buff wrote on 2020-04-18, 19:24:

First teaser. […]
Show full quote

First teaser.

Roughly 1.1Ghz, but heavily OC'd FSB on the Thunderbird.

Stock FSB

Overclocked FSB

That is impressive . I get about 9100 3Dmarks on a 1.4GHz 512K Tualatin on a 815 board (with PC133) using an FX 5900 (Windows 98 SE). Memory bandwidth limitations probably, as said before .

The Serpent Rider wrote on 2020-04-19, 19:58:

Here's global testing of chipsets on Vogons. 440BX and SIS DDR are the fastest chipsets for P6 CPU.

That was the thread I was thinking of!
In my 3DMark01 Megathread, user VooDooMan was able to just about match my VIA Apollo Pro 266T score with his SiS 635T. I'd say both of these DDR chipsets are ahead of BX, at least in 3DMark and Q3A.

My score @ 1575MHz on a QDI Apollo Pro 266T based motherboard was 13,685.

VooDooMan's score on an ECS SiS 635T based motherboard was 13,574, but he was running Win98SE, which typically produces slightly lower 3D'01 scores than XP. We used the same GPU and driver.

My scores with the SDRAM based 694T and i815 were 12,884 and 13,045, respectively.

I wonder how the X1600XT compares to the 6800GT? The X1600XT is the fastest AGP card I currently have. I'll throw it in for a quick test on the platforms tonight.

I'd say both of these DDR chipsets are ahead of BX, at least in 3DMark and Q3A.

They're ahead due to slight benefits of DDR for AGP cards with 4x mode. Then again, difference is too small.

I've been following this thread, and enjoyed the older thread where the different chipsets were pitted against each other.
I don't have a board with the Via DDR chipset, but I'm testing the following:

Asus P3B-F i440BX@133 (SDR-SDRAM)
Asus P3B-F i440BX @150 (SDR-SDRAM)
ECS P6S5AT SIS635T (DDR-SDRAM/SDR-SDRAM)
Compaq AP440 i840 (Dual channel RDRAM)
AOpen Ax6C i820 (RDRAM)
Asus TUSL2-C i815ep (SDR-SDRAM)
DFI CA64-TC Via 694T (SDR-SDRAM)

The Compaq has no options for OC'ing, the ECS supposedly has a BIOS which enables 166Mhz, but I have not seen that floating anywhere.

I try to stay as common as possible:
CPU: 1,4 Ghz Tualatin-S 512kb
OS: WIN98SE, every system a clean install
HDD: 16GB CF on Startech adapter
GPU: stock Asus V9570/TD/256 (56.54 drivers)
RAM: 256 MB of appropriate ram

What the correct way to run the quake2 timedemo?

As the AX6C is throwing a cmos checksum error fit, I had to swap it for an IWILL DS133-R.

After testing the i440BX; i840; i820:

Current winner in 3DMark '99 at stock speed: IWILL with 6025
Current winner in 3DMark '99 CPU at stock speed: P3B-F with 22052

Current winner in 3DMark '00 at stock speed: IWILL with 10347

Current winner in 3DMark '01 at stock speed: Compaq AP440 with 8315

flupke11 wrote on 2020-04-20, 21:47:

As the AX6C is throwing a cmos checksum error fit, I had to swap it for an IWILL DS133-R.

Hi,

I saw your PM, but I don't have enough forum posts to reply yet.

I have the AX6C BIOS v1.04, but I'm not sure if it's the latest one. Send me a PM with an email address, and I'll upload it to sprend.com.

edit: I think this board stores the BIOS settings in a flash memory (removing the CMOS battery won't clear it). I guess you've already tried clearing CMOS with the jumper.

If you happen to find a newer BIOS for this board, please let me know. The max multiplier is 8x (1066Mhz with 133MHz FSB).

