You used all single threaded applications to test the system, in the SMP cases, you barely used decent SMP-based applications.

There are plenty of SMP-capable software tools out there....flask encoding is one of them, and you could've used that as an accurate test.

Why would you ever buy a dual system that does piss poor versus a single processor system, for 3x the price?

You get a warrenty with a pre-built. Thats one of the big selling points. But you don't get a warrenty if you mess with the congifuration.
[QUOTE]The TIM is AMD-approved, so there's no reason to believe that it won't be adequate to transfer massive amounts of heat away from the Athlon MP processo/QUOTE]
So the TIM is fine. But then you go on to say that this isn't good enough because of you remove the HSF's the TIM is in the way.

It's a pre-built. The people who buy this PC will not be removing the HSF's.

RAID0 : need more be said? For a work station, I'd like some form of data security, not half what a basic home machine gets.

The PSU was loose. OK, maybe not loose. Not screwed in may be a better way of decribing it. In my eye's thats an instant failure for a prebuilt. Yet it got a passing mention. It's lucky you didn't get a DOA motherboard, and I can imagine the company disclaiming responsibility.

The benchmarks, never mention what settings were used. The two systems were in no way similar. Different RAM amount & Speed, Video card, HDD's. About the only thing the same was the CPU speed. No mention of what drivers were used for any piece of hardware.

Yes, this was posted anonymously, and I probably won't be back. I saw the way that R2P2 was attacked for questioning something here. It's pathetic. Grow up people.

This was posted as an honest set of constructive criticisim, not an attack on who ever wrote the review. You have to remember who the target audience of pre-built systems is, and it generally isn't anyone who actually will read this review.

[quote][i]> the system even comes with Windows XP's fruity GUI disabled.[/i][/quote]

Excellent! Wish more vendors would do this.

But ... are they as quiet as Dells?

[q]GamePC credits Enlight for the case, but it's identical to what you'll find from Chenbro, Antec, and others—they're all rebadging the same thing. If you can't beat 'em, rebadge 'em. Whoever builds this case isn't skimpin/q]

Chieftec is the American distributor of Chenming, the actual manufacturer of these cases.

And I just priced out the exact same machine at Newegg (with a better PSU...) for over $700 less than this computer.

What kind of a workstation resolution is 1024x768? I wouldn't dream of using a laptop at that resolution, let alone a desktop computer. Workstation resolutions for any serious graphics work or programming start at 1600x1200. 2048x1800 would be high end.

Diss - Re: SPECviewPerf and Quadros:

If you want to see a Quadro stomp all over a GeForce, get SPEC 6 and run the ProCDRS test. That one uses a wireframe rather heavily, and AA'ed, so the Quadros run at full tilt, while GeForces fall down, curl up, and die. Interestingly, The Radeon 8500 appears to support hardware wireframe and wireframe AA, as it walks all over the Ti4600 in the ProCDRS test.

How is this a review if there isn't an 11-point scale at the end? :(

looks like a nice workstation. would love to see how those dual Athlon MP chips thrash around a dual PowerMac G4 box. :)

Diss - That's a PNY Quadro4 750 XGL. Quadro = NV10. Quadro4 = NV17/NV25.

Also, it's XGL. GLX is the Xfree86 openGL extension. Dunno what GXL is.

[q]What makes a GamePC pre-built system better than a Dell?[/q]
Oh goodness...that question even has to be asked?

Thanks for the revie, Diss. I wonder how a dually 2600+ system would do....

[qut GamePC found that 3Ware's offering was fasterd and change the spec. Very cool.[/q]Fasterd, eh? Cool, indeed. ;)

[q]Can you look into getting a printable view of your reviews?[/q]I'd like it, too, but until TR finds a way to bankroll the servers with printable reviews, I'm thankful for what I get.

[q]The entire earth does not have broadband.[/q]Nor does it need it to read reviews. The largest graph is only 21KB!

#23: Take your own advice.

R2P2, forget your bran this morning?

R2D2:
stfu.

Odd, the "Read on to find out" link from the comments view goes to the comments view instead of the article. I thought the summary on the comments page came from the same source as the one on the front page; guess I must have misunderstood how the site works.

