So, a single core 65nm with 2 megs of cache is the way to go ... I'm impressed, otoh, there's like 33% less thermal output compared to Meatball's 4.66 dual core ... 🤣

It seems we should add a test to take advantage of the multi threaded netburst cpus. Any suggestions? Cinebench 11.5?

supercordo wrote on 2023-12-19, 00:38:

It seems we should add a test to take advantage of the multi threaded netburst cpus. Any suggestions? Cinebench 11.5?

Things are great as they are, but it sounds good. Perhaps memory performance can be measured also; Sandra or something?

This is getting to be a lot of work for you, though.

Meatball wrote on 2023-12-19, 17:11:

supercordo wrote on 2023-12-19, 00:38:

It seems we should add a test to take advantage of the multi threaded netburst cpus. Any suggestions? Cinebench 11.5?

Things are great as they are, but it sounds good. Perhaps memory performance can be measured also; Sandra or something?

This is getting to be a lot of work for you, though.

I dont mind.

Dammit, I had CineBench 10 running but didn't bother writing down scores 🤣

Merry Charismas, Everyone!

supercordo wrote on 2023-12-25, 05:39:

Merry Charismas, Everyone!

Thanks a lot, as soon as I finish the build with ASUS P4 NW 2.66@3.33, I'll try to do some benching to participate in this thread, then I have several P4s that I just have to put back into operation.

I think there are not many tests with SDRAM, so even though there is not much movement in this thread, I share my tests.
The CPU-Z voltage is wrong, my DMM reads 1.69v on the processor's VRM coils.
Since it is stable with less voltage and I think the limit is in the chipset, not the CPU, I made a small undervolt, that's why the voltage is low.
The graphics card is a Geforce 3 from Leadtek. The RAM memory is 7.5ns with Infineon chips.
It is not an extreme overclock, it is designed for 24x7 use and the computer case is closed. There are no modifications in terms of driver configuration, I had to use Powerstrip because the latest version of Rivatunner did not correctly detect the GPU frequency and I did not want to waste time. I will end up modifying the BIOS of the graphics card with these frequencies.
The motherboard is a polymoded Epox 4B2AE (https://theretroweb.com/motherboards/s/epox-pronix-ep-4b2ae).
CPU 16x128mhz, 512mhz FSB and RAM 170mhz.
The maximum stable FSB frequency is 136mhz but only with one RAM memory module, when adding the second one it starts to give errors in Memtest, I chose to stay at 128mhz because 128x4=512, and I found it funny, and I was looking for the stable 24x7 overclock to be able to use the computer calmly.
I opted for Windows ME because I thought it looked good on a Pentium 4 with SDRAM, both equally strange.
Edit, added photographs of the motherboard and graphics card.

That's not a bad overclock for a 1.6 socket 423!

H3nrik V! wrote on 2024-02-10, 20:57:

That's not a bad overclock for a 1.6 socket 423!

Thank you, it is my first experience with Socket 423, at the time I ignored it, only 462,370 and 478.

Scoreboard updated!

Here comes Cedar Mill at 5GHz!

P4 631 @ 5GHz, GTX980, Asus P5Q
3DMark 03 = 62967

I don't know how you guys are beating my 5.0 with your 4.6's, but I'm winning the Gigahertz race 😜

pixel_workbench wrote on 2024-02-14, 08:12:

Here comes Cedar Mill at 5GHz! […]
Show full quote

Here comes Cedar Mill at 5GHz!

P4 631 @ 5GHz, GTX980, Asus P5Q
3DMark 03 = 62967

I don't know how you guys are beating my 5.0 with your 4.6's, but I'm winning the Gigahertz race 😜

Wouldn't it be courtesy of GPU?

pixel_workbench wrote on 2024-02-14, 08:12:

Here comes Cedar Mill at 5GHz! […]
Show full quote

Here comes Cedar Mill at 5GHz!

P4 631 @ 5GHz, GTX980, Asus P5Q
3DMark 03 = 62967

I don't know how you guys are beating my 5.0 with your 4.6's, but I'm winning the Gigahertz race 😜

I was already at 5014 with Cedar Mill. Keep pushing.

Scores are updated.

Now it's ATI's turn, the FireGL 8800 with its 3.3ns memory is one of the fastest R200s, my graphics card already has a fairly high base frequency for the GPU with very little headroom, the memory has more headroom.
Both the Geforce 3 and the FireGL were released in 2001.
The R200 takes an advantage in 3DMARK2001.
The driver is forced installed as if it were a Radeon 8500 to have the best compatibility.

H3nrik V! wrote on 2024-02-14, 08:50:

Wouldn't it be courtesy of GPU?

So far I'm getting around 63000 with any one of these graphics cards: gtx980, gtx780ti, gtx580.
I'm thinking it's more related to this rule here "* Driver tweaks are encouraged." which can go to some ridiculous extent that I'd never use in any actual games.

Got another one.

3DMark 2000 = 27019
*P4 631 @ 5.07ghz
*2x 1GB DDR2-1126
*7900gtx
*Asus P5Q
*Win XP SP3

Don't know what magic voodoo it takes to break 30k on a P4, but I doubt the clock speeds are holding me back.

I think I remember reading that SP3 actually takes some performance overhead as compared to SP2? For my own benchmarks in this thread, experimenting with older drivers also has a positive impact ...

H3nrik V! wrote on 2024-02-16, 06:13:

I think I remember reading that SP3 actually takes some performance overhead as compared to SP2? For my own benchmarks in this thread, experimenting with older drivers also has a positive impact ...

This is true, XP1 or RTM both have even less overhead, and yes older drivers also lessen driver overhead issues.