New to Dual processor, need help and advice. thanks

[kelvin1704]

I am new to dual processor platform

1) can anyone give me hardware advice like Intel or AMD and the motherboard?

2) RAM issue mainly on ECC or non-ECC

I am mainly going to use it only on games, DVD, burning, video editing

thanks

 

[kelvin1704]

anyone?

 

[Jhereg]

The best bang for the buck these days are AMD XP/MP CPUs running on the Tyan Tiger motherboard.
check 2cpu.com for more info ( one of the best SMP dedicated sites on the web)

New MPX chipset boards are just coming out which will probably drive the Tyan boards to a lower price point.

ECC memory is good in the server market and since it is not exceedingly expensive nowadays , its worth the extra protection on workstations.

dual system will help only if you have applications which exploit it, very few games are written for SMP enviroments

 

[kelvin1704]

but do u think i need to buy the ECC ram?

when is the board coming out which will drive the price down?

 

[MrHelpful]

Another thing - you'll need registered memory if you plan on having more than 2GB of memory (IIRC).

 

[Jhereg]

I think ASUS and Abit are just introducing MPX boards , as is Tyan , check their sites.
ECC is not needed but I like the extra protection

 

[bozo1]

The MPX based boards are expected to be out mid-January. The Tyan Tiger is out now and requires Registered memory. It does not need to be ECC but the Registered most places sell is also ECC. The MPX based boards will not require Registered unless you go over 2GB of memory.

You may notice things slightly snappier with a dual board but unless your apps specifically support SMP, you aren't going to see the speed improvement that you may think.

 

[ScottMac]

If your apps don't use the SMP capabilities, they will run SLOWER, due to the administrative processes required to divy up the various jobs. Chances are you'd be able to do more things concurrently, but a single-non-SMP aware application (like most games) will run slower. I run three dual machines, mostly for video editing and file/application service (Adobe Premier is SMP capable...most consumer-grade video editing apps are not, and most CODECs are not).

*nix will generally work better for multi-processor. Otherwise, from MS, you're limited to Win2K or XP, the other OSs are not able to run multiple processors.

I'm just guessing here, but you'd probably be better off buying a hot single machine and dumping the extra cash into better video and more RAM.

FWIW

Scott

 

[OS]

<< If your apps don't use the SMP capabilities, they will run SLOWER, due to the administrative processes required to divy up the various jobs. Chances are you'd be able to do more things concurrently, but a single-non-SMP aware application (like most games) will run slower. >>



I have no idea why all of a sudden so many people here are saying this but that's simply not true. I've tested it myself, I've run game benchmarks with two cpus in and one cpu in on the same exact system. Run test with two cpus. Shut down, pull out a cpu. Run test again. There is no significant difference (I don't recall any variance greater than 1%).

Unless of course, you can show me some benchmarks that show otherwise.

 

[SUOrangeman]

Outer-

Take a look at the AMD XP 2000+ review over at GamePC. The dual config did score lower than single CPU systems in some tests. Wheter such a comparison is worth noting is up for debate.

-SUO, still going dual ... hehe

 

[ScottMac]

It will depend on what you're doing, and how the tasking can be acomplished, and how the application is written. Sometimes there's not much difference, somethimes there is.
Somewhere along the line, something in hardware or software has to decide what gets processed where, and that logic will slow down the processing while it's deciding. It's more latency than slower processing...but the result is that on a single processor, the task gets done sooner. In addition, some processing time is taken away from the primary task while "something" is handling the multiasking schedule. When you take cycles away from the task, it'll take longer to execute....try Seti (gui version) versus Seti (CLI version)...on a single processor, maintaining the GUI and doing the graphical output just about doubles the time.

Seti is not exactly a perfect example, because whether the results are on-screen or not, the calculations for the display are still being done. The Example was to illustrate the point of some cycles being taken away for other tasks, and the ultimate result is is the "main" process will take longer to finish.

You get just som many cycles, the more you spread 'em around, the less processing-per-task is gonna happen (plus the administrative time switching tasks).

FWIW

Scott

 

[OS]

Link? I can't seem to find it.

 

[OS]

<< Somewhere along the line, something in hardware or software has to decide what gets processed where, and that logic will slow down the processing while it's deciding. It's more latency than slower processing...but the result is that on a single processor, the task gets done sooner. >>



That's a nice theory but I've yet to see that played out in any benchmark. At anyrate, the OS is not going to juggle around a task unnecessarily for the sake of using up more CPU cycles.




<< In addition, some processing time is taken away from the primary task while "something" is handling the multiasking schedule. >>



And what makes you think a single cpu system doesn't have to do the same sort of scheduling?

 

[VicLavigne]

<< I am mainly going to use it only on games, DVD, burning, video editing >>


I don't think you'll get your money's worth with a dualie for this. Get an AthlonXP 1900+ cpu, RAID 0 array, DVD-ram, top of the line vid card, and mega ram and you'll probably have the system you need.

 

[MrHelpful]

here you go

 

[OS]

<< Take a look at the AMD XP 2000+ review over at GamePC. The dual config did score lower than single CPU systems in some tests. Wheter such a comparison is worth noting is up for debate. >>



Ok I just saw review on gampc.com and you know what, they aren't even using the same chipset when comparing dual to single. That's one huge ass variable don't you think?

