Turbo Drive: Two Dual Boards with 2000 MHz

G

Guest

Guest
WOW, what a fantastic in depth article!!! :-( there were more pictures than benchmarks ! Now how can you judge a dual CPU system on only 4 benchmarks ?? hello ?? Besides, we all know 3D Studio loves dual CPU.. same for Cinema 4D. Sysmark couldnt care less if there were 2 or 8 cpu's in the system; surely there must be better benchmarks around ?

But the real interesting question here, has not been answered. How does a dual Celeron or dual low-end P3 compare to a single CPU P3/P4/atlhon system ? What about some gaming benchmarks ? I wonder what a dual CPU config does under Win2K for playing games. Hello ???

Pffff.. I used to consider this website a "reference", but lately the qualitly of the reviews is going down so FAST. I already turned my back on Sharky's for being so anti-AMD for a long time.. Thank god for Anandtech, en especially Aces Hardware. Tom, have a look at this review, and compare it to your own pathetic review(s) of the P4:
http://www.aceshardware.com/Spades/read.php?article_id=15000196 (initial P4 test)
http://www.aceshardware.com/Spades/read.php?article_id=20000190 (in depth P4 architecture.. Highly recommended !)

Now there is a good article on the P4...
toms hardware "The Internet's premiere source for PC Hardware information" ? Dont make me laugh.
 

slvr_phoenix

Splendid
Dec 31, 2007
6,223
1
25,780
I know. THG has definately let me down a few times lately, and this is a PERFECT example.

Obviously a dual CPU system will perform twice as fast as a single CPU system for software that is properly multi-threaded. And when running software that is only single-threaded, it'll make almost no difference.

Did we really need a whole review article just to tell us that which we already knew?

And it doesn't tell us anything else. Who in the world would test such an early revision pre-production motherboard against against a normal production motherboard? I'd have been REALLY surprised had the pre-production won ANY of the benchmarks.

And besides, we already know that the i815 chipset is more efficient than any VIA chipset. So do we even need to compare them? I'd be incredibly depressed if the dual i815 chipset mobo after production were slower than the dual VIA chipset mobo.

And where is a test of the i840 dual? How does it compare when using PC600 RDRAM and when using PC800 RDRAM to at least the dual VIA? At least THAT mobo has been a production model for a while now. So why don't we see anything like that in the article?

And what about the KNOWN issue between Windows 2000, VIA chipset motherboards, and certain graphics cards like the Matrox Millenium G400? Do we see ANYTHING about this in the article? Yet a noticable number of people running dual CPUs are going to be using Windows 2000.

Frankly, this whole article seemed to be very poorly researched, lacking in hardware comparisons, AND incredibly rushed to have used even such an early revision engineering sample motherboard that is no where near being considered a final product. I just simply can't imagine how this article could be lacking in anything more. It's the worst article I've ever seen on THG. Don't they have ANY quality control? Dr. Pabst, why aren't you smacking Frank Völkel for this atrocity?

And bbaeyens, running benchmarks like Quake3 or Unreal Tournament or ... well ... probably ANY game aren't going to show you anything useful. Games are all single-threaded applications. So they won't see any noticable speed increase from a second CPU because nothing will ever be run on that second CPU. You might see as much as a 10% speed increase, and only because one CPU will be running Windows (and associated overhead) while one CPU can devote itself purely to the game. For gaming purposes, there really is NO point in multiple processors.

- Sanity is purely based on point-of-view.
 
G

Guest

Guest
I wans't that impressed either....

M

PS still waiting for your hardware?? :)



one of the first UK T-Bird users....
 
G

Guest

Guest
I would agree with that. The almighty John Carmack had a serious look at multi processor support for Q3 during its development, and he predicted a maximum of 15-20% speed improvement, coming from a total rewrite of the game engine. Knowing his pedigree, I would be loath to argue...

===
Do unto others before they do unto you...
 
G

Guest

Guest
very disappointing on the timing of the article (ie. preproduction board... WHAT?). also these aren't the only chipsets for dual socket 370 mobo's. the serverworks chipset may be expensive but it is an alternative (and there are several varients, which might put that nifty features chart to good use. someone might see a box checked or unchecked that they've never seen before, but curiosity and interest might lead to more problems around here than it's worth).
 

slvr_phoenix

Splendid
Dec 31, 2007
6,223
1
25,780
Actually, I've been lucky. I've been able to test my software every so often on the hardware today. (Only for about 5 to 10 minutes at a time though.) Debugging is always a major nuicance when you don't actually have the copy of VB to install onto the computer you're doing the testing on. I wish the people who made the hardware could have just made up a dummy box to link to my own computer so that I could test it here any time I wanted. That'd be so much easier than running back and forth from my desk to the application scientist demo labs.

