• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Dual Xeon w/Hyperthread

B

btomlins

n00b
Joined
Feb 10, 2004
Messages
33
Is the dual xeon w/hyper the way to go for hardcore folding? I have an AMD XP 2400+ and an Intel 2.4 w/hyper and started them folding both at the same time. The Intel has pushed out twice as many work units. Now with a 3.06ghz dually xeon, run 4 instances and basically add 12ghz to my farm saving $15 a month on power. Anyone have any thoughts they can add before I drop oh about $1200?

Oh yeah it won't be all for folding, would be my new file server, webserver, ftp server, game server. :D
 
That would be AWESOME. a xeon doesn't fold any different than a p4... but you can't run p4 dual. ... XP / MP fold very well, and are the clear winners in the "bang for the buck" catagory... but if you have the coin, yeah, dual xeon 3ghz would pretty much be increaible output for a single boxen. (you may consider high end MP chips and have almost the output for much less $$.
More power to you. that is very cool indeed. remember each instance of folding uses 20megs of ram (it could be 10, or it could be 20.. I like to play it safe) so remember that in your RAM purchase.

wowie. I'd like to have that in my house.
 
Up until last Friday, I had a Duallie Xeon 2.0 with 512MB RAM... I only ran one instance on each processor though... I wasn't sure how long I'd have it ("fixing" it for a co-worker's 12-yr old son)... I ended up having it for almost 2 months so it was a blessing for my stats :D but unfortunately, I had to give it back... I kept good track of it and my P4C 2.4 @ 3.0 in an excel spreadsheet... If I remember my averages right, the 2.4C@3.0GHz got about 4.25 pts/hr... and each Xeon (one instance) got about 3.75 pts/hr... But I guess that's the difference between the 400 vs 800MHz bus, and (maybe?) the cache involved...

I've been pricing a duallie to "replace" the borrowed one, and I've been leaning more towards an Athlon MP duallie... it's cheaper... however, there are much more dual Xeon and/or Quad Xeon boards out there... If you've got the $$ to throw at it, the hyperthreaded Xeon may be the way to go for you... Why don't you just get one of each :D and compare? ;)
 
well the MP dually will only handle 2 instances of folding. Where the P4 dual w/hyper can handle 4. I have a copy of windows 2000 Advance Server that I can toss on there so it can see 4 processors. Only bad thing about getting the Dual P4 Xeon is that it requires registered RAM and from the intel board I was looking at 4 sticks of RAM min.
 
Originally posted by btomlins
Is the dual xeon w/hyper the way to go for hardcore folding? I have an AMD XP 2400+ and an Intel 2.4 w/hyper and started them folding both at the same time. The Intel has pushed out twice as many work units. Now with a 3.06ghz dually xeon, run 4 instances and basically add 12ghz to my farm saving $15 a month on power. Anyone have any thoughts they can add before I drop oh about $1200?

Oh yeah it won't be all for folding, would be my new file server, webserver, ftp server, game server. :D
Click to expand...

1st I'd say that you don't have a flag set or something if the 2.4 Intel is pushing out twice the work. That 2 GHz XP should be at a minimum matching what the Intel does, and in reality beating it. If you set two flags when launching the XP CPU's, it will work faster. -advmethods and -forceSSE should do the trick.

As for spending $1200 on a 3 GHz Xenon ... seems like a waste if all you are going to do is Folding. Currently I have 3 Dual AMD's running (all using those flags) and I just picked up another Dual MB (just arrived yesterday). MB cost is about $140 or so... eBay auctions are great. I have converted some XP CPU's to look like MP's ... works great, and they are pumping out some serious work. Approximate cost should be less than $400 each if you watch your spending. Nice thing is, Duals generate less heat, use fewer electric dollars, and cost less over-all... I could set up at least 3 AMD Dual's for what you are wanting to spend on 1. No way that 1 system could keep up with what I could build for that. Something to think about. ;)
 
Dont go off the number of work units.
Work off the number of points per hour per Ghz.
That is a more accurate indication of how well your folding.
A good AMD should be running at just over 1.5 PPHPG.
A good Pentium at just under 1.5 PPHPG.

