VOGONS


AMD K6 3DMARK, aiming for stars.

Topic actions

Reply 160 of 269, by Mentat-vvo

User metadata
Rank Newbie
Rank
Newbie

All my NVIDIA cards and my PCI Radeon 9250 cause reboot of both Ga-5AA & Acer 5Ali61 , running W2kSP4. But, Voodoo5 5500 AGP finally was able to run 3DMark2k 🥴

2A56AB4C-8A94-40DF-8913-EFCC71ABEA1B.jpeg
Filename
2A56AB4C-8A94-40DF-8913-EFCC71ABEA1B.jpeg
File size
1.98 MiB
Views
724 views
File license
Public domain

Dual Opteron285/4Gb/RadeonHD3850/Audigy
2x P3Xeon900-2M/1Gb/2xVoodoo2/Q-DCC/AWE32
2x PPro@250-1M /TyanS1834D /512Mb/VBanshee
K6III+400@624/5MVP3G2/256Mb/GF3ti200@500
Iskra 1031 Issue 2 - i8086 PC/XT Clone from USSR

Reply 161 of 269, by Skalabala

User metadata
Rank Member
Rank
Member

Thats a really bad score 😮 But at lest you got one :p
Have you tried my guide to how to get it stable and fast?
A note to all, Super Socket 7 machines are super temperamental to what ram you use.
This is one of the big reasons why they are so slow, manufacturers had to get the motherboards stable.(making them slow)
Back then ram quality was poor. But later it got better. So now we can use PC133 CL2 6ns ram and then use custom timings to get the K6 faster and stable!
So please use ram that is way over rated for your K6 😁

Reply 162 of 269, by austinham

User metadata
Rank Member
Rank
Member
65C02 wrote on 2020-05-15, 01:40:
austinham wrote on 2020-05-11, 16:08:
from HWbot https://d1ebmxcfh8bf9c.cloudfront.net/u4702/image_id_1257775.jpeg […]
Show full quote

from HWbot
image_id_1257775.jpeg

I don't get it! How do you guys get such high scores? This is better than my Celeron 800MHZ with NVIDIA Ti4200 = 3469 pts. I also saw someone else getting more than 12000 with a P3 which I think my score should be closer to. Is it because I run Windows ME? Should I try 98 or XP instead?

tweaking. most folks on WHbot will spend hrs even days shooting for the best score. And there are so many odd ball tweaks to improve scores. Some even go as far as testings every bios version and theme in the OS. Some even play with the timings of CPU cache.

Reply 163 of 269, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie
Mentat-vvo wrote on 2020-05-15, 09:14:

All my NVIDIA cards and my PCI Radeon 9250 cause reboot of both Ga-5AA & Acer 5Ali61 , running W2kSP4. But, Voodoo5 5500 AGP finally was able to run 3DMark2k 🥴

Did you try using a good ol' GeForce 2 MX or MX 400 (128 bit versions only) with driver 7.76 ? They have very low power requirements (4 - 5 watts) and work great with SS7... albeit, you can't aim for the stars with them 😀 For that you probably need a GeForce 2 GTS / GeForce 3 Ti, but their power requirements are higher.
Your Voodoo 5 most likely works because it uses external power, from the 12V rail.
Also, make sure that you are not using any other devices (like USB2.0 PCI cards - those seem to kill the performance on many SS7 platforms).

2 x PGA132 / 5 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A
5 x Socket 370 / 5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: Ryzen 5 3600X
Backup rig: Core i7 7700k

Reply 164 of 269, by Mentat-vvo

User metadata
Rank Newbie
Rank
Newbie
bloodem wrote on 2020-05-15, 13:44:

Did you try using a good ol' GeForce 2 MX or MX 400 (128 bit versions only) with driver 7.76 ? They have very low power requirements (4 - 5 watts) and work great with SS7... albeit, you can't aim for the stars with them 😀 For that you probably need a GeForce 2 GTS / GeForce 3 Ti, but their power requirements are higher.
Your Voodoo 5 most likely works because it uses external power, from the 12V rail.
Also, make sure that you are not using any other devices (like USB2.0 PCI cards - those seem to kill the performance on many SS7 platforms).

