That did cross my mind as well. I do have at least 6 or 7 SS7 boards to try and will certainly do so. Just curious as this was my first tested board and feel like it’s optimized as best as can be.

Don’t get me wrong, it’s still pretty awesome getting this performance out of this chip, but something under the surface is apparently gimping the SpeedSys score and was interested to see if anyone else has tried this combination of hardware.

More to come!

Artex wrote on 2024-11-21, 23:39:

That did cross my mind as well. I do have at least 6 or 7 SS7 boards to try and will certainly do so. Just curious as this was my first tested board and feel like it’s optimized as best as can be.

Don’t get me wrong, it’s still pretty awesome getting this performance out of this chip, but something under the surface is apparently gimping the SpeedSys score and was interested to see if anyone else has tried this combination of hardware.

More to come!

If you have it I'd use a GA-5AX rev. 5.2. Same chipset just a earlier revision. You are right, maybe we expect too much from the pentium mmx cpu despite the impressive frequency. I'd suggest to do a quake benchmark on this configuration and then on the other one, it's notorious that is optimized pretty well for intel's fpu and you will be able to see if the problem is the cpu or the asus motherboard according to the result.

I do have 2 of those boards, both Rev 5.2, so I’ll try that next. The fact that we can even run these chips on these boards is amazing in itself. Getting over 60 fps in software mode Quake is sweet stuff.

I recently tested some MBs with Tillamook and found there are significant performace differences between various chipset, here's my table:
http://rayer.g6.cz/hardware/tillamok.htm#CPU_TEST_TABLE
E.g. some MBs with SDRAM performed worse than with FPM DIMM, some run slower at non-standard FSB over 75MHz and more compared to basic 66MHz and such quirks. Finding a good MB that supports Tilly and have good performance is not easy...

That's interesting.
I remember I did a round up of cpus at 200mhz some time ago (I also did a 250 but I didn't test the tillamook because I still didn't have it), with a Aopen AX59 PRO the performance were the same between a standard pmmx and a tillamook.

RayeR wrote on 2024-11-22, 01:20:

I recently tested some MBs with Tillamook and found there are significant performace differences between various chipset, here's my table:
http://rayer.g6.cz/hardware/tillamok.htm#CPU_TEST_TABLE
E.g. some MBs with SDRAM performed worse than with FPM DIMM, some run slower at non-standard FSB over 75MHz and more compared to basic 66MHz and such quirks. Finding a good MB that supports Tilly and have good performance is not easy...

This looks like an amazing deep dive. Thanks for referencing this and putting in all the work. I’ll check this out when I’m fresh tomorrow.

Artex wrote on 2024-11-22, 02:49:

RayeR wrote on 2024-11-22, 01:20:

I recently tested some MBs with Tillamook and found there are significant performace differences between various chipset, here's my table:
http://rayer.g6.cz/hardware/tillamok.htm#CPU_TEST_TABLE
E.g. some MBs with SDRAM performed worse than with FPM DIMM, some run slower at non-standard FSB over 75MHz and more compared to basic 66MHz and such quirks. Finding a good MB that supports Tilly and have good performance is not easy...

This looks like an amazing deep dive. Thanks for referencing this and putting in all the work. I’ll check this out when I’m fresh tomorrow.

Weird results indeed, I'm going to try to compare with the same frequency since it seems they are not so consistent:

1MB 	CPU 	fCPU 	fFSB 	fPCI 	Vcore 	Speedsys CPU/MemBW 	Memtest RAM/L1 	VESATEST 	Quake 320/640*
2GA-5AA 	Tillamook 233 	266 MHz 	66 MHz 	33,3 MHz 	2,0 V 	CPU: 197,84 / 285,26 MB/s 	145 / 174 MB/s 	34 MB/s 	57,8 / 16,0 FPS
3M590A 	Tillamook 233 	266 MHz 	66 MHz 	33,3 MHz 	2,0 V 	CPU: 198,64 / 154,45 MB/s 	- 	18 MB/s 	51,1 / 11,2 FPS
4M577 	Tillamook 233 	266 MHz 	66 MHz 	33,3 MHz 	2,1 V 	CPU: 198,62 / 410,06 MB/s 	102 / 102 MB/s 	43 MB/s 	51,4 / 18,2 FPS
5PA-2005 	Tillamook 233 	266 MHz 	66 MHz 	33,3 MHz 	2,5 V 	CPU: 197,84 / 222,52 MB/s 	138 / 171 MB/s 	55 MB/s 	52,8 / 17,7 FPS
6PA-2005 	Tillamook 233# 	266 MHz 	66 MHz 	33,3 MHz 	2,5 V 	CPU: 197,84 / 222,52 MB/s 	138 / 171 MB/s 	55 MB/s 	54,7 / 17,9 FPS

Apparently the best results are achieved with the GA-5AA, which is a at version of the GA-5AX, if i'm not mistaken?

