Dual Athlon MP Boards

[eternalseven]

I always like the dual board reviews at THG, but this time it seemed like they didn't release much data and comparisons at all. As if they didn't want to do their homework.

I want to see comparisons to other boards (Tyan) and comparisons to Xeon boards as well.

And why does THG keep including Lightwave benchmarks when using Athlon systems? Lightwave has been clearly optimized for P4 and cripples the Athlon.

No more text needed.


~ Eternal Seven

 

[Guest]

From the review it seems they are clueless about even how to use a dual board. Quake3 benches ? uhm, anyone buying a dual for quake3 needs a therapist.

Lightwave does much much better than what they showed it benching as in that review. They clearly have the program configure wrong.

The 3dsmax benches are also laughable. They need to look at some of the other benchmarks people have done who actually know how to configure the programs. 7% difference , lol.

hmm, other benches show rendering with one cpu 50 secs, 2 at 28, hmmm, looks like almost 50% difference to me.



Please get someone who knows how to properly bench smp systems or don't review them at all.

 

[Guest]

I agree that the benches for 3d studio max are off. I've built dual AMD systems for people who are Max fanatics. That's all they really wnat their comps for and the dual AMD is by far the best solution. The machines I've built for them are screemers. They are constantly telling me how much time it's saving them and they have computers of all sorts. From dual PIII systems to xeon systems to P4 systems to Single AMD systems. By far for max dual AMD is the only choice. It makes me wonder about those bench marks.

 

[Malc]

Another highly dubious benchmark under application benchmarks is the Linux kernel compile. The dual CPU compile is slower than the single CPU one. Obviously they didn't use both CPUs!

A less tangible benchmark that they never mention is how responsive an SMP box remains under really heavy load. I guess they don't ever do anything else on their computers after starting a long-running CPU-intensive process.

 

[shloader]

Looking through previous MB articles I didn't think to see a SMP article on the 760MPX and when I first saw it I thought it was a godsend. Let me compose myself...

This article was the biggest pile of crap I've seen since their last monitor review... of a monitor that seems to have little over a good Viewsonic. I've seen far better articles on other review sites where a dual 1.2 AMD slammed a dual 1.7 Xeon with far more comprehensive benchmarks. There's plenty of ways I could rip through the irrelivencies of this article but I only have a limited amount of time on my break so I'm going to throw just one out there. Notice how the writter says that SMP is more enjoyed by hobbiest as a way to encode Mpeg2 for DVD faster? Not a single benchmark on that subject, only DivX. I haven't touched divx in years. And what's the deal using Pinnacle software for mpeg2? You don't pay for xmpeg/flask and they use that fir Mpeg4? Use Tmpgenc 2.0 for Mpeg2. I'm sure the author would like the publicity on a widely known site for something that may not be free later.

Benchmarks... hell, reviews here have taken a dive on how comprehensive or relevent they've been lately.

I ordered my Asus A7M266D two days before this article hit. This didn't dishearten me in the least as he did nothing i would do on a SMP system. I still intend to load Win9X for games when I want to. I've never had a dual system... seems I know more than this article can give, though.

Oh and... MSI? They make an SMP board? Really. So does TYAN. Something they're known to do very well.

Intel is god, cleanroom is life, and the wafers are all that matter.

 

[Guest]

Kernel compiles do run rather faster on a dual board than a single one if you let make take advantage of both CPUs by spawning several concurrent threads ... From my dual p3-600 machine (the 'Elapsed' time is the one to look at!):

A single-threaded kernel build:
command time bash -c 'make clean && make dep && make bzImage && make modules && make modules_install'

802.45user 64.51system 14:45.25elapsed 97%CPU

A triple-threaded kernel build:
command time bash -c 'make clean && make dep && make -j3 bzImage && make -j3 modules && make -j3 modules_install'

819.93user 64.02system 8:35.66elapsed 171%CPU

A sextuple-threaded kernel build:
command time bash -c 'make clean && make dep && make -j6 bzImage && make -j6 modules && make -j6 modules_install'

829.03user 63.10system 8:29.75elapsed 175%CPU

 

[Guest]

Nice to see an actually useful comparison. I'm considering
getting a dual athlon and I've found the discussion not
as illuminating as I hoped (the review was useless, I agree). I expect to be running various kinds of high cpu
tasks (adaptive smoothing, feature detection, etc on
large images) while having about 30 different emacs' running
doing different other things, and compiling code in yet
another window. So my hope is that Linux will automatically
(at least if SMP is compiled... said he vaguely) set one
of the Athlons to get busy with my smoothing and the other
to do my everyday tasks and compiling etc., and I'll basically see close to a x2 performance improvement
relative to a single CPU, modulo issues of disk access
and having enough memory. Is this a realistic hope?
Jonathan
(who is never doing less than 5 things at once...)