I tried to use PCI cards with same effect. I tried GF4Ti, GF3Ti, GF6200, Radeon 9250 PCI. But, I do use USB 2.0 @PCI . What tool should I use to tweak K6 on NT systems? Thanks for hint on USB! Another thing : My GA5AA detects the CPU w/o plus sign. Maybe that is a problem. My MVP3 board boots XP through the SCSI controller much faster. Maybe should I try Ga-5AX bios?)
Waiting for better mobos to arrive )

Dual Opteron285/4Gb/RadeonHD3850/Audigy
2x P3Xeon900-2M/1Gb/2xVoodoo2/Q-DCC/AWE32
2x PPro@250-1M /TyanS1834D /512Mb/VBanshee
K6III+400@624/5MVP3G2/256Mb/GF3ti200@500
Iskra 1031 Issue 2 - i8086 PC/XT Clone from USSR

Reply 165 of 269, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie

You're welcome!
All the video cards you listed have very high power consumption, between 20 and 40 Watts, so it's very possible that your motherboard simply can't handle that power draw.
I currently have three SS7 boards (and I tested many others), and I can tell you that this is one of the most common problems with the SS7 platform: the AGP and PCI slots can't handle the load for power hungry video cards (Voodoo 3, GeForce 3+ cards, etc) and the behavior can be extremely strange: restarts, freezes, BSODs
So I do suggest you try a GeForce 2 MX/MX 400 (which has at least 4 - 5 times lower power consumption than the cards you have mentioned).

2 x PGA132 / 5 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A
5 x Socket 370 / 5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: Ryzen 5 3600X
Backup rig: Core i7 7700k

Reply 166 of 269, by Mentat-vvo

User metadata
Rank Newbie
Rank
Newbie
bloodem wrote on 2020-05-15, 13:44:

Your Voodoo 5 most likely works because it uses external power, from the 12V rail.

Forgot to say that I also tried FX5200 and even FX5900XT with power connector. Also with crash-reboot.
About the driver version you mentioned , is it for Win9x ? I use W2K.

Update: found 5ali61 rev F mobo w/o tag sram. I believe that’a Final revision of this mobo that allowed using 10-bit chipset tag sram. Also, maybe, the fastest. Gonna implant these chips into the Asus P5A-B.

69780B98-563A-4C01-8692-207823D8859D.jpeg
Filename
69780B98-563A-4C01-8692-207823D8859D.jpeg
File size
212.81 KiB
Views
672 views
File license
Public domain

Dual Opteron285/4Gb/RadeonHD3850/Audigy
2x P3Xeon900-2M/1Gb/2xVoodoo2/Q-DCC/AWE32
2x PPro@250-1M /TyanS1834D /512Mb/VBanshee
K6III+400@624/5MVP3G2/256Mb/GF3ti200@500
Iskra 1031 Issue 2 - i8086 PC/XT Clone from USSR

Reply 167 of 269, by bloodem

User metadata
Rank Oldbie
Rank
Oldbie
Mentat-vvo wrote on 2020-05-16, 05:28:

Forgot to say that I also tried FX5200 and even FX5900XT with power connector. Also with crash-reboot.
About the driver version you mentioned , is it for Win9x ? I use W2K.

Well, the FX series is a no-no for Socket 7. I tried it once and it didn't go well. 😀
As a general rule of thumb, whenever you work with a slow CPU/platform, you always want a video card driver from the same era, one that was also optimized (at least partially) for that slow CPU. And because older drivers only support older video cards (obviously), there's no point in trying newer cards. I know it's a bit counterintuitive (you always expect newer video cards to be faster, right?), but in the case of old motherboard/CPU combos, this is not the case. A newer video card will either not work at all (because it was not tested with such an old platform in the first place - this is also my experience with GeForce FX on SS7, especially with driver versions 5x.xx), or it will be MUCH slower, because of the extreme CPU overhead.
Regarding Win2K, even though I never used it (well, tried it a few times back in the day, but it wasn't my thing), I'm pretty sure you should also find the same drivers for it, although the old ones are probably more difficult to find, since very few gamers were actually using Win2K at the time.

