Thanks. So, I've noticed that the UDMA stability issues I was seeing with the 88SA8052-based SATA to PATA bridge went away with the mSATA version of the bridge (using either Transcend or Samsung mSATA drives); it's pretty weird.
The UDMA issues I was seeing also seemed linked to the quality of the IDE cable itself, eg. some 80-pin cables lead to errors why others work fine.
Well, 1725MHz Barton and 2080MHz Thoroughbred are equal in speed but Thoroughbred is damn hot vs. Barton.
So, I'll keep Barton, it's factory unlocked desktop version, week 29 of 2003.
Tried to get that multiplier-mod working with 2 wires in socket but no go. I cant get it to post, change multiplier in bios to x7 (x15 really) and fsb to 100 so it gives 1500MHz. But, when I start to raise FSB I can only get to 115MHz and after that no post of any kind. Tried lower multiplier and straight to 133FSB but no image. Really strange...
Without any mods the multiplier works flawlessly between 5-12,5.
Tomorrow I'll try to put all the 5 wires in socket, make it x17 and see what happens.
EDIT:
x17 was little too much, needed to bump up voltage. Lowered to x15 and FSB is 142 so little over 2,1GHz now.
This thing is quite fast, benchmark scores gets higher every 1MHz in bus speed.
This is so badly memory bottlenecked, seems it makes really no difference if the CPU clock 1,8GHz or 2,1GHz. Benchmarks shows it's all about FSB.
Thanks. So, I've noticed that the UDMA stability issues I was seeing with the 88SA8052-based SATA to PATA bridge went away with the mSATA version of the bridge (using either Transcend or Samsung mSATA drives); it's pretty weird.
The UDMA issues I was seeing also seemed linked to the quality of the IDE cable itself, eg. some 80-pin cables lead to errors why others work fine.
Note that this is with ICH2, I did not test with the VIA southbridge... But now that I have this working with my mSATA drives, I'll ditch the Intel drive (which probably has compatibility issues) and stick with this setup.
Thanks. So, I've noticed that the UDMA stability issues I was seeing with the 88SA8052-based SATA to PATA bridge went away with the mSATA version of the bridge (using either Transcend or Samsung mSATA drives); it's pretty weird.
The UDMA issues I was seeing also seemed linked to the quality of the IDE cable itself, eg. some 80-pin cables lead to errors why others work fine.
Note that this is with ICH2, I did not test with the VIA southbridge... But now that I have this working with my mSATA drives, I'll ditch the Intel drive (which probably has compatibility issues) and stick with this setup.
Thanks mockingbird. I remember having a similar mouse jerkiness issue in Win98 (see MSI MS-7253 / K9VGM-V (KM890, VIA8237A) , Athlon 64 X2 5050e, Radeon X800 XL) solved with nusb36/U98SeUSB; not using the 88SA8052 mSATA adapter though. I'd be interested to know the result of using the Startech + "ST6006C" combo with the VIA686B.
Thanks mockingbird. I remember having a similar mouse jerkiness issue in Win98 (see MSI MS-7253 / K9VGM-V (KM890, VIA8237A) , Athlon 64 X2 5050e, Radeon X800 XL) solved with nusb36/U98SeUSB; not using the 88SA8052 mSATA adapter though. I'd be interested to know the result of using the Startech + "ST6006C" combo with the VIA686B.
Not having to use the Promise and having a free PCI slot is thanks to you, good sir (not to mention that ICH2 performs better than the Ultra133TX2).
Sure thing... Will report back here with the findings on the 686B.
In short, yes the 686B works, and unlike the VT6421A, it passes TRIM. And for the long of it:
Well, first of all, Windows 98 proved once again that a preliminary Prime95 run under a Linux floppy micro-distro is no substitute for a proper stability test... My mobile 2400+ is not in fact stable at 2133 at 1.55 volts. I just lowered the FSB to 100Mhz instead of experimenting with it at this time, so I could get on with my testing. I may purchase a mobile 2500+ in the near future.
Now, while the 686B does work fine with the aforementioned combo, it does not do so without a caveat or shall we say altogether well. These screenshots speak for themselves (sorry that they're GIFs, I was limited to the primitive conversion filters in 9x MS Paint):
So the 686B has only half of the write performance to that of the ICH2 in this scenario.
Still, I would personally rather use the 686B than an add-in Promise Ultra133TX2, simply because it is generally always better to use the native implementation than it is to use an add-in card, and the speed is still quite acceptable (and not to mention, the read speed is perfectly fine, only the write speed is affected).
