1Ghz P3 Slot 1 cpu in pc currently with 500mhz cpu Slot 1??

Cuauhtemoc

These processors run different voltages 2.0v(500mhz) 1.7v(1ghz)

Can i adjust the voltage on the mobo to take the 2 1ghz cpu's instead of the 2 500mhz ones i have currently.

The motherboard is a dual cpu Intel 440BX AGPset with a max cpu speed of 700mhz (according to spec).

Are there any other compatibility problems that i'm not aware of??

Cheers.
______
C.

RicardoSmith

Would depend on the motherboard make and if theres a bios for the motherboard that supports the new cpus. Otherwise PowerLeap do adapters that would allow you to use them. Would it be cost effective though? Might be better selling the duallie and and getting a 2ghz+ single cpu system. It would be faster.

Cuauhtemoc

Cheers Ricardo, you're probably right.
Only place i can seem to find these 1ghz cpus is ebay uk. Seem to be rare enough and not that cheap.

____
C.

RicardoSmith

A cheap 1800xp or P4 1.8ghz would run rings around a Dual 1Ghz in any task too. Unless it was a dual threaded server application that was dealing with lots of small computational tasks and lots of I/O requests, like a mail server/router or something. Even then I think faster single CPU machine might be quicker. You'd have to find some server benchmarking to know for sure though. Lots of people seem to be using those dual 1.4 PIIIs with the large cache in servers. They most have good reason for it. Maybe their a lot cheaper than a dual Xeon box and maybe most won't risk a dual AMD in a critical server. What you're usage going to be?

Cuauhtemoc

eh dunno yet.
Picked up a few pc's from my last job(they're closing down so they were mostly for nothing), an optiplex 240, 2 dell gx1s and the precision 410 workstation with the dual processors.
Was going to patch them up and mess around with a home network or something(sad but anyway) and then maybe sell them on.

RicardoSmith

You'd probably be better selling them for a few bob and putting the money towards something else. Unless you need a home network for something....

Cuauhtemoc

Yep. Just sold the Optiplex 240 for €400 which coincedentally is the cost of my esat bb connection and rental.

po0k

I might be interested in the dual 500Mhz box.

Any more details you could PM to me?

Need one for a game+file server for GalwayLAN so she'd be getting a good home

Gerry

Ricardo, a dual machine is far more responsive, in both desktop and server environments. It never gets bogged down, even a 4ghz p4 will stall on you if one program hogs cpu or goes into an infinite loop or something. I've used a dual xeon 2.2ghz with a 10k scsi drive as my desktop machine, and I have to say it was very noticeable. Even though it was doing many strenous tasks at once, it always flew along.
Cuauhtemoc, before upgrading the cpu's, you'd need to check that your motherboard can supply a lower voltage, I'm sure it can but just make sure. Also check that it has coppermine support. Its unlikely to have 133mhz support ( unless its the asus p2b-d ) so you will be stuck to using p3 800's or 850's. Though you may be able to clock them up a little. Depends on the motherboard, please post up the make and model here if you want more help. Also bear in mind that a dual p3 500 is a very snappy machine, if you don't play games.

RicardoSmith

I think I made my point earlier. It depends what you are doing. If you were rendering or number crunching or gaming as you said, a single CPU which is faster would run rings around it. Server applications or I/O as I said earlier could be quicker. But if you have a fast I/O on a Single CPU machine it could well be faster than the dual if its I/O is slower.

We had a whole bunch of dual Pentium Pros 200s and PIII 400/500s in one place I worked and nothing we did was quicker on them. That was mainly running server applications on them, rendering and even as a scanning station. A P3 1Ghz or P4 1.5 simply blew them away speed wise. The duallies were stuffed with Ram too. Even a single 500 was faster than the dual Pentium Pros 200s.

A dual xeon 2.2ghz with a 10k scsi drive is a completely different beast. Technology wise and price range. I don't get your point at all about that at all or how it relates to what we are talking about. A machine with two already very fast CPU's probably 512/1mb cache and a SCSI system and a lightening fast drive is very fast. Cost probably 3-5K. If it wasn't very fast there'd want to be something seriously wrong with it. Even if there was only one CPU in it, it would be pretty "darn" quick.

Thats said a 3.2 cpu of the same spec would still whip it in 90% of tasks.