2 x PGA132 / 5 x Socket 3 / 4 x Socket 7 / 6 x Super Socket 7 / 5 x Slot 1 / 3 x Slot A
5 x Socket 370 / 5 x Socket A / 1 x Socket 478 / 2 x Socket 754 / 3 x Socket 939 / 4 x LGA775 / 1 x LGA1155
Current rig: Ryzen 5 3600X
Backup rig: Core i7 7700k

Reply 168 of 269, by Skalabala

User metadata
Rank Member
Rank
Member
Mentat-vvo wrote on 2020-05-16, 05:28:
Forgot to say that I also tried FX5200 and even FX5900XT with power connector. Also with crash-reboot. About the driver version […]
Show full quote
bloodem wrote on 2020-05-15, 13:44:

Your Voodoo 5 most likely works because it uses external power, from the 12V rail.

Forgot to say that I also tried FX5200 and even FX5900XT with power connector. Also with crash-reboot.
About the driver version you mentioned , is it for Win9x ? I use W2K.

Update: found 5ali61 rev F mobo w/o tag sram. I believe that’a Final revision of this mobo that allowed using 10-bit chipset tag sram. Also, maybe, the fastest. Gonna implant these chips into the Asus P5A-B.
69780B98-563A-4C01-8692-207823D8859D.jpeg

Interesting that your external tag is not there. Some of the 5ALi61 have the external tag with the Rev.G ALi.
People have bad luck with GA5AX stability. And I agree. The AGP slot is of the worst of SS7 and the board is slow on top of that.
I think the 5ALi61 is using 9148BF-36 Clock generator. This can not overclock I think.

Reply 169 of 269, by Mentat-vvo

User metadata
Rank Newbie
Rank
Newbie

Just tested GF2 MX400 with the same effect in 5MVP3. Looks like my W2K needs replacement). Just instant reboot.
p.s. Voodoo 5 5500 AGP was a bit faster in VIA-based mobo by ~150 pts ))

Dual Opteron285/4Gb/RadeonHD3850/Audigy
2x P3Xeon900-2M/1Gb/2xVoodoo2/Q-DCC/AWE32
2x PPro@250-1M /TyanS1834D /512Mb/VBanshee
K6III+400@624/5MVP3G2/256Mb/GF3ti200@500
Iskra 1031 Issue 2 - i8086 PC/XT Clone from USSR

Reply 170 of 269, by Mentat-vvo

User metadata
Rank Newbie
Rank
Newbie

Hello again, SS7-aimiers-for-stars! Stage one complete. Looks weird, but here we are: ACOSUS 😀

I've pulled a broken Asus P5A-B mobo (does not start) from Ebay. Post code card says F8.

1.jpg
Filename
1.jpg
File size
86.3 KiB
Views
608 views
File license
Public domain
2.jpg
Filename
2.jpg
File size
131.83 KiB
Views
608 views
File license
Public domain
3.jpg
Filename
3.jpg
File size
38.45 KiB
Views
608 views
File license
Public domain

An Acorp 5ALI61 "donor" w/o tag SRAM from the post above also showed F8 code. It was estimated that malfunction was caused by a dead south bridge. But, it was not dead )) it was just partly desoldered from the mobo.

We have IMPLANTED south bridge from ACORP into ASUS. Revision A1->B1. Now, our ASUS is alive.. But, it cannot detect IDEs.