I definitely need to do another roundup at 250mhz with all the cpus I have now, I will do it when I am in the right mood, that's the link to the very old one: Super socket seven round up at 250mhz

Nemo1985 wrote on 2024-11-21, 23:33:

Weird but the p5a 1.05 and 1.06 are notorious for being bugged with k6-x+ cpus maybe something similar happens with this one?
Do you have any other ss7 motherboard to try with?

Yeah, trying the 1k wt/wb resistor mod would be interesting. Shouldn’t hurt anything on tillamook I wouldn’t think. It may be the issue.

Sphere478 wrote on 2024-11-22, 03:39:

Nemo1985 wrote on 2024-11-21, 23:33:

Weird but the p5a 1.05 and 1.06 are notorious for being bugged with k6-x+ cpus maybe something similar happens with this one?
Do you have any other ss7 motherboard to try with?

Yeah, trying the 1k wt/wb resistor mod would be interesting. Shouldn’t hurt anything on tillamook I wouldn’t think. It may be the issue.

I remember there was an interesting topic about it, but it was tricky I remember something stopped working with such mod, have those issues been troubleshoted and solved?

Nemo1985 wrote on 2024-11-22, 02:59:

Apparently the best results are achieved with the GA-5AA, which is a at version of the GA-5AX, if i'm not mistaken?

GA-5AA has good clock steps settings but has poor memory/AGP/PCI bust throughput.
Look that FIC PA-2005 at 75MHz FSB beat GA-5AA at 83MHz. Crappy Ali chipset vs better VIA.

RayeR wrote on 2024-11-22, 06:07:

Nemo1985 wrote on 2024-11-22, 02:59:

Apparently the best results are achieved with the GA-5AA, which is a at version of the GA-5AX, if i'm not mistaken?

GA-5AA has good clock steps settings but has poor memory/AGP/PCI bust throughput.
Look that FIC PA-2005 at 75MHz FSB beat GA-5AA at 83MHz. Crappy Ali chipset vs better VIA.

You should also consider the pci clock, in the ali chipset at 75mhz is underclocked, while on the via motherboard is overclocked.
It may be interesting to see what's the difference between the ali chipset and the via mvp3.

Nemo1985 wrote on 2024-11-22, 06:11:

RayeR wrote on 2024-11-22, 06:07:

Nemo1985 wrote on 2024-11-22, 02:59:

Apparently the best results are achieved with the GA-5AA, which is a at version of the GA-5AX, if i'm not mistaken?

GA-5AA has good clock steps settings but has poor memory/AGP/PCI bust throughput.
Look that FIC PA-2005 at 75MHz FSB beat GA-5AA at 83MHz. Crappy Ali chipset vs better VIA.

You should also consider the pci clock, in the ali chipset at 75mhz is underclocked, while on the via motherboard is overclocked.
It may be interesting to see what's the difference between the ali chipset and the via mvp3.

The ALi chipset has a /3 PCI divider?

Here is my Tillamook-project:

- 266-MHz-CPU stable overclocked to 440 MHz (FSB 110 MHz x 4 / 2,5 Volt Vcore) using soldered wires directly on the CPU as shown on the previous pages in this thread (no adapter)
- Gigabyte GA-5AA Rev. 3.2, L2-cache of 512 KB enabled/active
- Radeon 9000 (64 MB)
- 256 MB SD-RAM at 110 MHz, CL 2

Results in Windows ME:

- GLQuake: 48,7 FPS
- Quake II: 30,8 FPS

Results in DOS:

- Quake I: 91,7 FPS
- DOOM: 156,6 FPS (477 realtics)
- in System Speed Test 4.78, the performance is similar to a Pentium II 300 MHz (I would expect something slightly better)

And here the rest of the pics.
BTW, the CPU-frequency shown as 450 MHz at POST is wrong - they are real 440 MHZ

Very nice^

Nemo1985 wrote on 2024-11-22, 03:43:

Sphere478 wrote on 2024-11-22, 03:39:

Nemo1985 wrote on 2024-11-21, 23:33:

Weird but the p5a 1.05 and 1.06 are notorious for being bugged with k6-x+ cpus maybe something similar happens with this one?
Do you have any other ss7 motherboard to try with?

Yeah, trying the 1k wt/wb resistor mod would be interesting. Shouldn’t hurt anything on tillamook I wouldn’t think. It may be the issue.