The only fair comparison would be to use the same motherboard, run a benchmark with one cpu and the same benchmark with two.

 

[kelvin1704]

thanks for all the comments..

i think i might still invest in a dualie just for the sake of trying out....and i do use concurently a lot of application...

another question, can a Athlon XP be use in a dual motherboard?

 

[Jhereg]

<< another question, can a Athlon XP be use in a dual motherboard? >>



Yes and No, AMD does not support XP on dual configs but the early on it was shown that XPs would run on them. Since then it seems that AMD has made a concerted effort to make MB makers adjust BIOSes to deny dual XPs.
Still it is possible to make XPs run in Dual ( XP amd MPs are exactly the same with the exception that MPs are guaranteed to work in dual configs).

Just be aware that if you do get XPs and they don't work on your dual board ... sucks being you

 

[Remedy]

OuterSquare, it is true that using Non Multithreaded apps on mulitprocessor system decreases performance. Xbitlabs showed the same results as did a couple of other not-so-reliable sources. Its a matter of the windows kernel and the ev6 architechure. But the same results were found on Pentium MP systems also. Just because you "tested" the theory doesn't make the result written in stone. A grain of salt is more like it.

For the Xbitlabs review. Click here

Besides, Windows doesn't break up or "Divy" up the X singluar thread in workload across 2 processor or more because the app was never multithreaded. Windows NT core is multithreaded. The only processor the single threaded app will go across is the first processor. Which is Cpu1, the slave processor will only host the operating system services.

What you can do is download SMPseesaw to manually divide the workload on the selected processor but then again, if your going to go MP then you might as well have the appliaction to take full advantage of it. Multitasking is only a minor reason to go that route. I can optimize windows to handle similar gruntloads with single threaded apps to efficently multitask. But hey its your money, do what you want.

But then again this is JMO i could be wrong.

 

[OS]

<< OuterSquare, it is true that using Non Multithreaded apps on mulitprocessor system decreases performance. Xbitlabs showed the same results as did a couple of other not-so-reliable sources. Its a matter of the windows kernel and the ev6 architechure. But the same results were found on Pentium MP systems also. Just because you "tested" the theory doesn't make the result written in stone. A grain of salt is more like it. >>



Atleast I've looked into it, as opposed to you who is working off hearsay.

At any rate, taken from the review you linked to;

Apparently, the situation is aggravated by an ill-implemented AGP support. As you might've noticed, dual-processor systems fall behind uni-processor ones both at lower and higher resolutions, when the main strain falls upon the graphics card. We can supply no other explanation but bugged AGP and IRQ router drivers. If so, there's some hope that improved drivers will come out soon and will increase the gaming performance.

By their own admission, the slowdown is caused by the AGP implementation, not necessarily dual CPU operation.

 

[borealiss]

OuterSquare

to a certain extent, this is true. but not for all apps. as far as the quake 3 comparison at gamepc is concerned, i would disregard it. they run it under xp, and anything based on the win2k kernel regarding quake 3 smp has always been iffy at best. you have to run the game in realtime under win2k to even get it working properly. if you want a true comparison of how smp and uniprocessor systems fare using quake 3, you need to run it in NT4, as it is the only OS that will support smp in q3 the way it was meant to run to my knowledge(linux?). but others like earlier versions of lightwave would run multiple threads slower. in some instances, apps run faster if the os loads the app to a specific processor and leaves the os handling to the other. but usually you have to specify cpu affinity for this to occur.

 

[Remedy]

For the record i don't go off of hearsay. I have dual systems and probably been in the game alot longer than you wish to think. So i have been doing this type of testing. Dual BX life time. Until you can come up with a better logic then mine i don't think you looked into this deep enough. Across different platforms in particular. If you have a hardware site that you run, post it up. I wouldn't mind reading your evaluation. I'm not going to sit here and beat a dead horse, here is one last link. Link

 

[OS]

<< I have dual systems and probably been in the game alot longer than you wish to think. >>



Why does that matter? I know people who drive cars everyday that couldn't add wiper fluid if their lives depended on it. Using something is not the same thing as understanding it. Not to mention riceboys, they daily drive civics and make all kinds of outrageous claims to beating expensive sports cars, but that hardly makes it a fact.



<< So i have been doing this type of testing. >>



Apparently not enough because your own freaking link shows that the dual P3 is in fact even consistently a little faster than the single P3!! Clearly there are issues with the dual K7 rig, but as mentioned above, one can't immediately ascribe the cause to dual cpu operation.




<< Until you can come up with a better logic then mine i don't think you looked into this deep enough. >>



Looking is not the same thing as thinking.

 

[Remedy]

Umm first of all buddy. My link to my rig is pretty old. I haven't touched it in months and even forgetten that it exist. I had P3's from all around on different VIA boards. But since you seem to nitpick instead of having a good discussion i'll leave it at where it is.

 

[viper007]

<< Outer-

Take a look at the AMD XP 2000+ review over at GamePC. The dual config did score lower than single CPU systems in some tests. Wheter such a comparison is worth noting is up for debate.

-SUO, still going dual ... hehe >>



I think that's sorta unfair, its comparing it against a KT266A chipset, which we all know is superior to AMDs