4.jpg
Filename
4.jpg
File size
118.1 KiB
Views
608 views
File license
Public domain

No problem, a patched BIOS from ACORP delivers 😀 UDMA66 in Asus P5A-B. I also tried enabling 10-bit internal TAG SRAM. It works! And another thing: ONE GIG SDRAM was properly detected (but, hangs) 😀
My Win2K needs replacement by a period-correct Win98SE

5.jpg
Filename
5.jpg
File size
115.31 KiB
Views
608 views
File license
Public domain

Dual Opteron285/4Gb/RadeonHD3850/Audigy
2x P3Xeon900-2M/1Gb/2xVoodoo2/Q-DCC/AWE32
2x PPro@250-1M /TyanS1834D /512Mb/VBanshee
K6III+400@624/5MVP3G2/256Mb/GF3ti200@500
Iskra 1031 Issue 2 - i8086 PC/XT Clone from USSR

Reply 173 of 269, by Lotosdrache

User metadata
Rank Newbie
Rank
Newbie

Hello to everybody here!

Nearly four month ago, Skalabala got in touch with me at my thread about my GA-5AX on overclockers.com and drew my attention to this contest here. I totally forgot that I also created an account here at vogons.org more than four years ago... 🙄

I read this thread and it's all about GeForce 3 (mostly) graphics card and most recently a Voodoo 5 on SS7 platform. I always preferred ATi cards because
1. after 8 years of working/playing on my AMD N80L286-16/S from 1990-1998, my second PC (Intel Pentium II 400 MHz) was equipped with ATi 3D Rage Pro AGP (8 MiB) "assisted" 😉 by Voodoo 2 (12 MiB)
2. ATi always had better support for DVD playback.

Now, led me introduce one of my Super Socket 7 systems first:
- Gigabyte GA-5AX [Rev. 5.2; ALi Aladdin V (Rev. H)]
- BIOS F4
alternatively: BIOS F5 Powered by Lotosdrache
- 512 kiB onboard Cache
- AMD K6-2+/570ACZ@600 (6*100 MHz)
- 2x 256 MiB Mircon Technology SD-RAM PC133U-222-542-Z (MT16LSDT3264AG-13EE3)
alternatively: 3x 512 MiB SD-RAM (vendor: Micron Technology or Xelo)
- 3Dfx Voodoo3 3500 TV incl. breakout cable
alternatively: Matrox G400 Max
- Matrox m3D
- 3Com 3CR990-TX-97
- Terratec EWS64 XXL (Rev 1.2)
- Gravis Ultrasound Max (Rev 2.1)
- IDE-HDD: Maxtor 6Y080P0 (80 GB), Maxtor 4R060J0 (60 GB)
- IDE-ODD: Kenwood UCR-421 True-X 72x, HP CD-Writer Plus (don't know the exact model number)
- 3.5" and 5.25" floppy disk drives

IMG_2100.JPG
Filename
IMG_2100.JPG
File size
1.15 MiB
Views
569 views
File license
CC-BY-4.0

That's how the system is intended to look like when it's finished. For this contest however, I will make a little change: I havn't got any GeForce 3 in my collection because GeForce is sooo mainstream. How boring! So, I looked through my stock and other systems and found a graphics card that should be equivalent to Geforce 3 considering time, features and performance:

ATi FireGL 8800 that found its way to me in an IBM IntelliStation M Pro (Pentium 4 with Rambus RAM) 😀

IMG_2142.JPG
Filename
IMG_2142.JPG
File size
1.92 MiB
Views
569 views
File license
CC-BY-4.0

As you can see, this combination is suffering some problems:
1. Although the chipset revision is H and supports internal tag RAM, Gigabyte still used an external one. The internal tag is deactivated. Therefore, it can only cache 128 MiB.
Solution: I reverse-engineered the BIOS and injected a patch that
a) disables external tag RAM,
b) enables internal tag RAM and now the board caches 512 MiB of RAM,
c) modifies a number of other chipset register settings:
- SDRAM Trc=7T (8T before) and Tras=4T (5T before)
- SDRAM Internal Page detection enabled
- SDRAM Enhanced Page Mode enabled
- LINEAR_WORD-Merge for Frame Buffer Cycle disabled
- Fast NAJ asserted in single write cycle enabled
- SDRAM tRP=2T@FSB100 (4T before)
- AGP Control Register II: Output delay control of AD_STB[1:0] : Default-1nsec=2,5nsec
The following chipset parameters were changed with modbin:
- Separate setting for Trcd and CL enabled
- Trcd=2T select when bit6 = 1 enable
These last two changes have no effect on RAM timings at all. They are just of cosmetic nature 😉

