Skalabala wrote on 2020-07-17, 16:09:

Dit you try to use higher VIO? P5A/P5AB is bad overclock with stock VIO.

Im running my Asus p5a in 616mhz 112x5.5
I cant get higher. I have not tried raising the VIO. How high do I need to push the VIO?
And whatnvoltage is considered safe ?

I am certainly not using the system 24/7. Maybe 20 hours per year.

What's the procedure to keep 3d Mark 99 Max running when you upgrade Direct X to 8 or nine?

Necrodude wrote on 2020-07-17, 22:31:

Im running my Asus p5a in 616mhz 112x5.5 I cant get higher. I have not tried raising the VIO. How high do I need to push the VIO […]
Show full quote

Skalabala wrote on 2020-07-17, 16:09:

Dit you try to use higher VIO? P5A/P5AB is bad overclock with stock VIO.

Im running my Asus p5a in 616mhz 112x5.5
I cant get higher. I have not tried raising the VIO. How high do I need to push the VIO?
And whatnvoltage is considered safe ?

I am certainly not using the system 24/7. Maybe 20 hours per year.

I ran mine on 4V for a long time and I think its damaged now because of it 🙁
For benching I would say 3.8V is safe. But first try 3.6V to see if that works 😀

Repo Man11 wrote on 2020-07-18, 00:48:

What's the procedure to keep 3d Mark 99 Max running when you upgrade Direct X to 8 or nine?

I have little experience with 99, it should still work if you install DX8 I would think.

Skalabala wrote on 2020-07-18, 11:55:

Repo Man11 wrote on 2020-07-18, 00:48:

What's the procedure to keep 3d Mark 99 Max running when you upgrade Direct X to 8 or nine?

I have little experience with 99, it should still work if you install DX8 I would think.

So here's what discovered; if you install 3D Mark 99 when you have Direct X 7, it works with no problem even if you update to Direct X 9 - but if you update beyond Direct X 7 then try to install 99, it refuses to install (unless there's some work around that I don't know of).

Mentat-vvo wrote on 2020-07-17, 11:13:

It seems that Asus P5A offers a little more. K6-2+ 500 @ 633 and K6-3+ mobile 400 both took 633 @ asus P5A. v1.06 and Gigabyte G […]
Show full quote

It seems that Asus P5A offers a little more. K6-2+ 500 @ 633 and K6-3+ mobile 400 both took 633 @ asus P5A. v1.06 and Gigabyte GA-5AX .
115 FSB is too heavy for a cache module. It needs replacement to a more frequency-capable one in Asus. Gigabyte is ok.
The screen is made on a Gigabyte GA-5AX mobo. I have both ATX Asus & Gigabyte where Asus offers 10% more in the memory bandwidth.
camphoto_684387517.JPG

Yo do know that the 1.05 and 1.06 asus boards perform slow with a + cpu? There is a mod on here somewhere that fixes the issue and makes the 1.06 a great choice as it can cache 512mb without a penalty where the 1.03 and 1.04 run fine with the mobiles but can only cache 128mb.

Repo Man11 wrote on 2020-07-18, 13:32:

So here's what discovered; if you install 3D Mark 99 when you have Direct X 7, it works with no problem even if you update to Direct X 9 - but if you update beyond Direct X 7 then try to install 99, it refuses to install (unless there's some work around that I don't know of).

This is not my experience at all...
Even now, I'm testing an Athlon 64 system, and right after the Win98 SE installation I went ahead and installed DirectX 8.1 + drivers.
After that, I installed 3DMark99, 2000 and 2001 in this exact order. They all worked fine.
And this is not the first time I've done this, never had an issue with 3DMark 99 not installing properly (on older & newer platforms, with Direct X versions 7.0, 8.1 & 9.0)

meljor wrote on 2020-07-18, 17:38:

Mentat-vvo wrote on 2020-07-17, 11:13:

It seems that Asus P5A offers a little more. K6-2+ 500 @ 633 and K6-3+ mobile 400 both took 633 @ asus P5A. v1.06 and Gigabyte G […]
Show full quote

