Aladdin 7 results

Discussion relating to Socket 7 hardware.
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

Fastest speed I have got for Superpi 1M is 283secs on my EPMVP3G2 running a K6-3+ at 617MHz (112x5.5) with 1M of onboard cache enabled and 384MB of ram running with 2-2-2-5 timings.

Stedman

DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

@stedman
great score too! I am quite sure it is due to the 1mb mainboard cache. according to the tests I undertook on ali-5 boards with mb-cache dis/enabled showed that superpi tremendously depends on second level cache. since the ali-7 boards don't have any mb-cache the good superpi results must be due to 133mhz sdram speed and 128bit memory access.

I really would like to see a superpi bench result for a board with 2mb cache. anybody ready to go for it? there are quit a lot of people here with via-boards (512kb, 1mb, 2mb) ... could not someone start a benchthread like we did on the sis530-thread?

@kalabok and others too

you wonder why agp-mode is disabled .... first I would like to say that "agp-mode" features a lot of things but the most important in terms of speed effects is aka "agp-texturing". both terms are broadly used synonymously. but "agp-texturing" is only the second fastest solution! the fastest solution is using "local" (ram onboard the gfx-card) video ram. so as long as the ram amount needed to render and display the grafics fits into the local memory there is no need for transfering texture data from mainboard ram into gfx-ram. so the bigger the amount of dedicated video-ram the better. the greater the memory interface of the gfx-card the higher is memory bandwidth (32, 64, 128 bit). the higher the memory is clocked the higher is memory bandwidth again. so as long as everything fits into the local video ram there is no difference in graphics speed between a pci card and an agp-card (given all other things kept equal) and agp-mode is not of concern. agp-texturing comes in 2 flavours. the first is the good old "dma" mode, where special ciruitry found on any mainboard (since 8088, 8086 times) manages to move memory around without consuming valuable cpu-time and stalling the cpu to wait for the task completed. it was on aceshardware.com or anandtech.com where I learned that nvidia wrote their drivers to use that "dma"-mode (back some years). they also found out that for example the people behind kyro wrote their drivers to use the "DIME"-mode, which is the second approach of agp-texturing. this mode needs the cpu to be engaged in the process of memory moving around. they (anandtech or aceshardware) argued that contrary to what the marketing people of diverse gfx-cards manufacturers back then broadcastet as the most modern and fastest implementation of agp-texturing namely dime-mode it was in effect plain "dma"-mode that was (still is?) the preferable way to go.

now back to the artx-onboard solution. since the artx video memory is in fact part of the main memory and any other additional memory needed because textures don't fit into that part of the "virtual" video memory is also from main memory I ask myself whether "agp-mode" enabled or not makes any noticeable difference.

User avatar
mamba
Senior K6'er
Posts: 160
Joined: Wed Jun 25, 2003 7:42 am
Location: Italy

Post by mamba »


User avatar
kalabok
Senior K6'er
Posts: 147
Joined: Mon Nov 06, 2006 11:40 am
Location: germany

Post by kalabok »

ok folks,

time to give you an idea what i did during last months ...
... i was creating my own ali 7 pcr table. gave up searching it in the internet ocean, so have a look on it - maybe it will help somebody out there.

mobo used: pc chips m583lmr

some explanations:

1. the cas latency

this offset cannot be switched in windows. it crashes the machine. what can be done is the only tweaking i found out for the ali 7 chipset - set the offset from 40 to C0. it gives about 10% more performance in everest.

2. onboard video and agp

agp is enabled only in win98 (agp 2x, not 4x as stated to work on this mobo). i could not force it to work on win200 nor xp. note, that the video drivers are only working for xp sp1, sp2 refuses to use them. if installed, agp is disabled but high resolution and 32bit are accesible.
the video amount can be found in two offsets of the pcr.

3. ram

use two equivalent dimms! it gives more performence (dual channel mode?) than only one dimm installed and is more stable than two different dimms. the blue marked fields in the picture above show offsets, that changed installing different amount/type and number of rams. of course more investigations could be done there.