Now, the chipset programming of my Gigabyte GA-5AX resembles Asus P5A nearly 😀

2. ATi FireGL 8800 is a workstation card and there are no drivers for consumer Win98SE/2000/XP.
Solution: I modified the inf-file of Catalyst 6.2 Win98SE driver and installed the card via device manager:
01firegl8800_win98sef3jf8.png 02firegl8800_win98se_mfkbk.png 03firegl8800_win98se_i9jrb.png 04firegl8800_win98se_vgjh8.png
😀

3. I saw that most of you had big problems to get ATi cards running on Aladdin V boards. My first trial on Thursday was a debacle, too. As soon as the system finished rebooting after driver installation, I could not open any program. Windows 98SE crashed immediately.
Solution: I had to limit Primary Frame Buffer Size to 16 MiB!
a) If you prefer using the original Microsoft Windows 98SE AGP driver then you must go into your BIOS-Setup -> Chipset Features Setup -> Primary Frame Buffer -> 16 MB
b) If you like to use ALi AGP driver (version 1.72, 1.82, 1.90, or 2.13 it doesn't matter) there is no need to change your BIOS setting. You can also use ALi AGP Control Center (aka ALi AGP Utility 1.40):
- click Smart Selection "Turbo" -> Apply
- go to "Manual"
- set "Frame Buffer Size" to 16M -> Apply
The other options (except AGP and SDRAM Delay Clock) have no effect on stability at least on my system. Frame buffer must be limited before Catalyst 6.2 installation. That's all. 😀

With this setting, FireGL 8800 runs absolutely stable albeit I only tested 3DMark2000 v1.1 with original BIOS F4, original MS Windows 98SE AGP driver, DirectX 8.1b, and ATi Catalyst 6.2 so far. The result was really disappointing:

BIOS F4
__________MS Win98SE____ALi AGP driver
___________AGP driver______1.72______1.82______1.90______2.13
128 MiB
DirectX 7.0a-------2900----------3019--------2845-------2979--------2837
-------------------------------------------------------------------(with 256 MiB: 2475)
-------------------------------------------------------------------(with K6-III+/400ATZ@550: 128 MiB=3441
------------------------------------------------------------------------------------------256 MiB=3257)


512 MiB
DirectX 7.0a-------2439----------2513--------2445-------2440--------2489


128 MiB
DirectX 8.1b-------2940----------2883--------2987-------2801--------2881


512 MiB
Directx 8.1b-------2482 🙁--------2460--------2469-------2537--------2474
-----------------2410----------2400--------2496-------2461--------2512


BIOS F5 Powered by Lotosdrache
__________MS Win98SE____ALi AGP driver
___________AGP driver______1.72______1.82______1.90______2.13
128 MiB
DirectX 7.0a-------2891----------2822--------2856-------2807--------2912


512 MiB
DirectX 7.0a-------2834----------2858--------2866-------2872--------2811


128 MiB
DirectX 8.1b-------2823----------2873--------2906-------2846--------2930


512 MiB
DirectX 8.1b-------2879 ----------2820--------2923-------2910--------2911
-------------------------------------------------------------------(onbord L2=L3 cache off: 2527)
-------------------------------------------------------------------(1 GiB RAM with MS files: 2474
--------------------------------------------------------------------1 GiB RAM with Rudolph Loew patchmem: 2837)


You see there is lot's of work considering BIOS and chipset settings, AGP and graphics card drivers, RAM capacity, DirectX version...
I will continue to fill this "table" (How can I create a smarter one?) from time to time. First, I have to look for better cooling of AGP voltage converter (it's getting really hot) and CPU (there isn't even heatsink paste between cooler and cpu).