"if one program hogs cpu or goes into an infinite loop or something" ...That generally causes a crash of the OS. Having two cpus won't stop that.

Theres only specific things that a dually is good at. Granted it can feel quicker in the GUI. But as soon as you hit a non dual threaded app then a faster single will beat it. The benchmarks all prove this over and over again. Which is why a faster single P4 usually beats a Dual G4 or Dual Althon in many of those kinda benchmarks.

RicardoSmith

BTW back to the original post for some reason I though that board was a dual 370, not a dual Slot 1. I don't think you can get an slot 1 adapter. If the bios on the board doesn't support the new one then it won't work with it.

Gerry

Originally posted by RicardoSmith
I think I made my point earlier. It depends what you are doing. If you were rendering or number crunching or gaming as you said, a single CPU which is faster would run rings around it. Server applications or I/O as I said earlier could be quicker. But if you have a fast I/O on a Single CPU machine it could well be faster than the dual if its I/O is slower.

We had a whole bunch of dual Pentium Pros 200s and PIII 400/500s in one place I worked and nothing we did was quicker on them. That was mainly running server applications on them, rendering and even as a scanning station. A P3 1Ghz or P4 1.5 simply blew them away speed wise. The duallies were stuffed with Ram too. Even a single 500 was faster than the dual Pentium Pros 200s.

A dual xeon 2.2ghz with a 10k scsi drive is a completely different beast. Technology wise and price range. I don't get your point at all about that at all or how it relates to what we are talking about. A machine with two already very fast CPU's probably 512/1mb cache and a SCSI system and a lightening fast drive is very fast. Cost probably 3-5K. If it wasn't very fast there'd want to be something seriously wrong with it. Even if there was only one CPU in it, it would be pretty "darn" quick.

Thats said a 3.2 cpu of the same spec would still whip it in 90% of tasks.

"if one program hogs cpu or goes into an infinite loop or something" ...That generally causes a crash of the OS. Having two cpus won't stop that.

Theres only specific things that a dually is good at. Granted it can feel quicker in the GUI. But as soon as you hit a non dual threaded app then a faster single will beat it. The benchmarks all prove this over and over again. Which is why a faster single P4 usually beats a Dual G4 or Dual Althon in m
any of those kinda benchmarks.

I didn't dispute that a faster single cpu would beat a dual machine in single threaded applications.
One program hogging the cpu shouldn't cause an os crash. At worst it should cause an application crash. On a dual machine, while this app is crashing, the machine will still be responsive. I'm sure you've seen internet explorer hog the cpu because of some script it doesn't like.
Also, why should an infinite loop crash the os? Unless its in the os itself. Code up an infinite loop in the language of your choice. If it runs as a high priority task, it will appear as if the os has crashed, but its just waiting for that single cpu to be freed up. My post was purely about the faster response on a dual machine. I have used faster machines than the 2.2ghz xeon, built plenty of them.

Nietzschean

Originally posted by RicardoSmith
I think I made my point earlier. It depends what you are doing. If you were rendering or number crunching or gaming as you said, a single CPU which is faster would run rings around it. Server applications or I/O as I said earlier could be quicker. But if you have a fast I/O on a Single CPU machine it could well be faster than the dual if its I/O is slower.

We had a whole bunch of dual Pentium Pros 200s and PIII 400/500s in one place I worked and nothing we did was quicker on them. That was mainly running server applications on them, rendering and even as a scanning station. A P3 1Ghz or P4 1.5 simply blew them away speed wise. The duallies were stuffed with Ram too. Even a single 500 was faster than the dual Pentium Pros 200s.

A dual xeon 2.2ghz with a 10k scsi drive is a completely different beast. Technology wise and price range. I don't get your point at all about that at all or how it relates to what we are talking about. A machine with two already very fast CPU's probably 512/1mb cache and a SCSI system and a lightening fast drive is very fast. Cost probably 3-5K. If it wasn't very fast there'd want to be something seriously wrong with it. Even if there was only one CPU in it, it would be pretty "darn" quick.

Thats said a 3.2 cpu of the same spec would still whip it in 90% of tasks.

"if one program hogs cpu or goes into an infinite loop or something" ...That generally causes a crash of the OS. Having two cpus won't stop that.

