Wow, 2.5ghz at 1.65V... that's a pretty nice chip you've got.

As for FSB vs. multiplier overclocking, your particular board seems to have some quirks with higher multis that make higher FSB preferable, but I wouldn't make hard conclusions based on that. It's pretty well known that clock speed is what really matters on Athlons, and it doesn't make that much difference which method you use to get to a particular speed. I see that in your screenshots, the difference in scores is pretty consistent with the ~120mhz difference in clock speed.

(Of course, if you were looking at extremes, then 200fsb is certainly going to perform quite a bit better than 100fsb... but even the 25% jump from 133fsb to 166 only made for ~5% performance difference in a few things.)

If you're interested in testing that out further, try comparing 17x133 versus 15x150... both are at ~2.2ghz, and they should show to be within a couple percent of each other. You may find a slight difference when using the PCI V5, since I believe it's a little bit bus-limited and scales a certain amount with PCI speed.

Regarding the voltage of the CPU, the BIOS set up area tells me that it is 1.71V. At the Win98 desktop, Sandra reports it as 1.72V. If I then set the CPU's default multi to 16.5x (from its boot up "speed" of just 6x), then this value drops to 1.70V. I guess this "overvolt" of about 0.05V is acceptable. BTW, you can "undervolt" a CPU using this mobo's BIOS set up area options.

Regarding further testing, in particular testing at 17x multi - unfortunately this will require me to "operate" on the XP-M, because its default factory unlocked multi is "capped" at 16.5x. I'm a bit reluctant to operate on it, because a) I've only just got it, and b) it wasn't cheap!

Regarding the "trade off" between FSB increase and multi increase, I did do some relevent testing here (at the very bottom of this page).

Can anyone please recommend a product that does the following -
Inside the BIOS set up area, I can see the voltage infomation for the processor, AGP, etc. But I want to be sure that this information is accurate. Is the device I need called a multimeter? If so, can someone recommend the name of a good one?

If I type in the word multimeter in to ebay, 1000s of results are returned. However, if I type in multimeter pc, then only 9 results are returned. These listings have either the word USB or RS232 in them. Am I searching for the wrong thing? Thanks for any advice.

"multimeter pc" gives you those that you can hook up to a computer to record the results.

For this purpose IMO you'd need one with better accuracy than the 10€ parts. Also you need to figure out where on the mainboard that specific voltage can be measured, and in case you slip with the probes you can ruin everything.

h-a-l-9000 wrote:

... you need to figure out where on the mainboard that specific voltage can be measured ...

Can I just double check something? In order to measure what the PSU is doing, in terms of its voltage supply, I must use a multimeter device that has its own built in probes, and have these probes attached to an area on the mainboard. I assumed (incorrectly) that I need to search for a device that "sits between" the PSU and the mainboard, a little bit like a PSU power cord adapter, including some kind of display.

Will this multimeter device with built in probes tell me about the values of the different PSU voltage rails? For example, it would be very interesting to learn what power is being used on the 5V rail.

When you ask for a multimeter you get this: http://en.wikipedia.org/wiki/Multimeter
You point the probes (right word?) at the spots where the voltage is i.e. at the back side of the ATX power supply connector.

While a device that you plug in between power supply and mainboard to monitor currents would be really cool I doubt they are widely available. And you can't monitor Vcore on the power supply connector as its generated on the mainboard.

A multimeter may not even be sufficient to diagnose power supply problems because it displays the average voltage and ignores noise or very short spikes that may well cause the system to crash.

Thanks very much for the info. I searched google (yes I did! 😉 ) and found this interesting tutorial -

http://articles.techrepublic.com.com/5100-108 … 11-1056686.html

The article title is called "Power-supply troubleshooting with a multimeter"

I ran a couple of Windows GLide benchmarks - Unreal Tournament and Unreal. The system hasn't changed from before, and that includes the OC'd V5. For 1024x768 resolution, I get:

Unreal Tournament "utbench.dem": 68.73 FPS
Unreal "intro fly over": 117.something FPS

Regarding the 68.73 UT "utbench" benchmark - I read one of my previous posts here, and I notice that I was testing an AGP V5 (not OC'd) on a slightly slower system. It scored 67.2 FPS, which is only about 1.5 FPS less. For my current tests above, it appears that the V5 will not respond with any noticeable FPS performance increase by having a faster CPU (at around ~ 2.5 GHz), and also being OC'd itself. I guess this means that the V5 itself has more or less reached its maximum potential, and it cannot perform any faster.

Also, my previous tests indicate that the PCI V5 is slower than the AGP V5. I wonder if these scores above cannot be improved, even if you put either type of V5 in to a more modern and powerful system?

Old Thrashbarg wrote:

If you're interested in testing that out further, try comparing 17x133 versus 15x150... both are at ~2.2ghz, and they should show to be within a couple percent of each other. You may find a slight difference when using the PCI V5, since I believe it's a little bit bus-limited and scales a certain amount with PCI speed.

I just unlocked my 3000+ rated mobile barton. I used a sowing needle, and opened up an L6 bridge, and I can now choose a multi greater than its default multi. So I can go beyond 16.5, and choose up to 24x. I currently have a humble Voodoo3 2000 video card installed, and so I ran 3DMark 99 Max. It can run all of its tests, except for one. I get these scores:

133 FSB, 17x multi (2267 MHz clock speed) = 5838 (32705 CPU)
150 FSB, 15x multi (2252 MHz clock speed) = 5843 (34363 CPU)

The 3D score is almost identical! Just for a bit of fun, I set the FSB to 140, and set the multi to 18, giving me a speedy clock speed of 2521 MHz. I run 3DMark 99 Max again, and get 5836 (35302 CPU). BTW, in order to capture all of this info in a screenshot, I reran this test in order to produce the image below, and the 3DMark CPU score is less, probably because I left other apps running in the background.

Update: I'm a bit of a moron. I've just realised that the V3 card has reached its limit at this speed, and in order to produce more meaningful results, I need to increase the power of the video card by replacing it with a faster one. Such as the V5. I'll get around to doing that tomorrow...

Yep, definitely looks like you're GPU-bound at this point.

I have replaced the AGP Voodoo3 2000 card, with an AGP GeForce3 Ti 500 card. Here are the 3DMark 99 Max scores:

I'm using the 43.45 drivers. They seem good for GF3 cards. I also increased the CPU's VCore by just "one notch" in the BIOS. It's on 1.7V. Sandra thinks it's 1.72V. Also, the BIOS AGP is set to 4x with an aperture of 128M. I'm using one stick of 256MB PC-133 SDRAM.

Edit: Included table of 3DMark 99 Max scores above rather than text, as a table is a bit easier to read. The last row in the table above is featured below as a screenshot.

Interesting toppic, THX!

Im am currently thinking about my SK41G equipped with Quadro FX 700 as a AMITHLON/WIN9X/2K/Linux Multiboot.
Currently the machine has a Athlon--Mobile CPU with 256KB Cache. I am thinking about changing to a more performat Barton core.

Win2K in 1024x768 FarCry I get about 17-30fps mostly. Need to check via benchmark to get more reliable scores.

Machine: https://archive.shuttle.eu/2004/en/fx41.htm
Mainboard: https://archive.shuttle.eu/2004/en/fx41.htm#mainboardfx4
CPU-Supportlist: https://global.shuttle.com/news/productsSuppo … t?productId=470 (I currently use a mobile-cpu XP-M that is not listed here!)
GPU: https://www.techpowerup.com/gpu-specs/quadro-fx-700.c1351