Swiftech MCW200 Cools an ASUS P3V4X

When Gabe at Swiftech first politely tried to dissuade me from using the MCW 2000 Barebones system with my Asus VIA P3V4X, I was a bit reticent. After all, I have had pretty fair results using this motherboard with three 500E’s and a 650E.

The 500E’s all hit 750MHz using an Alpha cooler; the 650E had run with stability at 877 MHz using my patchwork water cooling set-up.
But given Gabe’s knowledge and experience, I felt it would be a foolish move to discount his advice. After receiving my new system, I reveled in GHz splendor for a week, running benchies in Sandra and 3DMark and delighting in Unreal Tournament’s gory frames per second.

Gabe’s advice to go with a proven BX board – nearly a year old – the P3B-F paid off big time!

My new 700 cB0 blossomed into a 1050 MHz ice-plant, succulent with the fragrance of sweet success. At ambient temperatures in the high 70’s F, she remained rooted in frosty temps from 0 to -7C.

“How sweet I roamed from field to field and tasted summer’s pride.” (W. Blake)”:O}

I returned my “old” 650E to its accustomed Asus VIA home. Though still needing to be shaded by lower FSB settings when the temperatures began to climb beyond 70 F in Seattle, she continued to bench and run with stability as long as I didn’t ask too much of watercooling unaided by a peltier.

Then I began to wonder: The P3B-F was without question a fine motherboard. Yet topping off at 150 FSB left room to ask “How high an FSB could a peltier cooled 700 cB0 be incited to climb?”

Looking at the Swiftech watercooling apparatus surrounding my P3B-F, another question arose: “How hard would it be to swap out motherboards?” After all, the Asus VIA P3V4X had settings that reached a stunning 166 FSB.

What about that VCRAM which would only run on a VIA board? Would it give the kind of service at a higher setting that the Mushkin ENHANCED currently did on my BX board? Just how curious was I?

The swap turned out to be fairly easy: There was enough flex in the water lines to allow me to slip the MCW encased CPU out of it’s Slot 1 berth by easing it past the radiator and 120mm fan sitting next to and behind it. I was able to rotate it a half-turn and up. Then I zip-tied it to the case rails above, out of the way. There proved to be no further hindrance to removing the motherboard as one normally would.

Pulling the PV34X out of my other box and slipping it into place beneath the MCW 2000 was not a all difficult. I’ve always felt that O/C.com should give out certificates for such procedures that would read something like:

This task has been successfully performed by Daniel Edgar and therefore we certify it as idiot-proof.

So, after loading up my motherboard and filling the slots with RAM and GeForce, Sound and Promise cards, pressing the IDE ribbon cables and case connects into place, I hit the big button and, to my amazement, it booted first time.

I started off using the HSDRAM as I wanted head-to-head benches between the P3B-F and the P3V4X. BUT as this was a motherboard comparison and an overclockability comparison at that, I went for the highest FSB I could grab and get to run with stability.

With my Motherboard RAM and CPU, this proved to be 1085 MHz. I emphasize “my” here because I would expect to see enough variance between motherboards, RAM and CPU to make this comparison useful only in general, or in helping me decide how best to distribute my gear between two systems.

A disappointment at the onset – I was only able to run my HSDRAM at the AUTO setting, even at 1050 MHZ. This of course adversely effected my memory scores in Sandra. “AUTO” defaults this RAM to 133 MHz and a 3-2-2 RAM timing. On the BX board, it runs at 150 MHz and a 3-2-2 RAM timing.

Since a 1085 MHz setting did not cost the P3V4X anything (as opposed to a 1050 MHz setting), I used this setting to run my benchmarks.

SiSoft Sandra Benchmarks:

CPU Bench Mark

Dhrystone: P3V4X 2941 P3B-F 2894
Whetstone: P3V4X 1455 P3B-F 1442

Integer : P3V4X 3419 P3B-F 3311
Floating Point P3V4X 4552 P3B-F 4408

CPU P3V4X 403 P3B-F 467
FPU P3V4X 469 P3B-F 520

Drive Bench Mark
P3V4X 19,595/ 19,370/ 20,532
P3B-F 21,011/ 21,290/ 21,469

As we can see, for my purpose (ie find the faster board) the P3V4X out-performed the P3B-F in CPU and Multi-Media benchmarks. However,it lost out badly in the memory benchmarks, due to at least to three factors:

  • First, the higher FSB used by the P3V4X gave it an advantage in CPU and Multi-Media benches;
  • Next, the P3V4X’s inability to run my HSDRAM at a 1:1 ratio on 150 MHz FSB hurt its memory scores;

  • Finally, the slight advantage the BX chipset still offers in memory bandwidth. At these FSB speeds, the BX beat the VIA board bloody.