What's with posting a multi-page article in the middle of the afternoon? Needed a little extra time to proofread, dissonance? Should have taken a little more:

On p4, in the paragraph under the fan picture, it says "Hust remember", which should obviously be "Just remember".

A more important mistake is in the table with the system configs -- The dually is listed with the Ti4600 and the KT333 is listed with the Quadro. Congrats on using HTML for the tables, though; much nicer than those Excel screens you were using for a while. ;)

Remember when AMDZone was asking for submissions for time to complete a SETI unit? He provided a setup so everyone could use the same work unit for comparison. Maybe that can be done with SETI, F@H, RC5, etc.

I see what you mean I think. You are saying that as F@H runs it switches instruction mix, branching, etc as it proceeds through the work unit? That must be why I see around 5% relative percent variation between successive CliBench scores. I'll fool around rebenching a single and duallie and then start a thread in the forums when I have some results compiled. One might still be able to take a stab at this and provide an estimated error of measurement by periodically rerunning one of the benchmarks between each of the other benches. Then one has an estimate of "drift".

Good review. I wonder what it would cost to build that exact system mail order.

I'm fairly confident there is a change in loads between frames. They certainly vary in time to complete, anyway, even when the system is untouched for hours.

Originally Posted by dissonance
Mr. Bill: I considered doing something like that, but my concern is that the load from F@H may not be the same with each benchmark run.

Ideally I\'d love to be able to run the same work unit each time, and have the actual load for that work unit remain constant throughout. I\'m not sure if, while crunching a work unit, there\'s any deviation in actual load.

I've always liked GamePC's lab reviews. They always have stability as highest priority.

Hmmm, You have to set the percentage (less than 100 mind you) and get F@H running. Then go into the task manager and set FahCore_65.exe to Realtime priority. Then run the bench. CliBench Mk III SMP 0.7.15 seems to get lower scores by an appropriate amount. I set the core at 40% and realtime priority on my duallie. The system load on task manager reads as 20% on each processor. Then I ran CliBench with two threads. Scores range from 65 to 77% of scores without interference. Memory throughput unchanged. Dhrystone hops all over the place but Whetstone, Eight queens, Matrix operations, Number crunch, and Floating point are reproduceable within plus/minus 5%.

a la aceshardware.com?

[q]SOME OF YOU may consider ordering a pre-built system one step short of sin,[/q]

Geoff, you meant "one step BEYOND sin" right? :)

ChangWang #6:
[q]Can you look into getting a printable view of your reviews? I'm on the road alot and its easier if I can get it all in 1 page, print and be done with it.[/q]

YES!!!! PLEEAASSEEE!!! do this. The entire earth does not have broadband. 10 page reviews are, howdoyousay, a [b]BITCH[/b] to download! and then try to keep open to read offline... even with the use of a real browser (Opera). :)

Respectfully,

JQ

I just realized we have a nice instant load already available, the Folding@home client. Just open a command prompt; change to the F@H directory; run Fah3console -config; choose advanced options; then select the CPU percentage desired. I'm testing it right now and it seems to work fine.

It would be nice to see a program like 3D Studio, Maya, or even Lightwave included but I guess that would drive the cost up too much.

fejji, I have the same thoughts. Below is a proposal I sent in for a Friday night topic.

Ever since one of Tech-Reports reviews talked about being able to run quake on a duallie while rendering in the background, I've been wondering how benchmarks would perform in a real multitasking single/multi CPU environment. We need an application whose sole purpose is to steal a
selectable fixed percentage of CPU time using a fixed mix of instructions. Later one could elaborate and allow selection of instruction mix sets. I'm thinking the way to get a fixed percentage of CPU time is to have the application run at real time priority but interrupt to do its "work" at some fixed percentage of the available load. Then run the bench of choice and see how different CPU's degrade in performance while the OS and CPU are interrupted by multitasking.

Maybe this could be a friday night topic. How to do it.

Umm... is it GPG or PGP? Make up your mind, already. :P

Well at least TR isnt one of those bulky ass sites with menus and all kinds of crap on the top, bottom, and sides of the site. It's far more printer friendly than some other sites....

Off topic but isnt it funny that TR's comment system organizes older posts at the bottom and new posts at the top, yet TR's forums do the exact reverse. Come on, lets have some consistancy!!!