Theres only specific things that a dually is good at. Granted it can feel quicker in the GUI. But as soon as you hit a non dual threaded app then a faster single will beat it. The benchmarks all prove this over and over again. Which is why a faster single P4 usually beats a Dual G4 or Dual Althon in many of those kinda benchmarks.

would whip it in 90% of tasks? I'm sorry but i don't quite agree with you, unless you are doing some serious high cpu requiring single threaded app the dual cpu makes quite a big difference. From the original post it sounds more like he would use the box for miscellanious tasks, Dual cpu can be quite and for this as pointed out explorer can't drag u down to a crawl by taking all the cpu. In a single task a faster cpu is better, but the last time i was doing a single task i can't remember(and this is not a server enviroment i'm talking about) a few copies of explorer, winamp and other such things, this is windows, stuff breaks often eating up a cpu...dual cpu(or not quite as good but still handy hyper-threading that i have) makes quite a large difference to every day applications......

RicardoSmith

I don't think anyones memory is that short that you need to quote my post in its entirety, twice, right after the original post.

If you have an endless loop or an IE event that is hogging the cpu, it doesn't hog just one cpu it hogs both of them. It uses all the processing power it can. You can't cherry pick which CPU it runs on.

If an app its dual threaded it will use both CPUs at the same time and if it isn't then it will use 100% of one and then start using the the other one until its at 100% aswell. I would assume that IE is dual threaded. Apps like winamp aren't. If you ran 10 copies of winamp you don't get 5 on one cpu and 5 on the other. You will get 10 on one cpu until it hits 100% then it will start using the the other CPU. Look at the CPU utilisation graphs if you don't believe me.

What makes it seem faster is that some of the OS tasks like I/O are dual threaded so they respond slightly quicker. If some thing crashes it doesn't just crash on one CPU it crashes the OS. Even though parts of the OS is dual threaded, it manages all tasks and the division of threads. So if it crashes it doesn't matter one iota if the machine has one cpu or 80. Its outta there.

Theres a reason most machines aren't dual processor. Because in the majority of applications the performance gained is not worth the extra cost, of the hardware or the development of the multi-threaded software in the first place.

Nietzschean

Originally posted by RicardoSmith
If you have an endless loop or an IE event that is hogging the cpu, it doesn't hog just one cpu it hogs both of them. It uses all the processing power it can. You can't cherry pick which CPU it runs on.

In my experence that would be completely inaccurate, IE hogs up one cpu and thats it. Other cpu just run's everything. And you were saying about winamp, its easy to set the affinity of the application so 5 copies would run on each cpu.


And i have looked at the cpu utilisation graphs, it does exactly that.

And General System Lockups i find are a hell of alot less frequent on the multiple cpu machines. Also beyond windows, if you take a look at the linux box i'm using now, i've several times come across mozilla after die'n some how and started hog'n 100% of the cpu, found 3 cpu's like this and i can still use the machine no problem to login and kill all the processes. Much higher system stability imo.

Cuauhtemoc

Ooops..forgot about this thread...due to no money to buy bits for it. I think the board is an Intel 440BX AGPset the dell tag...RZ1PX for any more details.

If anyone has the highest speed processors that it takes plus working hdd and compatible ram(256mb dimms) i'd be interested. Will post this in the wanted forum later.

Its a dual slot 1 board and currently has 2 P3 500's in it. 1 the original that came with it.
From what i've seen of the dell's tho they can be a bit inflexible about adjusting things on mobo.

Syxpak the box has no hdd, ram or video atm. I doubt you'd be interested in it in its current state.
Once i have it up and running i'll probably post it for sale. Maybe with a retail version of xp pro if i'm feeling generous

Cheers for the help.

___
C.

RicardoSmith

Originally posted by Serialkiller
its easy to set the affinity of the application so 5 copies would run on each cpu

How do you do that on a Windows box?
Maybe its because you are running linux that its more stable? Never Linux on a dual cpu box myself.

Heres two links to add to the discussion.
http://www.digitalvideoediting.com/2002/07_jul/features/cw_macvspc2.htm
http://dbforums.com/t586281.html


Most Dells can take a standard intel bios instead of the Dell one. Though you do it at your own risk. But an intel bios may allow you to use newer CPUS than the Dell one. I flashed a GXa with a intel bios which allowed my to use newer cpus than the original Dell bios had.