I've got absolutely no idea where this journey ends up. Stay tuned!

Last questions to you:
How often do you loop the benchmark with the aim of cache training?
Are your systems fully equipped with sound cards, network interface cards and so on or do you use minimal configuration with only a graphics card installed?

EDIT 1:
I finished benchmark run with original Gigabyte BIOS F4 (see table above). I did most of the runs only once and that with 512 MiB RAM and DirectX 8.1b twice. However, I think I can make/confirm following remarks/statements:
1. The result range is considerably high. My estimation for a given configuration is: average +/- 50 points. Therefore it's hardly possible to declare a winner.
2. The differences between both DirectX versions are small (128 MiB / 512 MiB):
- DirectX 7.0a: 2923.2 / 2465.2
- DirectX 8.1b: 2898.4 / 2470.1
- Δ..................... -24.8 / 4.9
3. The differences between all 5 drivers are small (128 MiB / 512 MiB):
- M 98SE: 2920 / 2443.7
- ALi 1.72: 2951 / 2457.7
- ALi 1.82: 2916 / 2470
- ALi 1.90: 2890 / 2479.3
- ALi 2.13: 2877 / 2491.7
4. 128 MiB RAM are much faster than 512 MiB RAM (256 MiB RAM gave similar results as 512 MiB):
- 2910.8 <-> 2468.48
- Δ -442.32
5. K6-III+/400ATZ@550 MHz (5.5*100 MHz) is much faster than K6-2+/570ACZ@600 MHz (6*100 MHz) albeit a little bit more unstable (128 MiB / 256 MiB):
- K6- 2+: 2837 / 2475
- K6-III+: 3441 / 3257
- Δ............ 604 / 782
Therefore, more on-die cache is more helpful than higher clock.
6. With K6-2+/570ACZ@600 MHz on GA-5AX, I hardly achieve 3000 points with FireGL 8800 and standard chipset configuration 🙁

To comments about ALi M1541 AGP driver that drove me crazy:
1. Version 1.72 doesn't enable SDRAM Internal Page Detection (Offset 49/ Bit 1) when set to Turbo Mode as newer versions do.
2. All versions disable Force Snoop INV (Offset 43/ Bit 7) upon installation on this board (it's enabled by default through Gigabyte's BIOS) and enable it again after installation of graphics card driver.

To be continued...

EDIT 2:
I finished testing of my BIOS patch:
1. Differences between both DirectX versions are small again (128 MiB / 512 MiB):
- DirectX 7.0a: 2857.6 / 2848.2
- DirectX 8.1b: 2875.6 / 2888.6
- Δ...........................18 / 40.4
3. The differences between all 5 drivers are small again (128 MiB / 512 MiB):
- M 98SE: 2857 / 2856.5
- ALi 1.72: 2847.5 / 2839
- ALi 1.82: 2881 / 2894.5
- ALi 1.90: 2826.5 / 2891
- ALi 2.13: 2921 / 2861
4. 128 MiB RAM are no longer faster than 512 MiB RAM:
- 2866.6 <-> 2868,4
- Δ 1.8
😀
From this I can conclude that
- the patch works and the board is now able to cache 512 MiB of memory with its internal tag RAM 😀
- decreased results with original BIOS F4 are possibly due to smaller cacheable area
5. Onboard L2 cache (= L3 cache with K6-2+) still impacts performance even with K6-2+ and its internal L2 cache
When switching onboard cache (=L3 cache with K6-2+) off, I only got 2527 points.
6. Upgrading RAM to 1 GiB decreased performance again:
- 2474 -> Here I am again out of the cacheable area with MS Win98SE original files.
Interestingly, using Rudolph Loew's patchmem files for systems with more than 512 MiB RAM kept the result high:
- 2837 -> Either the patch limits the system to the lower cacheable 512 MiB, or it reprograms Win98SE to fill the cacheable area first. I don't know.

