win2k or win98se

[MechCD]

I have a puter that will readily accept win2k, but has win98se installed. Would i get a decent improvement if I upgraded it to win2k, using a FAT32 filesystem? I don't really want to repartition, but is NTFS actually faster? It probably doesn't matter that much for G@H, but this puter has a 20gb harddrive and could be a candidate for dumping CDs to and burning from

[zhotfire]

I'd guess that NTFS would be slightly faster as it's a more efficient file system, but i doubt you'd notice the difference over a defragged FAT32....

[FlybyKnight]

FWIW: Some time ago I had done benchmarks for GAH on identical pc's, using 98 and win2k.

windows 98 runs Genome at home 20-30% faster than on the same machine with Win2k. This difference was even greater on weaker machines (I.E a k6-2/450)

I love win2k and use it on some machines, but for GAH, 98 will
be significantly, more productive.

FBK

[Benjamino]

That's funny that you say that... I found 2K to run it faster. Maybe there are some hidden variables here. You think?

[MechCD]

I haveAdvanced. There is an option to prioritize background stuff, or applications and games. It'sin the sytem properties page. It doesn't even need a restart :D

[zhotfire]

Originally posted by FlybyKnight
FWIW: Some time ago I had done benchmarks for GAH on identical pc's, using 98 and win2k.

windows 98 runs Genome at home 20-30% faster than on the same machine with Win2k. This difference was even greater on weaker machines (I.E a k6-2/450)

I love win2k and use it on some machines, but for GAH, 98 will
be significantly, more productive.

FBK
Hmmm... phil did the same benchmark and came up with an opposite result... i know that if logged in as admin in win2k you will be chewing up resources that you wouldn't if logged in as a regular user... let me see if i can dig up the thread...

[zhotfire]

Hmmm... i seem to be confused! :o The only comparison that i can find that phil did was between Linux & win2k.... win2k winning the battle by a small margin. Gawd! I'm gonna have to compare for myself... all in the name of science, right? :rolleyes: :D

[zhotfire]

Ok, research underway with the latest version of genomespy. Running under win2k now, will switch over to win98 when i go to work this afternoon, then report back after work sometime after midnight Atlantic time. Anyone else care to undertake the endevour for the sake of comparison? :)

[zhotfire]

Ok... i've obviously done something wrong here... i'm getting the same result for both operating systems.... argh! Is there something i need to clear before switching to the other os? :confused: I'm running the same client(nonet), same gene, same machine(dual boot)....

[pelligrini]

I don't see why you think something is wrong. If the speeds are the same, they're the same. I really haven't noticed a difference between NT, 2k or 98se. I haven't done any serious testing between them though.

[zhotfire]

Originally posted by pelligrini
I don't see why you think something is wrong. If the speeds are the same, they're the same. I really haven't noticed a difference between NT, 2k or 98se. I haven't done any serious testing between them though.
Ummm... call me paranoid? :rolleyes: ;) :D

[MechCD]

Why in the world would logging in a Admin make a performance difference? The same services and apps are running when you are a regular user, but the regular user doesn't have access to everything

Could running G@H as a service makes a difference? I have it running like normal, with the window sitting there poopin out results :D

[Benjamino]

I ran it as a service for about a week... it always seems slower to me. That could have something to do with the fact that I can't actually "see" it crunching. :D

[Dave S]

speed dif not worth worring about but if you want somefing thats not gonna crash everytime you do somefing go W2K:)

[TDKozan]

Originally posted by Dave S
speed dif not worth worring about but if you want somefing thats not gonna crash everytime you do somefing go W2K:)

Amen.

Win2K is delightfully stable compared to 9x and Neanderthal Technology.

TK ;>

[zhotfire]

Well the intial results are in! For a 1g duron(morgan) @ 1.1ghz with 128mb pc100 ram @ 110mhz on an ecs k7s5a crunching a 61aa gene:

win98se = 6 hrs 04 mins per gene (very consistant)
win2kpro = 5 hrs 56 min per gene (average over 3 genes and dropping... lowest time on last gene = 5 hrs 50mins)

so according to theMath(tm) (if it holds at 5 hrs 50 mins).... that's an extra 1.11 genes(61aa)/13.71 woonits per week or 58 genes/713 woonits per year.

I'm going to run the gene under win2k for a few more till the times stabilize... i'll let y'all know what happens. :)

[Azzuron]

if you have time, why not try the same test without overclocking, some applications i have noticed react differently to overclocking, such as Half Life which cant run with my FSB at 110 mhz :\ grr ! I would just be curious if the same ratio pops out.

Ill run a test my self on my system. ill install win 98se, win me, win 2k and win xp... and give them all a run. thank the lord for partition magic :D Someone save me a very small fast gene :P I cant live without my system for long lol, so its gotta be able to finish the thing in the time im at school. that sbout 6-2 (8 hours) on average, some times longer. ill run first at 900 mhz then at 1000 see how the results turn out :)

[zhotfire]

4 more genes are in:

5 hrs 59 mins
5 hrs 58 mins
5 hrs 57 mins
5 hrs 58 mins

So it looks like the advantage i calculated should be cut in half... which means there is very little difference genome wise between win98se and win2kpro. Switch to win2k for just under a woonit/day? Of course! Progress is progress! ;) :D

[Bruce]

When the hardware is running pure code, it doesn't matter what OS started that code. The actual OS services (e.g.- Disk I/O) are a small enough part of the G@H process that nobody should make a decision based on that.

There should be virtually no difference between Windows OSs if you configure them so there is nothing else running. Differences may show up when you start using the computer for other things, but that's not easy to benchmark.

Comparing Windows to *nix, however, will give different results since Stefan used different compilers to generate the code.

[Azzuron]

I still think if we had a software engineer look at the asm...They could reallly speed up the calculations. im sure the complier put alot of stupid junk in the exe. Infact ill probably look at the new stuff they come out with when i get farther through school. see what i can dig around in. Maybe if i make good improvements, stefan will use it.... altho, who knows. I guess all he would need is the calculation algorythm in asm, the rest can still be C++ or whatever else is used... then just link the asm in.

[Bruce]

Originally posted by Azzuron
I still think if we had a software engineer look at the asm...They could reallly speed up the calculations. im sure the complier put alot of stupid junk in the exe. Infact ill probably look at the new stuff they come out with when i get farther through school. see what i can dig around in. Maybe if i make good improvements, stefan will use it.... altho, who knows. I guess all he would need is the calculation algorythm in asm, the rest can still be C++ or whatever else is used... then just link the asm in.

The fundamental calculation code is in fortran. It would probably be a real challenge to rewrite it in asm. Somewhere in the website faq there was a reference to who Stefan got it from.

[Azzuron]

hum. well it is compiled :) its already in ASM. just tweak it out. but you gotta be somethin good to do that... im not gonna take it on :P i dont hve any experience with x86.