Gerry

Originally posted by RicardoSmith
If you have an endless loop or an IE event that is hogging the cpu, it doesn't hog just one cpu it hogs both of them. It uses all the processing power it can. You can't cherry pick which CPU it runs on.

Sorry ricardo, this makes no sense. You seem to be saying that a program will split itself up dynamically if its hogging one cpu, and take over the other cpu(s). Its possible to bind some things to cpu's in windows, I know that some high end network cards allow you to specify which cpu handles the interrupts.

Originally posted by RicardoSmith
If an app its dual threaded it will use both CPUs at the same time and if it isn't then it will use 100% of one and then start using the the other one until its at 100% aswell. I would assume that IE is dual threaded. Apps like winamp aren't. If you ran 10 copies of winamp you don't get 5 on one cpu and 5 on the other. You will get 10 on one cpu until it hits 100% then it will start using the the other CPU. Look at the CPU utilisation graphs if you don't believe me. [/B]

You seem to have little understanding of threads. Again you seem to be saying that a single threaded application will spread itself across 2 cpu's. Its not a case of single or dual threaded, its more like single vs multithreaded. From what I've seen, IE starts up a new process every time you click the IE icon on the desktop. When you open a new window within IE, it seems to start a new thread. I say this, because if one of these windows crashes, it takes out all IE in its group. This is due to these threads sharing the same address spaces.
I believe IE uses additional threads as well to fetch different parts of the page in parallel, but these do not cause the crashes in general.

Originally posted by RicardoSmith
What makes it seem faster is that some of the OS tasks like I/O are dual threaded so they respond slightly quicker. If some thing crashes it doesn't just crash on one CPU it crashes the OS. Even though parts of the OS is dual threaded, it manages all tasks and the division of threads. So if it crashes it doesn't matter one iota if the machine has one cpu or 80. Its outta there.

Theres a reason most machines aren't dual processor. Because in the majority of applications the performance gained is not worth the extra cost, of the hardware or the development of the multi-threaded software in the first place. [/B]

What makes it faster is that the machine can actually do two things at once. In my experience, even on windows which has a reputation for instability, one application crashing does not bring down the whole OS. I'm speaking about windows nt based products here, which has has private address spaces for user programs, and also separation between OS and user address space. Normally windows nt/2k only crashes on me due to dodgy drivers. I get plenty of application crashes ( mostly IE ), but I don't have to reboot.

RicardoSmith

NO no no....did you read the links I posted?

I am saying that a singled threaded app will NOT use the second processor unless it needs to. A proper multithreaded app will ALWAYS use both processors. One works in series the other in parallel. If the singled thread app works in series and it needs 110% of a cpu it will use 100% of one and some will also use 10% of the other. Mosat only use 100% of one cpu. Actually its the OS doing this not the application. Whereas a dual threaded application will use 55% on one and 55% on the other.

BUT I'm saying the OS splits the tasks/processing. If a crash occurs it crashes in the OS and at the OS level. There is a point where the OS decides which thread is processed where. At this point the OS is not multihreaded. So its only running on one CPU.

You seem to think that a crash in one thread doesn't effect a thread on the other cpu. That could be true. But if the crash causes a problem with the OS at the point its not multi threaded then the OS will crash. If the App crashes it doesn't have to effect the OS. But this is the same for a single CPU as a dual machine.

The benchmarks and performance in those links illustrate only one thing. That a dual cpu machine is not faster due to having dual cpus unless the application is almost completely dual threaded. Theres hardly any such apps. Even the oft quoted photoshop isn't.

po0k

Originally posted by Cuauhtemoc
Ooops..forgot about this thread...due to no money to buy bits for it. I think the board is an Intel 440BX AGPset the dell tag...RZ1PX for any more details.

If anyone has the highest speed processors that it takes plus working hdd and compatible ram(256mb dimms) i'd be interested. Will post this in the wanted forum later.

Its a dual slot 1 board and currently has 2 P3 500's in it. 1 the original that came with it.
From what i've seen of the dell's tho they can be a bit inflexible about adjusting things on mobo.

Syxpak the box has no hdd, ram or video atm. I doubt you'd be interested in it in its current state.
Once i have it up and running i'll probably post it for sale. Maybe with a retail version of xp pro if i'm feeling generous

Cheers for the help.

___
C.

You'd be surprised

If you get bored of it I would appreciate a PM from you.
Would take it off your hands at the very least, if not make it do something useful (to me and the GL crowd/housemates/computer society).