All in all:
- DirectX 7.0a seems to be less stable on this system than DirectX 8.1b. At least, I got more abnormal terminations with version 7.0a than with 8.1b.
- With the new BIOS, the system seems to be a little bit slower than with the older one albeit of faster RAM timings. At least, I never reached 3000+ points as I did once before with 128 MiB RAM.
Therefore, I have to check whether SDRAM Enhanced Page Mode is really so enhanced. I tested 4 CPUCLKs here. Other settings are 8, 12, or 16 CPUCLKs. However, I never saw a board where this is enabled. Maybe I will disable it again. For this tests I'm going to use a Matrox G400 Max...

Last edited by Lotosdrache on 2020-06-05, 16:43. Edited 7 times in total.

Reply 174 of 269, by meljor

User metadata
Rank Oldbie
Rank
Oldbie

I used a minimal setup, just getting scores without any sound etc. installed.

The really high scores can only be gotten by using a geforce2/3 with earliest drivers as possible. Every other card will be around half the points I guess. Compared to the ATi, I think even a G400 will score better?

PLEASE share that BIOS! I would like to try it out on one of my gigabyte boards...

asus tx97-e, 233mmx, voodoo1, s3 virge ,sb16
asus p5a, k6-3+ @ 550mhz, voodoo2 12mb sli, gf2 gts, awe32
asus p3b-f, p3-700, voodoo3 3500TV agp, awe64
asus tusl2-c, p3-S 1,4ghz, voodoo5 5500, live!
asus a7n8x DL, barton cpu, 6800ultra, Voodoo3 pci, audigy1

Reply 175 of 269, by Lotosdrache

User metadata
Rank Newbie
Rank
Newbie
Skalabala wrote on 2017-05-25, 10:28:

Some candy for Super Socket 7!! Yes those are 64X128K 4ns Cache chips!!

Those 1 MiB chips (64K128) are the same as on my DFI K6XV3+/66 (REV:B2) with 2 MiB L2 cache in total:

DFI K6XV3+_66 REV_B2.JPG
Filename
DFI K6XV3+_66 REV_B2.JPG
File size
1.68 MiB
Views
558 views
File license
CC-BY-4.0

I'm looking for one or two of those chips, too. Why?
It would be interesting to see whether Aladdin V chipset recognizes them as 1 MiB L2 cache or only as 512 kiB chip. According to the chipset data sheet, it is necessary to use two 512 kiB (64K64) chips in order to get 1 MiB L2 cache in total. Those high capacity chips are not mentioned. Maybe because the data sheet is too old?

I would really like to exchange the chip on my Jetway 542C (Aladdin V, Rev. G) to see whether it works or not. If not, a BIOS patch may help. 😉 There are programmable chipset offsets for setting cache size and number of banks. On my board, the 512 kiB (64K64) chip is configured to have one bank. Maybe those 1 MiB are organized in two banks? This information should be mentioned in the data sheet of these chips. Does anyone have it?

Together with the internal Tag RAM and an external 32K8 Tag, the chipset would be able to cache the incredible amount of 4 GiB RAM 😀. That's the dream I have for my Gigabyte GA-5AX. But first as I don't like to risk it, it has to be proven on another board. 😉

meljor wrote on 2020-05-21, 09:57:

PLEASE share that BIOS! I would like to try it out on one of my gigabyte boards...

There is still one last thing to do: At the moment, the individual patches are spread over 3 independent files. I like to summarize them into one single file to save space in the final BIOS. That's not complicated it's just a question of motivation. It will be done at the latest when I finished filling the first table above and proceed with the second one...

Reply 176 of 269, by Mentat-vvo

User metadata
Rank Newbie
Rank
Newbie
Lotosdrache wrote on 2020-05-21, 12:12:

I would really like to exchange the chip on my Jetway 542C (Aladdin V, Rev. G) to see whether it works or not. This information should be mentioned in the data sheet of these chips. Does anyone have it?