Plus Hypertreading only adds about 15% more work.
So you wont be adding 12Ghz. More like 7Ghz.

Luck.......... :D
 
the P4 2.4 is running two instances of folding, the XP 2400+ is only running 1, and I have all the flags set.

Now, 3 Dually AMDs....6 procs at about $110 a piece, 3 mobos, 3 cases, 3 sticks of ram, it would be close...

With a dual P4 Xeon w/Hyperthread, one case, one HDD, less space, pumps out the same work as 4 PCs...

I already have 7 PCs in my computer room at home, I am trying to add more folding power, move my server apps all to once machine, and also have the game server for the LAN parties I attend that can run numerous game servers at once.
 
Well, I don't want to waste time arguing about it, but a HyperThreaded CPU does not do the work of 2 CPU's, more like 1.25 at the most... Unless you sprinkle some magic dust on the CPU's, they don't work that way. Many here who are experts at hardware have tested this and that's the way it is. At best that setup would do the work of 2.5 CPU's ... not bad, but not super magical... :p

If you want I can post some links to some cool AMD MP stuff where 2200 cpu = $65 each, 512 ECC memory = $65 and a Tyan Dual MB for $120... all are used but guaranteed to work.

Either way, more CPU's of any type is a good thing. ;)
 
Originally posted by btomlins:

Is the dual xeon w/hyper the way to go for hardcore folding? I have an AMD XP 2400+ and an Intel 2.4 w/hyper and started them folding both at the same time. The Intel has pushed out twice as many work units. Now with a 3.06ghz dually xeon, run 4 instances and basically add 12ghz to my farm saving $15 a month on power. Anyone have any thoughts they can add before I drop oh about $1200?
Click to expand...
I have two HT P4 boxen that are clocked at least 3.4GHz. These machines are running two F@H clients each. For Gromacs WUs, the 3GHz+ HT machines are nearly equivalent in production to my dual AMD Barton machines @ ~2GHz. All else being equal, a dual Xeon @ 3GHz+ should be comparable to two dual Athlon boxen @ ~2GHz.

However, with Tinker WUs, it's a different story. AFAIK, Tinkers don't utilize any SSE optimizations. The P4 without SSE support although no slouch at high clockspeeds, is outcompeted by the Athlon. AMD processors have a better designed, more efficient FPU, and thus excel over their Intel counterparts in these scenarios.

If you add the flag to receive Gromacs WUs instead of Tinkers, and the flag to utilize SSE optimization, it will ensure that Intel based architectures are running at optimal rates of production from a stats standpoint. Personally, if the cost is comparable, I wouldn't hesitate to purchase a dual Xeon, and this is coming from someone who's running 3 dual Athlons and no Xeons -- yet. ;)
 
Well, here is what I see looking at info from Stanford and Statsman:
Code: 
APOLLO . . . . . . .  15 Procs - 661 pts last 24
BTOMLINS . . . . . .  15 Procs - 511 pts last 24
TheWeatherDude . . .   7 Procs - 517 pts last 24
I'm keeping my slow dual Athlons and you guys can have those fast Hyperthreaded Intel CPU's :D
I rest my case. :cool:
 
I run all Intel at work, except 1 MAC wich I just borg'd today. At home, I have 3200+, 2600+, 2400+, 1100 Duron, 750 Athlon, 850 Laptop, and P4 2.4

I take what I can get and fold on whatever I can.
 
Originally posted by LPerry:

Well, here is what I see looking at info from Stanford and Statsman:
Code: 
APOLLO . . . . . . .  15 Procs - 661 pts last 24
BTOMLINS . . . . . .  15 Procs - 511 pts last 24
TheWeatherDude . . .   7 Procs - 517 pts last 24
I'm keeping my slow dual Athlons and you guys can have those fast Hyperthreaded Intel CPU's :D
I rest my case. :cool:
Click to expand...
Um, not quite.