RicardoSmith

Generally duals have a higher ram capacity. Stuffed with ram it would make a great webserver, email server, firewall etc.

Gerry

Originally posted by RicardoSmith
NO no no....did you read the links I posted?

Yep, one was a comparison of a dual mac versus a single p4. This was to prove your point about dual cpu's not being twice as fast as a single cpu, a point which I did not dispute.
Dual cpu's are very rarely twice as fast, this because of dependencies in the instructions which result in serial code, a reduction in memory bandwidth because there are now 2 cpu's fighting for the same bandwidth, arbitration and synchronisation between the cpu's, and a lack of parallelisation in most code.

Originally posted by RicardoSmith
I am saying that a singled threaded app will NOT use the second processor unless it needs to.

Ok, once more. A thread is a single flow of execution. It runs on ONE cpu at a time. If an app uses two cpu's, it has more than one process, or more than one thread.

Originally posted by RicardoSmith
A proper multithreaded app will ALWAYS use both processors. One works in series the other in parallel. If the singled thread app works in series and it needs 110% of a cpu it will use 100% of one and some will also use 10% of the other. Mosat only use 100% of one cpu. Actually its the OS doing this not the application. Whereas a dual threaded application will use 55% on one and 55% on the other.

The single threaded app will just use 100% of one cpu. It can only run on one cpu at a time. The concept of 110% doesn't really help matters. See above.

Originally posted by RicardoSmith
BUT I'm saying the OS splits the tasks/processing. If a crash occurs it crashes in the OS and at the OS level. There is a point where the OS decides which thread is processed where. At this point the OS is not multihreaded. So its only running on one CPU.
[/B]

Theres a difference between a crash in a user program, and an os crash. NT seems to be able to carry on ok after a program crashes ( say a segmentation fault ). But of course it crashes catastrophically if this happens to an os process, which is running in kernel space. This normally takes down the machine.

Originally posted by RicardoSmith
You seem to think that a crash in one thread doesn't effect a thread on the other cpu. That could be true. But if the crash causes a problem with the OS at the point its not multi threaded then the OS will crash. If the App crashes it doesn't have to effect the OS. But this is the same for a single CPU as a dual machine.
[/B]

Nope, if the threads are in the same process, it will be affected. Thats why I was talking about Internet Explorer, and the fact that when one window crashes, it takes all of them out, because they are threads ( if you started the windows with ctrl - N ). Yep, it makes no difference if this is a single or multi cpu machine.

Originally posted by RicardoSmith
The benchmarks and performance in those links illustrate only one thing. That a dual cpu machine is not faster due to having dual cpus unless the application is almost completely dual threaded. Theres hardly any such apps. Even the oft quoted photoshop isn't. [/B]

Applications which don't require much memory bandwidth often get major speedups. It also depends on the bus technology used. For example the Digital EV6 bus, which is used by AMD, offers each cpu a dedicated connection to main memory. This setup shows better scaling with 2 cpu's than the intel shared bus architecture.

Here's an example.

The opteron gets a 78% speedup with the addition of another cpu in lightwave, a popular 3d rendering program. The xeon gets 66%, which still isn't bad.

RicardoSmith

One of the links was between a single P4, a dual AMD and a G4. The tests were in applications that can take advantage of the dual cpus. You'd expect therefore the dual cpus to walk away with it. They don't. In fact the P4 is the quickest in some tests.

Second link is about SQL server running on a dual cpu machine. Basically is shows that SQL server though dual threaded is not maximising the performance of the dual cpus. So thats another consideration poorly coded dual threading really effects the performance.

So a dual cpu machine is not more crash resistant then?

If a machine has a faster memory bus, then it will be quicker. Lovely. Thats also a feature of single cpu machines aswell as dual cpu machines.

Ok rendering in 3D apps is one area where dual cpus are a big advantage. But rendering is something you do 10% of the time. So thats a lot of expense to have a dual machine to save 30% of rendering time of 10% of your work. Unless your a film studio or something.

In summary a dual cpu is slower than a faster single cpu machine in the majority of tasks and applications.

Ironically I actually do a lot of stuff myself where a dual cpu machine would be quicker. But most of the time I've bought a single cpu machine because a dual machine was too expensive.