I remember there was an interesting topic about it, but it was tricky I remember something stopped working with such mod, have those issues been troubleshoted and solved?

The p5a 1.05 1.06 has a whole host of issues. Only one of them is solved by that mod.

Search asus p5a simple mod. The thread will pop up.

I swapped to my AOpen AX59Pro and had some much better results, although for whatever reason I had to pump a lot more voltage into the chip to get it stable than when using the Asus P5A 1.06 board. I am not comfortable running it with this voltage for any length of time.

Processor: Intel Pentium 266 w/MMX "Tillamook"
Motherboard board: AOpen AX59Pro (1MB L2 Cache)
Memory: 1 x 256MB PC133 SDRAM
BIOS: R2.36
Chipset: VIA VT82C598MVP (Apollo MVP3)

Processor Mods:

• Green (VCC2DET)
• Yellow (L2 Cache)
• Blue (BF2 4X Multi)

SpeedSys: 303.08
Quake: 74.6 fps

Jumper Config:

CPU BUS Frequency Multiple (4x) (but according to the manual this is a setting for 2X):

• S1: on
• S2: off
• S3: off

FSB (100Mhz)

• Jp4: 1-2 short
• JP5: 1-2 short
• JP6: 2-3 short
• JP25: 2-3 short

Voltage (2.7V)

• S4: on
• S5: on
• S6: on
• S7: off
• S8: off

Artex wrote on 2024-11-23, 17:50:

I am not comfortable running it with this voltage for any length of time.

It would be not very correct what I mean, but, I think, if you stay at any voltage less than 2,8 V (= standard Vcore for the "standard" Pentium MMX P55C, before Tillamook), you will not damage this CPU (or the risk of damage is not too high).
I mean, a Tillamook-chip is not too different from the P55C (even they are differences of e.g. no support of dual-CPU-environment etc.) - the mean reason for using 1,8 to 2,0 Vcore is the laptop-ability for the Tillamook (longer battery operation), as well the 250 nm of Tillamook versus the 350 nm of P55C. Of course, the less the Vcore is, the better.

Artex wrote on 2024-11-21, 23:48:

I do have 2 of those boards, both Rev 5.2, so I’ll try that next. The fact that we can even run these chips on these boards is amazing in itself. Getting over 60 fps in software mode Quake is sweet stuff.

I cannot seem to find the right combination of jumpers + switches on this board. All the documented FSB and multipliers seem to not work correctly with this chip. Anyone have this board and know the correct combination to get to 100FSB with 4x multiplier or even 100FSB with 4.5x for 450mhz (or anywhere inbetween using FSB of greater than 100Mhz)?

Hi, tried my Tillamook on a Shuttle HOT-579 which has jumper settings to set 2,1v lowest voltage. So I set it to 2.1v, measured the output of the power transistor just to make sure I won't fry it, installed and voila.
Bios said Tillamook-MMX and then froze.

So I did the yellow and blue mods (not the green for the forced voltage detection, not the red for the L2 and not the orange), installed it back and it just worked at 4x66=266 😁
Unfortunately, this motherboard is unstable at 83MHz FSB with any RAM and any CPU (tried PC133 SDRam at the slowest settings and a K6-2 500 which works on another board at 83x6), so I ended up at 75x4=300MHz. That way it completed all of the benchmarks just fine.

It a nice little boost, considering that I didn't expect miracles, just an uplifted P-MMX. It of course loses to K6-2 due to raw MHz power of the latter.

GeorgeMan wrote on 2024-11-29, 07:56:

Hi, tried my Tillamook on a Shuttle HOT-579 which has jumper settings to set 2,1v lowest voltage. So I set it to 2.1v, measured […]
Show full quote

Hi, tried my Tillamook on a Shuttle HOT-579 which has jumper settings to set 2,1v lowest voltage. So I set it to 2.1v, measured the output of the power transistor just to make sure I won't fry it, installed and voila.
Bios said Tillamook-MMX and then froze.

So I did the yellow and blue mods (not the green for the forced voltage detection, not the red for the L2 and not the orange), installed it back and it just worked at 4x66=266 😁
Unfortunately, this motherboard is unstable at 83MHz FSB with any RAM and any CPU (tried PC133 SDRam at the slowest settings and a K6-2 500 which works on another board at 83x6), so I ended up at 75x4=300MHz. That way it completed all of the benchmarks just fine.

It a nice little boost, considering that I didn't expect miracles, just an uplifted P-MMX. It of course loses to K6-2 due to raw MHz power of the latter.

Nice!

Maybe disabling onboard hardware like IDE and using external card would help?