I’m inclined to see the differences in drive scores as insignificant. This is due in part to Ed Sroligo’s recent article which shows current drive benchmarks possess little value when it comes to predicting drive performance or its effect in a given system. Also, there is a relatively small difference in the measured performance between the two contenders. Still, for whatever it’s worth, the BX board showed a consistent edge.

Then it was time to look into what I really came here to determine: Video card performance. Here the VIA board should really shine. It holds several separate advantages over the P3B-F BX board. Well, kind of:

It offers 4x AGP speeds while the P3B-F is stuck with the default 2x. But as the Accelerated Graphics Port only comes into play when the video card is accessing the CPU and System RAM (it currently only requires 2x bandwidth) this factor would make little, if any, difference.

Then there’s FastWrites. This allows a Video card to communicate directly with the CPU without having to go through main memory.
Unfortunately I could only get 3DMark to complete a benchmark with FastWrites turned ON.

Lastly there is, of course, the 55 MHz CPU speed advantage the VIA board offers over the P3B-F BX board.

Here I was frustrated once again. At 1085 MHz, my Elsa Gladiac was slammed back to the desktop again and again while playing Unreal Tournament. In fact, I was never able to even get a FPS reading even at a reduced 1050 MHz!

So the following P3V4X 3DMark benchmarks were recorded at 1050 with VIA’s 4x AGP set to 2x as 4x would not run the benchmark either! FastWrites turned off – a level playing field by default.

Every chance of a VIA advantage gone. Every reason I had for possibly preferring the VIA board, given the high quality of my components and their proven ability to run with stability at 150 MHz FSB, gone.
Yet mightn’t VIA still do well?

After all, it was a level playing field. So I cranked up 3DMark, shut down very icon in sight and began a benchmark. Crash. I began backing down on my Geforce overclock.

I had run the P3B-F at a setting of 240 core speed 386 Memory clock. By the time I found a stable enough setting to benchmark, I was at 225 GPU core and 376 MHz Memory clock – not what I had hoped to find.

  • P3V4X

640 X 480 16 bit 16 bit Z buffer with T&L= 8679
640 X 480 32 bit 24 bit Z buffer with T&L= 7983

1024 X 768 16 bit 16 bit Z buffer with T&L= 7302
1024 X 768 32 bit 24 bit Z buffer with T&L= 5236

1280 X 1024 16 bit 16 bit Z buffer with T&L= 5685
1280 X 1024 32 bit 24 bit Z buffer with T&L= 3457

  • P3B-F

640 X 480 16 bit 16 bit Z buffer with T&L= 9194
640 X 480 32 bit 24 bit Z buffer with T&L= 8442

1024 X 768 16 bit 16 bit Z buffer with T&L= 7724
1024 X 768 32 bit 24 bit Z buffer with T&L= 5456

1280 X 1024 16 bit 16 bit Z buffer with T&L= 5936
1280 X 1024 32 bit 24 bit Z buffer with T&L= 3592

Now, What about that VCRAM? As I could not get it to run in a stable fashion at these FSB settings, I have no benchmarks. Can’t say why, but I spent a lot of time trying with nothing to show for it.

With as much as a 415 point difference in the 640 x 480 range and as little as 135 point advantage at 1280 x 1024, the P3B-F again rose to prominence but not by a whole heck of a lot, especially considering that not many GeForce cards are run at less than 1024 x 768 settings.

But overall, as the reasons for preferring the P3B-F piled up benchmark by benchmark, my choice became clear. The P3B-F would retain its place beneath the MCW 2000 and the P3V4X would be returned to it’s makeshift watercooling system in my other box.

My 650E would continue to run at 877 MHz when the weather permits and to be clocked downwards when house temperatures rise past 72 F.

I’d like to thank Gabe at Swiftech for designing a great watercooled Peltier system and for steering me to a motherboard which gives it the stability, features and settings I need to let my 700 cB0 at 1050 MHz rock my world.

Email Dan

Be the first to comment

Leave a Reply