Together with the internal Tag RAM and an external 32K8 Tag, the chipset would be able to cache the incredible amount of 4 GiB RAM 😀.

There is no reason keeping external tag sram active. Chipset can handle 2 gigs.

My Asus P5A-B have a space for 0603 SMD component and “1M/512k” text near PBSRAM module. Thus, 1Mb cache must be supported. Gonna implant 1Mb/4ns one as the donor mobo arrives.

Dual Opteron285/4Gb/RadeonHD3850/Audigy
2x P3Xeon900-2M/1Gb/2xVoodoo2/Q-DCC/AWE32
2x PPro@250-1M /TyanS1834D /512Mb/VBanshee
K6III+400@624/5MVP3G2/256Mb/GF3ti200@500
Iskra 1031 Issue 2 - i8086 PC/XT Clone from USSR

Reply 177 of 269, by Lotosdrache

User metadata
Rank Newbie
Rank
Newbie
Mentat-vvo wrote on 2020-05-21, 15:06:

There is no reason keeping external tag sram active. Chipset can handle 2 gigs.

Where does your information come from and how has the chipset to be configured for this?
I refer to the data sheet here pages 45, 50, 77, 82, 121, and 122. 2 GiB are not mentioned there.

Mentat-vvo wrote on 2020-05-21, 15:06:

My Asus P5A-B have a space for 0603 SMD component and “1M/512k” text near PBSRAM module. Thus, 1Mb cache must be supported. Gonna implant 1Mb/4ns one as the donor mobo arrives.

That's an encouragement. 😀
Please, keep me up to date.
Can you show me a detailed picture of your board? I never saw this remark on any board.

Reply 178 of 269, by Mentat-vvo

User metadata
Rank Newbie
Rank
Newbie

The info on 2gigs are from somewhere , the data sheet is a more reliable source, of course.

Here’s the proof on 1Mb of Cache.

575C8290-A7C5-4E5D-9BF7-35EE2D1AE9A0.jpeg
Filename
575C8290-A7C5-4E5D-9BF7-35EE2D1AE9A0.jpeg
File size
255.17 KiB
Views
527 views
File license
Public domain

2gigs Online , but freeze. Must patch bios for this, I suppose. Will try to test 3x1Gb modules.

D6A1B495-65BC-43B5-A1DF-A7C5A9E4988D.jpeg
Filename
D6A1B495-65BC-43B5-A1DF-A7C5A9E4988D.jpeg
File size
1.2 MiB
Views
527 views
File license
Public domain

Dual Opteron285/4Gb/RadeonHD3850/Audigy
2x P3Xeon900-2M/1Gb/2xVoodoo2/Q-DCC/AWE32
2x PPro@250-1M /TyanS1834D /512Mb/VBanshee
K6III+400@624/5MVP3G2/256Mb/GF3ti200@500
Iskra 1031 Issue 2 - i8086 PC/XT Clone from USSR

Reply 179 of 269, by Lotosdrache

User metadata
Rank Newbie
Rank
Newbie
Mentat-vvo wrote on 2020-05-22, 02:56:

Here’s the proof on 1Mb of Cache.
575C8290-A7C5-4E5D-9BF7-35EE2D1AE9A0.jpeg

It's written in the Asus manual. That's really a nice piece of information. Thanks a lot for this 😀
I havn't got an Asus SS7 board so I never read that manual carefully.

Has anyone ever managed to get the Aladdin V BIOS option "Host Read DRAM Command Mode" (aka "CPU read DRAM command mode" or "Fast DRAM Read") working stable when it's set to "Bypass"?
I always get sporadic general exception errors, Windows 98SE Explorer errors and so on. The more the faster the system is tweaked. 🙁
Together with my modded BIOS, the board does not even boot any system whereas it works well when it's set to "Synchronous mode".