Gerry

The point about the memory bus was not pure speed, but whether it is shared or not. AMD dual systems have a direct line to memory, they don't need to wait for the other cpu to finish reading. Whereas an intel dual system entails a lot of waiting around for the bus.

Regarding SQL server, well it was suggested that most of the operations were happening in a serial fashion. For a single sequence of related transactions, this will frequently mean that one thread can't read from the database until the other has committed a change to it..
Whereas if you have a load of unrelated transactions happening, the cpu's will both be able to run without waiting, and you will get both running at 100%.
I've seen this with oracle on a quad xeon box, we were frustrated by the fact that only about 1 1/2 cpu's were being used at a time. By rearranging the requests coming in, and cutting down on locking, we were able to get full cpu usage.

RicardoSmith

Originally posted by Gerry
...AMD dual systems have a direct line to memory, they don't need to wait for the other cpu to finish reading. Whereas an intel dual system entails a lot of waiting around for the bus

Didn't help it that much in the link I supplied. Whats the speed increase 5%-10% at best.

Originally posted by Gerry
...Regarding SQL server,....

I think thats demonstrates that its only very specific tasks specifically coded that benefit from the dual processing. Kinda narrows the scope of the advantages of multi cpus.

Then theres also another approach where you could throw more machines at the problem. So in the example you give, use a specific machine for the transactions (transaction servers) or even a number of them. In fact when single cpu servers are cheap it can make sense to have a whole bunch of them, one for the webserver, transactions, one for the DB etc etc. Even for rendering say in Lightwave. Anyone working on this kinda stuff will have a rendering slave machine or even a network of them.

It can make a lot of sense, if for the price of a multicpu machine you can have 2 or even 4 single machines to go that route. I thought it was interesting the route Overclockers.co.uk too with their seti farm. They built a stack of machines that all run from zip drives, and have the minimun amount of ram, a cheap cpu and board all networked. They had a built a buch of these low cost machines just to crunch seti. Hugely cheaper than building dual athlon machines. I'm not saying that there aren't scenerios where a multi cpu machines is not the best option. I'm just saying that for 90% of the time (maybe less) it isn't.

actually this is the spec they used...

PC Chips 810LMR Athlon motherboard
Athlon XP 1700
128Mb PC133 SDRAM
Taisol CEK733092 cooler
ATX 250w PSU
Internal 100Mb ZIP Drive
Zip Disk
Power Cable
1m network patch cable
and heres a mad photo of it .link

So looking at the original machine a dual 500, I think he'd be better selling it to some one who would use it as a server and build a cheap 1Ghz+ machine with the money. It would be more practical as a machine to muck about on. Even linux with something like KDE 3.1 takes a fair PC to make it useable

po0k

multi-cpu systems > single cpu systems, same platforms at same clock. FACT.
Nevermind what type of software you're running or any of that.
2 or more is better then 1 on it's own.

a P4 @3Ghz will kick the living shit out of a dual 8052 setup.

And any serious software should be coded to take advantage of culti-core/thread platforms, considering a decent workstation will be a dual setup.
There's a reason for this Ricardo and you could do well with heeding what's being said.
Similar reason why parallelism is the cheapest and quickest way to get more done in a shorter time then simply running a chip faster and faster.
Look at any of the major research iinstallations, "super computers" and clusters.
All based on the same principle of breakingone big job down into lots of smaller chunks which are processed at the same time over multiple cores/threads.

RicardoSmith

I'm not disputing that they can be quicker with the right software. Or that at the same clock they are quicker. Did I say that anywhere? I'm saying at the price of a dual you can buy a single cpu machine thats a heap faster in 90% of tasks. Case in point. Dual 500. For the cost of an upgrade to dual 1Ghz/1.4Ghz PIII's do you not think a single XP2000+ or P4@2.4ghz would wipe the floor with it in 90% of tasks and be a heck of lot cheaper? If not you've been inhaling too much of that artic silver. Heres another link ok this on IS more Mac biased but its a lot more relevent to the topic of this thread than talking about "supercomputers" and "major research installations". I'm mean he has a dual PII 500. Hardly on the same planet.
ww)

Incidentally I've a buddy running a Xeon 1.8 dual (SCSI) which is the fastest dual workstation I have had access too. But my P4 1.8@2.4ghz is faster than it in all of the graphics and multimedia applications I can do a comparision on. His machine cost about 3k mine well under 1k. He also has a GF2Quadro and I've a GF4ti4200. For working in big 3D models and big 2D images my machine is visibly a lot faster.