It seems that Asus P5A offers a little more. K6-2+ 500 @ 633 and K6-3+ mobile 400 both took 633 @ asus P5A. v1.06 and Gigabyte GA-5AX .
115 FSB is too heavy for a cache module. It needs replacement to a more frequency-capable one in Asus. Gigabyte is ok.
The screen is made on a Gigabyte GA-5AX mobo. I have both ATX Asus & Gigabyte where Asus offers 10% more in the memory bandwidth.
camphoto_684387517.JPG

Yo do know that the 1.05 and 1.06 asus boards perform slow with a + cpu? There is a mod on here somewhere that fixes the issue and makes the 1.06 a great choice as it can cache 512mb without a penalty where the 1.03 and 1.04 run fine with the mobiles but can only cache 128mb.

My mobo came with this mod, thanks.
A small update : cache module was replaced by 4ns version. Now 115 seems to be ok, must go up to at least 124 MHz of system bus. Runs stable @ 633 in memtest86. Chipset voltage is set to 3.6v.

Update2: I was able to successfully run memtest86 at 630 MHz (140x4.5) with External cache off. Chipset voltage was set to 3.8v.

Mentat, you always get me excited when you post 😁
So did it help to go from 5ns to 4ns? Are you going to try 1024Kb?
I would like to see the performance of Rev 1.04 Vs Rev 1.06 😁

Skalabala wrote on 2020-07-20, 08:49:

Mentat, you always get me excited when you post 😁
So did it help to go from 5ns to 4ns? Are you going to try 1024Kb?
I would like to see the performance of Rev 1.04 Vs Rev 1.06 😁

I can compare a couple of K6-2+ @ 620 or 630 without cache on, because my P5A-B mobo does not go over 100 MHz with cache ON.
Well, it helped to get stable 115 MHz with Cache On but it did not go higher, unfortunately.
I'm not going to replace cache with one meg, until I exactly know what I'm doing.
It could be necessary to install a TAG SRAM module. I have a donor of both tag and 1 meg of cache, but that VIA mobo does not post (yet).

This is the reason why I never installed 1024Kb on my P5AB. We have missing info.
Interesting that your P5AB does not overclock with cache on. Is the problem the clock gen? The Tag? The Cache? The mem controller? Not easy to tell.

Skalabala wrote on 2020-07-20, 10:38:

This is the reason why I never installed 1024Kb on my P5AB. We have missing info.
Interesting that your P5AB does not overclock with cache on. Is the problem the clock gen? The Tag? The Cache? The mem controller? Not easy to tell.

My P5A-B was purchased as a dead one. Noone knows how it was tortured. Maybe, someone ran it at 4.0v for years )))

Skalabala wrote on 2020-07-20, 08:49:

Mentat, you always get me excited when you post 😁

Hope I did it again 😅
However, personally I got disappointed from this Freeway , because of the following:
1) It does not support Plus CPUs (post code 0D). TMC BIOS fixed this.
2) It does not like VIO overvoltage (3.6v no post)
3) It lacks small-step CPU voltage increment (maybe, hidden?) TMC user’s manual helped a lot, small-step increase exists.
4) Clock generator supports max 133Mhz.
5) My specie of FW mobo is stuck at 100Mhz FSB with all the 2 megs of fast 4ns cache. useless.
6) It has badly-designed AGP power . GF3 Ti200 causes system crash and reboot when enters 3D.

And the biggest fail. I’ve downloaded stock bios via awdflash and have patched it for plus cpu support. And flashed it back. Got dead mobo. Error 41. Looks like awdflash failed to download bios file correctly and/or it downloaded only a half of the image, since the binary file is filled with FFs by half (lower part).
Now I’m going to trace schematics to enable second megabyte of cache in Epox 5mvp3G2. And this Freeway will be a donor of 512kbit SOJ TAG SRAM till I design a TAG SRAM doubler board.

So just use a EEPROM flasher to flash it back to the stock and/or reflash it with the modified BIOS.