VIA 4.43 was installed, and DMA was enabled.
But I'm going with the P4 system over the KT7A, simply because ODCM throttling is so good, the KT7A simply cannot compare with its flexibility in that regard. And my P4 board has SB-Link, so I'm covered there.
This disrepancy between read and write speeds is interesting. On my KT7A, using the Marvell 88SA8052-based mSATA to 2.5" PATA bridge, plus a 2.5" to 3.5" adapter, and a Transcend 128 GB mSATA drive, I get roughly the same read and write speeds, but it tops at 50 MB/s, even though in theory UltraDMA 100 is used (according to the BIOS and Linux kernel log). These speeds are consistent across Win98/XP and Linux.
This disrepancy between read and write speeds is interesting. On my KT7A, using the Marvell 88SA8052-based mSATA to 2.5" PATA bridge, plus a 2.5" to 3.5" adapter, and a Transcend 128 GB mSATA drive, I get roughly the same read and write speeds, but it tops at 50 MB/s, even though in theory UltraDMA 100 is used (according to the BIOS and Linux kernel log). These speeds are consistent across Win98/XP and Linux.
You know, it might be this decades-old IDE 40pin to 44pin adapter that's causing the issues with the Ableconn adapter (it also has the same mouse jerkiness on the 686B)... A new one is on its way in the mail... Will report back.
Your results do look good by the way, but it seems like you're being limited to UDMA66 instead of UDMA100 (despite what you're being shown by the display with regard to that setting). Which controller does your Transcend module use?
Also, the stability issues were caused by a PC133 module that was randomly picked out of a pile for this particular test that happens not to actually work at 133Mhz... Replacing it with a Hynix module fixed everything, and the CPU is indeed stable at 1.55V @ 2133Mhz.
It's great that you found the cause of your stability issues! Indeed not all PC133 modules are created equal 😀
About my mSATA drive, it is the Transcend MSA230S (TS128GMSA230S), with a SM2258-H controller (Silicon Motion). I've also had good results with Samsung mSATA 860 EVO disks (MJX controller S4LR030), but they are harder to find now.
Damn, i've been struggling since 2 months with an abit kt7a non raid 1.3 rev ( used ) plus Tbred B 2600 with bios updated to the last ver (B4) and an epox 8kta3+PRO ( new ).
I decided to assembly it with a 512 mb ram 133mhz, a 320 gb hdd using it in dual boot mode, an awe64 gold, a s3virge 4mb and a couple of v2 12 mb ( win98 ), and a geforce 4 ti 4600 for win xp.
Where am i stuck? I can't get any response from the agp slot apparently. Every time i try to initialize the pc via Agp output ( setting it by the bios ) the pc hangs with 3 beeps and refuses to post, so i can use just the pci slot for video output ( where, instead, s3 and the voodoos are working greatly ) so far. Got via 4in1 4.43 installed, no gain at all.
Win98 & winxp don't even recognize and list the agp card in the device manager.
I have even tried to unmount everything and disabling usb, com1&2, lpt device ecc...but nothing...
So, thinking of a bad caps issue affecting the agp slot, i replaced the abit with an epox ep 8kta3+PRO brand new OEM. Unfortunately, same shit as above is happening, but with a big difference: even if the pc won't boot&post from the agp side ( getting an FF signal from the mobo's led ), when i boot win98 by pci mode for example, it recognize the agp card's presence so that i can install the nvidia drivers but, obviously, following this the pc hangs because of the conflict beetwen the pci card ( s3 virge ) and the agp one. Therefore here we are again.
BUT, the weirdest thing by far , is that if i mount the geforce 4 on an Acorp 7kta3 of a friend of mine ( so the same couple socket/via chipset kt133a-VIA686 as for abit and epox ), using the SAME parts ( ram sticks, psu, ecc...) but the cpu which is a palomino 1700+, i can boot and post using both the pci and thr agp card...
Well, let's try the Usual Suspects 😉 ... check PnP OS is enabled in CMOS Setup so board can allocate resources correctly... if that doesn't help, try a power unplugged and battery out, and short it's terminals, CMOS reset to clear the ACPI tables like you bleached them, and see if that improves matters.
Unicorn herding operations are proceeding, but all the totes of hens teeth and barrels of rocking horse poop give them plenty of hiding spots.