Cuauhtemoc

Hmm..I think i have all i need from this thread. Weighty as all the replies have been i think the best processors i could put in it from Gerry's point of view were the Slot 1 P3 800/850's, for some reason he reckons the 1ghz slot1 is crap. They seem to be very pricey as well.
I have a games machine already, this was just about upgrading this dual system as far as possible which from what i can work out would be Dual p3 800/850 and 1GB of ram and maybe scsi hdd if i felt like spending the money, but probably just stick with a standard hdd.
Cheers for all the replies.

___
C.

RicardoSmith

Those Intel 1Ghz slot1's have been in huge demand from the beginning. You#re not stuck to Slot1 CPUS though. You could possibly use two slockets which would allow you to use 370 PIII's (100fsb only probably) in it either. That would give you a wider range of CPU's to try. PCPRO did this with a dual slot1 board they had/have. Another approach would be to stick something like a single CPU 1.4ghz celeron (they won't work in dual) on a slocket and use that. You should get one of those celerons cheap enough. Should be a decent boost over a dual 500. I've a (one only) slotket for sale if you are interested. www.powerleap.com have a bunch of upgrades you could use too.

But I still reckon you could sell it and get a cheap 2nd hand socket A board with integrated video/sound/lan with a 1600XP or better with the money. That would be a lot more useful to play with as 2nd machine either a windows/linux box/server etc. As for the discussion above ah well thats what the forums are for init? Nice to have natter.

Gerry

Ricardo, you probably think that I'm being pedantic by now, but I'm sticking to my guns. I don't care that its way off topic, the topic is being handled as well. You are attributing things to me that I didn't say. In single threaded apps, a faster single cpu machine will of course be faster every time. But the trick is to use multithreaded applications. If you waste your dual cpu machine by doing one thing at a time, thats your problem. The future lies with combining more and more cpu's, as we get closer to the limits of silicon fabrication. IBM already have a dual core power4 cpu out. Intel launched hyperthreading to take advantage of the whole chip, knowing that most of the time, you will have 2 threads wanting to run. Sun are launching chips which can run 4 or 8 threads at a time. Yeah, it needs better programming. But its not impossible.
The clusters you refer to are still multicpu machines, they are cheaper because they are not bleeding edge, they don't need crazy chipsets, they can do without a 6 layer motherboard, etc. They still need to have the work split up and doled out. In fact, cray has gone with single cpu nodes for their new athlon 64 based supercomputer, which did suprise a lot of people but costs less due to the reasons above.

In other news, the 1ghz p3 I referred to using 100fsb is starved of memory bandwidth compared to a 133mhz fsb chip. This is true in a single cpu machine, and so is much worse in a duel cpu machine where the 2 are fighting for memory bandwidth. So you don't get the full benefit of paying more for the faster chips.

RicardoSmith

You could since we're on and off topic that we've multiple threads running in the same threads... or is that waaaay to geeky?



Coding multithreaded applications is a lot more complicated. So if you don't have to do it, its a big cost saving. Using multiple CPU nodes/machines is very cheap and scalable, you'd be mad not to do it. Hyperthreading does give you a boost ok. Dunno how much extra work is required coding wise to take advantage of it though. Certainly using a dual cpu machine does not give you the cost savings and outright performance that having two machines gives you. You don't even need to rewrite your software. Its a win win situation.

I don't see having multicpus and having multiple machines as the same thing at all. Theres a world of difference.

Not to be picky but I've never noticed a real world difference between 100/133fsb. 66 and 133fsb yeah ok, and maybe 133 and DDR. You'd have to be running something really intensive for it to have an effect though.

Gerry

Well, pretty much everyone is finding that they have to code up applications as multiple threads these days. It can actually make programming easier, in a lot of cases. Because each thread only worries about one aspect of the program. For example, starting java will probably create 20-30 threads.
With multiple machines, surely you still have to split up your workload, or run separate things on the 2 machines to get a boost? It would be easier to make all the machines in the cluster appear as one, so you could run ordinary programs on it.

With regards to fsb, its only really noticeable in games. Particularly quake3, which craves memory bandwidth. You can see the difference in fps even between 133 and 150.