which has occurred.