. Unfortunately, same shit as above is happening, but with a big difference: even if the pc won't boot&post from the agp side ( getting an FF signal from the mobo's led ), when i boot win98 by pci mode for example, it recognize the agp card's presence so that i can install the nvidia drivers but, obviously, following this the pc hangs because of the conflict beetwen the pci card ( s3 virge ) and the agp one. Therefore here we are again.
there should be no conflict between agp and pci, most likely computer freezes because of bad agp power.
Well, let's try the Usual Suspects 😉 ... check PnP OS is enabled in CMOS Setup so board can allocate resources correctly... if that doesn't help, try a power unplugged and battery out, and short it's terminals, CMOS reset to clear the ACPI tables like you bleached them, and see if that improves matters.
If i recollect well, i've tried already all of these steps. Again: no gain. Even IF i'm beginning to realize that on the epox the issue either could be a memory addressing problem or an irq affair. Otherwise i cannot explain why the epox won't boot by agp inizialization but win98 recognize it in device manager...
Anyway, soon i will post some screenshots of what i have set up so far in bios...
Update: i forgot to mention that epox bios is 05-2003 rel and that the tbred is working at 16x ( 2133mhz ) according to the optimized defaults after a last cmos clearing
rasz_plwrote on 2022-02-09, 22:25:20 year old Abit, it needs recapping. You can check power lines with oscilloscope if you have access to one. […] Show full quote
. Unfortunately, same shit as above is happening, but with a big difference: even if the pc won't boot&post from the agp side ( getting an FF signal from the mobo's led ), when i boot win98 by pci mode for example, it recognize the agp card's presence so that i can install the nvidia drivers but, obviously, following this the pc hangs because of the conflict beetwen the pci card ( s3 virge ) and the agp one. Therefore here we are again.
there should be no conflict between agp and pci, most likely computer freezes because of bad agp power.
Do you think this last one could be bypassed using and agp with an own power connector like that of a gf 6000 series? Now you make me remember that the user who sold me the abit stated to have used on its own an ati all in wonder...
Can you check with a scope the 5v/3.3v for the agp slot?
Already checked. it says 1.5v. I've even tried to raise it by 0.1. No change at all. Tried everything, agp mem entrance, 1x, 2x, 4x, fast write...nothing and nothing...what about memory addressing? i've tried so far00, ff ( default is DA )...
Last edited by Kaisersoze on 2022-02-09, 22:53. Edited 1 time in total.
rasz_plwrote on 2022-02-09, 22:25:20 year old Abit, it needs recapping. You can check power lines with oscilloscope if you have access to one. […] Show full quote
. Unfortunately, same shit as above is happening, but with a big difference: even if the pc won't boot&post from the agp side ( getting an FF signal from the mobo's led ), when i boot win98 by pci mode for example, it recognize the agp card's presence so that i can install the nvidia drivers but, obviously, following this the pc hangs because of the conflict beetwen the pci card ( s3 virge ) and the agp one. Therefore here we are again.
there should be no conflict between agp and pci, most likely computer freezes because of bad agp power.
Do you think this last one could be bypassed using and agp with an own power connector like that of a gf 6000 series? Now you make me remember that the user who sold me the abit stated to have used on its own an ati all in wonder...
Do you mean in bios voltage control or are you referring to the psu 5+ line? Cause the psu 5+ line has about 25-30 A of output if i'm right ( near 250 of 3.3 + 5 aggregate wattage, but i will check this )...
20 year old Abit, it needs recapping. You can check power lines with oscilloscope if you have access to one.
no such thing as 20 year old brand new 😀 its old stock. Sadly capacitors dry even when unused, especially from that bad Taiwanese caps period
there should be no conflict between agp and pci, most likely computer freezes because of bad agp power.
Do you think this last one could be bypassed using and agp with an own power connector like that of a gf 6000 series? Now you make me remember that the user who sold me the abit stated to have used on its own an ati all in wonder...
Can you check with a scope the 5v/3.3v for the agp slot?
Already checked. it says 1.5v.
oscilloscope https://www.gamersnexus.net/guides/2053-power … e-and-relevance
20 year old dried capacitors from smack in the middle of bad Taiwanese cap era are prime suspect, even if they "look" good. You can either measure power rails with a scope, or use Chinese clone of https://www.mikrocontroller.net/article ... _(English) ESR meter (~$10 on ebay) to measure them directly, third option is spending that $10 on caps instead and just replacing blind hoping for the best.