4. system divider

there is no offset - as for the mvp3 - that determes the fsb divider. no change in booting with 66 100 or 133 mhz.

hmm ... ok. have a look on it. i hope, it is somehow clear ...
Attachments
ali pcr.PNG
ali pcr.PNG (36.43 KiB) Viewed 8836 times

User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

Very nice work. I have found some hidden BIOS settings on the 583LMR using AMIBCP that may be helpful for more tweaking. I have modified the BIOS to unhide these settings. I'll report more on this later. I haven't had much time lately to do more testing myself.

Here's an article I stumbled on with a referance to the dual channel function of the ALI 7 chipset:

http://www.vanshardware.com/articles/20 ... nForce.htm

Scroll down to the bottom and it reads:
One last note, ATi is quietly working on an advanced integrated chipset. Through its ArtX subsidiary, ATi already has some experience with a dual channel SDRAM chipset, the ALi Aladdin 7 designed for the AMD K6-III+. Ironically, ATi now appears to have wedded itself with Intel. Perhaps ATi will be the source of an integrated P4 chipset that can begin to compare with nForce.

lazy_kalabok

Post by lazy_kalabok »

interesting too,

which different settings did you found ? it would be great if you could send me that naked bios!

hope to hear some results from you, uranium!

User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

Here some of the hidden chipset settings that I found with the default settings in bold print.

Memory DATA Current
05/07/09/0F/1F/3F

Memory Address Current
05/07/09/0F/1F/3F

RAS to CAS Delay
3 Clock/2 Clock

RAS Precharge Time
2 Clock/3 Clock

Write Recovery Time
2 Clock/1 Clock

Min Page Active Time
5 Clock/6 Clock

CPU-PCI Write Posting
Enabled/Disabled

I0Q1Deep
Disabled/Enabled

NA Pin
Enabled/Disabled

DRAM Configuration
Safest/Fast Micron

User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

I was able to get some testing in on the M583LMR this weekend. Performance was not much different from the P561A.
Here's a few quick observations:
-The 2x/6x multiplier remap doesn't seem to work on this board. When I setup for 600 (6x100) it boots at 200Mhz (2.5x80)??? The multiplier, FSB and voltage settings are jumperless and done in the BIOS.
-This board didn't like nearly all of the PCI video cards I tried, just like the P561A. I did get a 4MB 3Dlabs Permedia2 to fully work with better cooperation than with the P561A. A PCI video card didn't seem to improve overall system performance.
-The onboard ArtX video has only a max of 32MB of shared memory available compared to 64MB available on the P561A. Performance was no better with only one driver version available. Most 3D video benchmarks crashed even with conservative BIOS settings.
-The 124 FSB worked but with many Memtest errors. The 133 FSB would not work. This was done in dual-channel mode. I'll try single channel later.
-Memory Read results were similar to the P561A but Memory Write results were higher probably due to the additional chipset settings available.

I'll post some numbers tonight if I get a chance.

lazy_kalabok

Post by lazy_kalabok »

hello uranium,

great, we have similar boards, i am testing mine right now with the modded bios u mentioned. here are my observations:
Uranium235 wrote:-The 2x/6x multiplier remap doesn't seem to work on this board. When I setup for 600 (6x100) it boots at 200Mhz (2.5x80)???
yes, its true - the x6 multi works only with cpucool in windows. i dont know, why the cpu inst remapping the x2 as x6, but the 2,5x80 are k6-III+ power saving features, which are working on every board. the cpu is running at the 4/5 speed of fsb u set manually.
The multiplier, FSB and voltage settings are jumperless and done in the BIOS.
but note one interesting thing - there are pins (allthough not soldered on) and settings description on the board.
This board didn't like nearly all of the PCI video cards I tried, just like the P561A. I did get a 4MB 3Dlabs Permedia2 to fully work with better cooperation than with the P561A. A PCI video card didn't seem to improve overall system performance.
hmmm ... no, not had this experience. i have only one video card - the blaster banshee (creative) 16mb video size, and the everest numbers do encrease if i use pci video, not onboard agp. what do you mean by "dindnt like video cards". this - "Most 3D video benchmarks crashed even with conservative BIOS settings"? i had no problems at all - even @133mhz. ok, depending on what benchmarks u were usiing.
-The onboard ArtX video has only a max of 32MB of shared memory available compared to 64MB available on the P561A. Performance was no better with only one driver version available.
yes, with the normal bios. the bios i modded, as u did also, had the option to select "reserved" in the video amount selection. the board used 64mb video onboard then ...
-The 124 FSB worked but with many Memtest errors. The 133 FSB would not work. This was done in dual-channel mode. I'll try single channel later.
very improtant, uranium!!!: set the agp aperture size to the minimum availible (4mb for me). the system gets by far more stable. it bootet even win98 @133 fsb, and seemed to be stable - ok, not rock solid, some strange errors appeared from time to time, but 98 did not crash.