Wasn't dual channel SDR supposed to be faster than DDR on Tualatin boards?

buff wrote on 2020-04-19, 22:23:

I wonder how the X1600XT compares to the 6800GT? The X1600XT is the fastest AGP card I currently have. I'll throw it in for a quick test on the platforms tonight.

much slower
6800GT has for example 16ROPs and TMUs, x1600xt has 4, not to mention 256bits ram vs 128bits.

the x1600xt might shine in SM3.0 compared to the 6800GT, but still

anyway 3dmark01 with this level of CPU is heavily CPU limited.

feipoa wrote on 2020-04-21, 00:51:

Wasn't dual channel SDR supposed to be faster than DDR on Tualatin boards?

I'll test my P3TDE with the same config. I suppose with the faster timings, and equal bandwith as single channel DDR-Sdram, it should be marginally faster.
One to add to the list 😊

flupke11 wrote on 2020-04-21, 09:45:

feipoa wrote on 2020-04-21, 00:51:

Wasn't dual channel SDR supposed to be faster than DDR on Tualatin boards?

I'll test my P3TDE with the same config. I suppose with the faster timings, and equal bandwith as single channel DDR-Sdram, it should be marginally faster.
One to add to the list 😊

It's also on my infinite list somewhere.

flupke11 wrote on 2020-04-20, 21:47:

As the AX6C is throwing a cmos checksum error fit, I had to swap it for an IWILL DS133-R. Tualatest1.JPG […]
Show full quote

As the AX6C is throwing a cmos checksum error fit, I had to swap it for an IWILL DS133-R.
Tualatest1.JPG

After testing the i440BX; i840; i820:

Current winner in 3DMark '99 at stock speed: IWILL with 6025
Current winner in 3DMark '99 CPU at stock speed: P3B-F with 22052

Current winner in 3DMark '00 at stock speed: IWILL with 10347

Current winner in 3DMark '01 at stock speed: Compaq AP440 with 8315

Oh, better hide that mains-power-outlet-measurement thingie .. It will only make you sad 🤣 and don't forget - energy spent in retro hardware is energy saved in heating .. 😉

H3nrik V! wrote on 2020-04-21, 18:18:

Oh, better hide that mains-power-outlet-measurement thingie .. It will only make you sad 🤣 and don't forget - energy spent in retro hardware is energy saved in heating .. 😉

Thank the heavens for sunshine and solar panels 😀

The serverworks chipset does not play nicely with Win98SE... 3DMark00 and 01 crash.

I'll test under WinXP and if that works, I'll redo all tests under WinXP. We'll use 512 MB of ram, to properly compare all chipsets under that OS.

@ OP, I don't want to hijack the thread, I'd just like to add some data to give the poor Tualatin some credit in its fight with the mighty Athlon 😀.

edit: The P3TDE6 will have its winXP on a SCSI 36GB disc, as the IDE is throwing fits as well.

edit2: AAARGH, 3Dmark99, 00 and 01SE all crash or halt on this system. I know the AGP on Serverworks chipsets is far from flawless, but what is the real issue here?

Try the AGP at 1x. I have a Tyan board with this chipset which can run dual tualatins and it needed AGP set to 1x, while the Supermicro variant could handle 2x.

feipoa wrote on 2020-04-21, 22:52:

Try the AGP at 1x. I have a Tyan board with this chipset which can run dual tualatins and it needed AGP set to 1x, while the Supermicro variant could handle 2x.

Thanks Feipoa!

Great, that did the trick! The whole board is really stable, and I never encountered these issues before. In AGP Operations, a BIOS setting, you can chose AGP1X/2X/4X. 4X means a beep concert at boot. We can say that 2X means stable unless you run benchmarks 😀
I'll reinstall Win98SE later this week. Off to bed now.

Could you also run them in either W2K or XP? On 2 of 3 of my dual Tualatin systems, I don't have w98SE installed, so I won't be able to compare with your results.

Was able to push the 1Ghz Athlon higher with a copper heatsink. The aluminum blob was letting it get too hot.

buff wrote on 2020-04-19, 22:23:

I wonder how the X1600XT compares to the 6800GT? The X1600XT is the fastest AGP card I currently have. I'll throw it in for a quick test on the platforms tonight.

Quite poorly, ati had to release the x1800 gto to compete with the 6800gs which ties a 6800gt. The x1600xt was a little faster than a 6600gt in sm3, but in older games it would loose to a 6600gt.