Some of those processors are sub-500MHz. The stats you displayed don't tell the story in its entirety. I only have two P4 processors. One can equally assume after knowing my network specs somewhat more intimately, that my choice in the favor of employing AMD over Intel was perhaps the primary reason I do not produce as much as I should... ;)

My personal observations are still valid. A HT P4 over 3GHz produces comparable results with Gromacs WUs to a dual Barton. Case closed.
 
Originally posted by APOLLO
Um, not quite.

Some of those processors are sub-500MHz. The stats you displayed don't tell the story in its entirety. I only have two P4 processors. One can equally assume after knowing my network specs somewhat more intimately, that my choice in the favor of employing AMD over Intel was perhaps the primary reason I do not produce as much as I should... ;)

My personal observations are still valid. A HT P4 over 3GHz produces comparable results with Gromacs WUs to a dual Barton. Case closed.
Click to expand...


Before I overreact, I just want you to clarify this rather bold statement:
A single P4 with HT over 3 GHz outproduces a dual Barton around 2200MHz?
 
Now is that a dual barton 2200+?

2200 actual mhz is a XP 3200+
Didn't know they made dual AMD systems like that, or I haven't seen anyone overclock a MP to run at those speeds.
 
Originally posted by btomlins
Now is that a dual barton 2200+?

2200 actual mhz is a XP 3200+
Didn't know they made dual AMD systems like that, or I haven't seen anyone overclock a MP to run at those speeds.
Click to expand...

No... if you take a standard Athlon MP 2800+ Barton 2.083GHz with a 266 mhz FSB it will easily overclock to 2200 mhz... more than capable of kicking some serious butt. ;) About $167 each... ;)
 
I've got a pair of 2400+'s at 17 multi and 138 FSB @ 1.75 volts on a MSI K7D for a nice 2x2346 Mhz.
Just not sable enough to fold at 140 FSB.
Must get around to overclocking the rest of my 2400+'s.

Luck......:D
 
Originally posted by mwarps:

Before I overreact, I just want you to clarify this rather bold statement:
A single P4 with HT over 3 GHz outproduces a dual Barton around 2200MHz?
Click to expand...
No, not exactly. My statement is that a P4 clocked over 3GHz with HT enabled running two clients is comparable to a dual Barton clocked over 2GHz with Gromacs WUs. It doesn't outproduce it.

I understand many on this forum have been devoted to AMD for many years, and so am I. The very last thing I wish is to instigate an AMD vs Intel flame war on the DC forum of all places. Since action always speaks louder than words, I'll post the following log files to better illustrate my point.

Dual Barton @2200MHz:

Code: 
Folding@home Gromacs Core
[03:51:14] Version 1.55 (December 22, 2003)
[03:51:14] 
[03:51:14] Preparing to commence simulation
[03:51:14] - Assembly optimizations manually forced on.
[03:51:14] - Not checking prior termination.
[03:51:14] - Expanded 327846 -> 1861229 (decompressed 567.7 percent)
[03:51:14] - Starting from initial work packet
[03:51:14] 
[03:51:14] Project: 858 (Run 212, Clone 3, Gen 0)
[03:51:14] 
[03:51:14] Assembly optimizations on if available.
[03:51:14] Entering M.D.
[03:51:20] Protein: p858_p53dimer858
[03:51:20] 
[03:51:21] Writing local files
[03:51:21] Extra SSE boost OK.
[03:51:21] Writing local files
[03:51:23] Completed 0 out of 2500000 steps  (0)
[04:16:51] Writing local files
[04:16:53] Completed 25000 out of 2500000 steps  (1)
[04:42:28] Writing local files
[04:42:30] Completed 50000 out of 2500000 steps  (2)
[05:08:08] Writing local files
[05:08:10] Completed 75000 out of 2500000 steps  (3)
[05:33:49] Writing local files
[05:33:51] Completed 100000 out of 2500000 steps  (4)
[05:59:58] Writing local files
[06:00:00] Completed 125000 out of 2500000 steps  (5)
[06:27:43] Writing local files
[06:27:45] Completed 150000 out of 2500000 steps  (6)
[06:55:49] Writing local files
[06:55:51] Completed 175000 out of 2500000 steps  (7)
[07:24:36] Writing local files
[07:24:38] Completed 200000 out of 2500000 steps  (8)
[07:53:19] Writing local files
[07:53:21] Completed 225000 out of 2500000 steps  (9)
[08:19:57] Writing local files
[08:19:59] Completed 250000 out of 2500000 steps  (10)
HT P4@~3.35Ghz:

Code: 
[09:32:58] Folding@home Gromacs Core
[09:32:58] Version 1.55 (December 22, 2003)
[09:32:58] 
[09:32:58] Preparing to commence simulation
[09:32:58] - Assembly optimizations manually forced on.
[09:32:58] - Not checking prior termination.
[09:32:59] - Expanded 328923 -> 1861229 (decompressed 565.8 percent)
[09:32:59] - Starting from initial work packet
[09:32:59] 
[09:32:59] Project: 858 (Run 52, Clone 0, Gen 0)
[09:32:59] 
[09:32:59] Assembly optimizations on if available.
[09:32:59] Entering M.D.
[09:33:05] Protein: p858_p53dimer858
[09:33:05] 
[09:33:05] Writing local files
[09:33:05] Extra SSE boost OK.
[09:33:05] Writing local files
[09:33:07] Completed 0 out of 2500000 steps  (0)
[10:01:29] Writing local files
[10:01:31] Completed 25000 out of 2500000 steps  (1)
[10:30:00] Writing local files
[10:30:02] Completed 50000 out of 2500000 steps  (2)
[10:58:31] Writing local files
[10:58:33] Completed 75000 out of 2500000 steps  (3)
[11:27:01] Writing local files
[11:27:03] Completed 100000 out of 2500000 steps  (4)
[11:55:33] Writing local files
[11:55:35] Completed 125000 out of 2500000 steps  (5)
[12:24:07] Writing local files
[12:24:09] Completed 150000 out of 2500000 steps  (6)
[12:52:31] Writing local files
[12:52:33] Completed 175000 out of 2500000 steps  (7)
[13:20:49] Writing local files
[13:20:51] Completed 200000 out of 2500000 steps  (8)
[13:49:07] Writing local files
[13:49:09] Completed 225000 out of 2500000 steps  (9)
[14:17:20] Writing local files
[14:17:22] Completed 250000 out of 2500000 steps  (10)
These log files show the first 10 completed frames of one client on each of the machines in question. The WU is a p858_p53dimer858 protein. These results were completed overnight while I was asleep, and therefore the machines should not have experienced any significant activity. The average time per frame for the Dual Barton is ~27min and the P4 is ~29min. My other P4@~3.5GHz is even closer in performance, but it wasn't working on the same proteins and could not be evaluated for this comparison.
 
Before I jump to any conclusions here, I'd like to see the comparison with the identical WU. If someone wanted to take one for the team and copy the same WU from one computer to another and compare frame times, then we'd be able to make the best apples to apples comparison we can - I've seen seemingly identical proteins run different times on almost identical machines.
 
Plus running 2 copies on each machine.

If both AMD & non-HT Intel are running at around 1.5 points per day per Ghz.
And HT gives you around 17.5% more work.
Then a 2Ghz AMD duallie = 1.5x2x2 = 6 points per hour.
And a 3.4Ghz Intel = 1.5x1.175x3.4 = 5.9925 points per hour.

Interesting.

Luck............:D
 
WHAT's FOLDING?

Couldsomeone point me to an information source? I'm using a dual pIII setup for a fileserver now and it's time to upgrade....(more)distributed computing, here we come?
 
Saving the world through distributed computing science.

100000.gif
60000.gif
 
You must log in or register to reply here.
Back
Top