- Sanity is purely based on point-of-view.
 

slvr_phoenix

Splendid
Dec 31, 2007
6,223
1
25,780
Yeah. A lot of software really doesn't gain much of an advantage by using multi-threading. You still have to keep the threads synchronised because it does no good to draw a scene if you're still trying to process the appearance of several of the figures in the scene. So you have to make sure that the other threads have processed those appearances before you can draw the whole scene. And in doing so, 3D games just don't gain much advantage.

The only real applications that do gain a significant advantage are ones that do a lot of differing algorythms or routines simultaniously. Which is why the software that my company is developing has part of the code multi-threaded. In there we have one process moving servos on a camera-like device to collect images from different angles. And at the same time we send these images from the 'host' collecting them to the 'client' that wants to actually use them. And at the same time as that, we can also have several software filters applied to the images.

So by multi-threading we can actually speed things up because there normally is a lot of potential for wasted CPU time as the software would be waiting for a responce from the hardware collecting the images and moving the servos and a lot of wasted CPU time between processing the image filters and actually sending the images across the network.

But most software doesn't have opportunities like that to actually gain anything from multi-threading their applications. (Except that if that application were to run on a dual CPU system it would be using both CPUs and thus about twice as fast. But most people don't use dual CPUs, so it's usually a very low priorty to do for this reason.)

I really couldn't see how any game software could improve much from multi-threading because the CPU is just about always going to be running maxed out anyway. It really shouldn't have idle time.

- Sanity is purely based on point-of-view.
 
G

Guest

Guest
I am lucky too really , we are regarded V.highly and generally get what we want(god knows why), we have our own mock systems and access to the live data whenever we need it...anyway good luck with your debugging , I am trying to finish a payroll integration system for 30000 timesheets a week and it is such a pain to try and test....

M


one of the first UK T-Bird users....
 

slvr_phoenix

Splendid
Dec 31, 2007
6,223
1
25,780
Yeah. I just don't know what is going on with THG lately. It feels like they're being pressured into releasing articles that just aren't really ready to be released.

And when I had seen the article I had high hopes because I've been thinking about what would be good to put into a multiple CPU system for myself lately. Even though I don't NEED two CPUs, I can find times when it would be handy, even if just for running an MP3 player while I'm doing something else and not have to suffer any CPU load problems.

So anyway, when I saw just how POINTLESS the article was and how little information it gave, I was rather depressed. I mean it hadn't even managed to give any new information over what we already know.

- Sanity is purely based on point-of-view.
 

slvr_phoenix

Splendid
Dec 31, 2007
6,223
1
25,780
My company has the opposite view. I think it's something like, "Software engineers should be seen and not heard."

Good luck on your software too. I've had friends who worked on similar things. I've heard how annoying database software can be.

- Sanity is purely based on point-of-view.
 
G

Guest

Guest
http://www.tomshardware.com/mainboard/01q1/010201/index.html
says:
Presently, there are two different chipsets on the market for the socket 370 platform supporting the operation of two processors: These include the Intel 815E and the VIA Apollo Pro 133A - also known as VIA 694XDP
What happened to the i840? Did they stop making the i840 already? I was just about to buy a dual board based on this.

He also lists the 840 as slot only - but the Tyan Thunder i840 is a dual socket 370 board. What does this mean?

(strange that Tom didn't include the Serverworks chipset either)
 
G

Guest

Guest
And the conclusion is the topper!!!

A single system is priced @ $1760
A dual system is priced @ 2090

The conclusion is:
Especially the price for a second CPU makes a big difference and raises the cost of the dual system immensely

First, duh. (insert related story about how buying twice the memory does in fact increase cost...)

Seoncd, the price increase is only ~5%. I think the story is that because these dual CPU boards are very inexpensive compared to the high-end dual boards, the cost increase is not "immense" but pretty small. Now how else can you increase Max by 50% by spending 5% ?
 

Crashman

Polypheme
Former Staff
840 is still around. Tom only uses it when comparing RD to SD or DDR ram. The 840 would have had an unfair advantage in his testing because it uses RD ram, which tends to perform better than SDRAM in the few applications that can actually take advantage of SMP, and worse in aplications that do not (genrally speaking-it's actually an issue of latency vs. transfer rate).

Suicide is painless...........
 

SoulReaper

Distinguished
Dec 31, 2007
194
0
18,680
I would buy dual processors for one reason only. It would cut down on my compile times. When i compiled my last half-life map for valve it took over 49 hours to compile. Dual processors would cut that in half if not more. Thats 49 hours on a 1ghz CPU with 768mb ram, all scsi. I can't wait for dual athlons to come out. Garanteed i will be a pioneer for the dual athlons. BRING IT ON!!!!

--SR