I wouldn't necessarily just junk the board because of a bad flash.

cyclone3d wrote on 2020-07-21, 14:48:

So just use a EEPROM flasher to flash it back to the stock and/or reflash it with the modified BIOS.

I wouldn't necessarily just junk the board because of a bad flash.

I don’t give a junk ))
Stock bios was wasted by awdflash) I’ve found some bios and will try Plus CPUs with it tomorrow.

Update:
Freeway TI5VGF is a clone of TMC TI5VGF So, I flashed the mobo with TMC bios and , unfortunately , it was unable to run with more than 100 MHz FSB and cache on. Moreover, 3DMark2000 run caused reboot.

Update2:
A DFI baby AT board K6BV3+/66 B1 (1Mb of fast cache) also ran well at 124 MHz FSB as my Epox 5MVP3G2 Mobo. A Sharptooth 450 tried some superPi @500 MHz. However, AGP power is also a problem here. And the mobo is unable to set 112,124 and 133 FSB via jumpers or from BIOS.

Lots of people say that MVP3G5 does not overclock at all. I am wondering if the problem is not signal filters. Or the 32 pin TAG.
We have evidence the the Elite MT 1024Kb chip is stable at overclocking.
We do not have proof of 32pin TAG to be stable with overclock.
Also I think VIA SS7 does not like overvolt on VIO.

Have you seen this?
http://dualsocketworld.blog134.fc2.com/blog-entry-158.html

Skalabala wrote on 2020-07-22, 15:37:

Lots of people say that MVP3G5 does not overclock at all. I am wondering if the problem is not signal filters. Or the 32 pin TAG […]
Show full quote

Lots of people say that MVP3G5 does not overclock at all. I am wondering if the problem is not signal filters. Or the 32 pin TAG.
We have evidence the the Elite MT 1024Kb chip is stable at overclocking.
We do not have proof of 32pin TAG to be stable with overclock.
Also I think VIA SS7 does not like overvolt on VIO.

Have you seen this?
http://dualsocketworld.blog134.fc2.com/blog-entry-158.html

They use ultra version of the board and some PCI power fix mod. However, 600 MHz (6x100) is not much enough)
32-pin TAG SRAM (512 kbit) could be replaced by 1024 kbit (hard to find fast ones) or a couple of 256 kbit (taken from fast donors).
It should be easy to just swap 512kbit TAG from Freeway into Epox to make sure that TAG SRAM is the issue. The datasheets are available. Upper address lines must be grounded.

Here is the fastest TAG
https://www.datasheetarchive.com/pdf/download … rm=SB61L256BT65

Skalabala wrote on 2020-07-23, 08:59:

Here is the fastest TAG
https://www.datasheetarchive.com/pdf/download … rm=SB61L256BT65

It’s funny, but freeway engineers have put 5-volt TAG , a 61512 in their board )) that’s the problem.

TAG SRAM is a generic SRAM in SOJ package. One could design small breakout board with AND logic gate with inverted output to handle /CE of a second SRAM IC when A15 address line becomes active. A couple of of fastest available 3.3v SRAMs (7nS?) in any available package and a dual pin grid with a 1.27mm step.

Mentat-vvo wrote on 2020-07-23, 13:45:

Skalabala wrote on 2020-07-23, 08:59:

Here is the fastest TAG
https://www.datasheetarchive.com/pdf/download … rm=SB61L256BT65

It’s funny, but freeway engineers have put 5-volt TAG , a 61512 )) that’s the problem. TAG SRAM is a generic SRAM in SOJ. One could make small breakout board with an inverter to handle /CE of a second SRAM IC through A15 address line and a couple of fastest available 3.3v SRAMs (7nS?) and pin grid with a 1.27mm step.

Yes with the correct circuit you should be able to make a daughter board. Will be supper cool.
Fastest TAG is 6.5ns 😁 in the link I supplied 😀
But this is in tssop form. If you make a daughter board then there are other SRAM modules as fast as 3ns 😳