ok, good luck, uranium!

User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

I must be using a different BIOS version than yours because I don't have "reserved" as an option for the video memory. What is your BIOS version dated as?

The PCI video cards I tested are relatively newer ones that may have some compatibility issues with the M583. A Geforce4 MX4000, a Radeon 9100 and a Radeon 9000 that I have would not let the board POST. The older PCI cards I have, a Permedia2 and a S3 Virge worked.

lazy_kalabok

Post by lazy_kalabok »

i downloaded the 2k0925s version, modded it with the tool u mentioned and it worked for me - dont know, there was no "reserved" selection in amibcp, but after the flash i got them additionaly. now i am able to select 64mb onboard video ...
... as i do not have any experience with newer pci video cards, i cant help you concerning this. maybe they need too much power?

some benchmarks i did last night for my 500acr and ali7:

k6-III+ 500acr @ 133x4,5 (600mhz)
2x 128mb apacer pc133 2225
16mb blaster banshee, (2mb onboard video - i cant disable it)
winxp pro:

everest
446 read
191 write
147,2 latency

superpi 1M 4:35

hotcpu 1276

cpumark 62,5

uranium, can you confirm those numbers? are u using xp?

seems to be a powerful chipset if onboard video is disabled ...

User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

Those numbers are outstanding and the best I've seen anywhere. :) I'm just trying to duplicate those results. I had been using the 01525S.rom BIOS version. I'll try the version your using, maybe it's more stable. Thanks.

User avatar
KachiWachi
K6'er Elite
Posts: 507
Joined: Wed Sep 21, 2005 10:53 am
Location: Pennsylvania, USA

Post by KachiWachi »

OFF TOPRC (but related)

@superpi -

Does this program create any known registry entries, etc...?

Thanks.
Moderator - Wim's BIOS

PC #1 - DFI 586IPVG, K6-2/+ 450 (Cyrix MII 433), 128 MB EDO. BIOS patched by Jan Steunebrink.
PC #2 - Amptron PM-7900 (M520), i200 non-MMX, 128 MB EDO
PC #3 - HP8766C, PIII-667, 768 MB SDRAM
PC #4 - ASUS P3V4X, PIII-733, 256 MB SDRAM
PC #5 - Gateway 700X, P4-2.0 GHz, 768 MB PC800 RDRAM
PC #6 - COMPAQ Evo N1020v laptop, P4-2.4 GHz, 1 GB PC2700 DDR
PC #7 - Dell Dimension 4600i, P4-2.8 GHz, 512 MB PC2700 DDR
PC #8 - Acer EeePC netbook, Atom N270 @ 1.60 GHz, 1 GB RAM
PC #9 - ??? ;)

lazy_kalabok

Post by lazy_kalabok »

@ kachiwachi

... as far as i know, no, it does not. its only a program determing the calculation time needed for different number of pi digits. it creates only some files during the process, new registry entries seem to be too complicated for a 500kb program ...

User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

I agree with Kalabok on SuperPI. It's just a simple math calculation which probably only requires a batch file to execute. One way to be sure is to try a program like JV16 Powertools which can show any registry changes after it's installed.

Last night I flash the BIOS to the 2k925s version and retested at the 124 and 133 FSB's. No change at all...124 POSTS but freezes almost immediately and 133 will not power up. I'm thinking it's my RAM, I'm using 512MB (2x256) which may be more than the chipset can handle at high speed. I don't have any smaller PC133 in a matched pair.
Kalabok, what RAM are you using?

Post Reply