ragefury32 wrote on 2020-12-20, 18:05:

The "more kosher" way of doing it is by using an ephemeral server, that is, one that you only spin up when you absolutely need to with SMBv1/NTMLv1 support

Or you just use smb.conf options like these which allow for SMB1, 2 and 3 simultaneously:

1[global]
2        min protocol = CORE
3	max protocol = SMB3

I use it at home to connect all of my machines (98SE, 2000, XP and 10) to the same network share without downgrading the security of Windows 10. The network share is not exposed to the internet. Doing SMB on the internet is a sh*tty idea anyway, regardless of the protocol version.

schmatzler wrote on 2020-12-20, 19:34:

Or you just use smb.conf options like these which allow for SMB1, 2 and 3 simultaneously: […]
Show full quote

ragefury32 wrote on 2020-12-20, 18:05:

The "more kosher" way of doing it is by using an ephemeral server, that is, one that you only spin up when you absolutely need to with SMBv1/NTMLv1 support

Or you just use smb.conf options like these which allow for SMB1, 2 and 3 simultaneously:

Copy code to clipboard

1[global]
2        min protocol = CORE
3	max protocol = SMB3

I use it at home to connect all of my machines (98SE, 2000, XP and 10) to the same network share without downgrading the security of Windows 10. The network share is not exposed to the internet. Doing SMB on the internet is a sh*tty idea anyway, regardless of the protocol version.

Yeah, but by doing so, you are allowing older protocols (smbv1/ntlmv1) on the server-side, which defeats the purpose of various vendors disabling them in the first place. It’s your home network and your decision, but I do not allow SMBv1/NTLMv1 on my NAS.
If I do run an SMBv1 share for working with the older machines, I only run them temporarily on a dedicated server off a specific VLAN (hence the emphermal server). I disable it immediately after I am done.

ragefury32 wrote on 2020-12-20, 19:14:

It wouldn't. If it was released in Q3'01 it would not be an 815EM, it'll be an 830MP Almador (which is much better from a raw chipset capability standpoint).
I would also have more respect for Toshiba if it was an 830 or 440ZX machine with a GeForce2 and the legit YMF - the 815EM machines were a bit of a hack job (the Inspiron 8000/8100s also use the same chipset). Same 256MB/slot SODIMM limitations as the 440, 100MHz FSB, and doesn't even show much (if any) improvements in performance to the old 440 chipset despite being 2 years younger. The Sat2805 might be the "ultimate" Win98/DOS retrogaming laptop, but the 384MB max RAM count totally hobbles it for WinXP.

Sat 2805-s402 appears to have been released in April of 2001.

I'm surprised you're down on the mobile 815 chipset. The ATA100 capability on the IDE controller has caught my eye. Going to have to see if I can scrounge an SSD for it now. I tend to leave XP for bigger machines so the 384MB memory ceiling is plenty for me.

Now a Pentium 4-M with maxed RAM (2GB?), Radeon 9000 mobility 64MB, and ESS Allegro might work for an ultimate WinXP/DOS retrogaming laptop. Online specs say the Gateway 600 YG2 is such a machine. I wonder though if it would work in practice. No doubt getting all that to work would be a headache. But then again it might just be the Allegro OPL. 😉

I need to read up PC sound tech. It's not intuitive to me that laptops with chipsets introduced after ac97 are accidentally legacy soundblaster compatible in DOS without an engineer's attention. Who's going to boot DOS on a laptop that shipped with WinME or WinXP?

ragefury32 wrote on 2020-12-20, 19:14:

the 815EM machines were a bit of a hack job (the Inspiron 8000/8100s also use the same chipset). Same 256MB/slot SODIMM limitations as the 440, 100MHz FSB, and doesn't even show much (if any) improvements in performance to the old 440 chipset despite being 2 years younger.

Inspiron 8100 is not the same as 8000. Inspiron 8100 has 815EP chipset, 1.26 GHz Pentium 3-M Tualatin (133MHz FSB), ESS Maestro 3i, can optionally work with 128bit 64MB Radeon 9000 (Radeons 9000's in some laptops are crippled to 64bit bus)

adalbert wrote on 2020-12-21, 01:36:

ragefury32 wrote on 2020-12-20, 19:14:

the 815EM machines were a bit of a hack job (the Inspiron 8000/8100s also use the same chipset). Same 256MB/slot SODIMM limitations as the 440, 100MHz FSB, and doesn't even show much (if any) improvements in performance to the old 440 chipset despite being 2 years younger.

Inspiron 8100 is not the same as 8000. Inspiron 8100 has 815EP chipset, 1.26 GHz Pentium 3-M Tualatin (133MHz FSB), ESS Maestro 3i, can optionally work with 128bit 64MB Radeon 9000 (Radeons 9000's in some laptops are crippled to 64bit bus)

From a system standpoint, it's close to the same - they both use the i815EP chipset underneath, and the 8100 can support Tualatin CPUs. Both use PC100 SDRAM, and both also have the maestro 3i. Dell probably introduced some minor differences between the 2 (possibly bigger EEPROM, different BIOS flashes, a bigger fan, possibly different glue chips underneath, maybe reposition ports on the board) but the build-of-materials are probably very similar. It's certainly not a major retool like the one required to use the Almador...but then Dell might deal with the embarrassment of having a Tualatin/Almador machine beat its early P4 mobile flagship. The early Williamette/Norwood P4s were not terrific performers compared to the P3s they replaced.

As for whether the Radeon M9 is 64 or 128 bit, that depends on how the video daughter board is constructed and which M9 chip they went with. From what I remember the default lowest cost M9 has 32 MB of RAM embedded into the package and connected via a 64 bit internal bus, and then there is one with that M9 but with an extra 32MB of RAM outside of the package on an external 64 bit bus (giving you 64MB with 128 bit on the datapath). There's a 3rd variant with 64MB embedded but with a 128 bit databus. Whether you have the one with 64 or 128 bit data bus is really dependent on how much VRAM was on the machine when purchased - the same phenomenon can also be seen on the ATi Mobility cards from the M1 all the way to the M10 - the rule of thumb is that the more VRAM you buy, the wider your data bus, but there are weird exceptions, and even then, the performance difference isn't noticeable until you tease it out in certain games or benchmarks. The 3DMarks2001SE difference between the M7 (Mobility Radeon 7500) with 16MB and 32MB VRAM on an XGA screen equipped Inspiron 4150 (Norwood 1.7GHz) is only around 500 points, and maybe 5-7 fps on UT2004SE. I have no idea whether the difference is due to having more VRAM loaded, or the wider datapath.

snickersnack wrote on 2020-12-21, 00:59:

I'm surprised you're down on the mobile 815 chipset. The ATA100 capability on the IDE controller has caught my eye. Going to hav […]
Show full quote

ragefury32 wrote on 2020-12-20, 19:14:

It wouldn't. If it was released in Q3'01 it would not be an 815EM, it'll be an 830MP Almador (which is much better from a raw chipset capability standpoint).
I would also have more respect for Toshiba if it was an 830 or 440ZX machine with a GeForce2 and the legit YMF - the 815EM machines were a bit of a hack job (the Inspiron 8000/8100s also use the same chipset). Same 256MB/slot SODIMM limitations as the 440, 100MHz FSB, and doesn't even show much (if any) improvements in performance to the old 440 chipset despite being 2 years younger. The Sat2805 might be the "ultimate" Win98/DOS retrogaming laptop, but the 384MB max RAM count totally hobbles it for WinXP.

I'm surprised you're down on the mobile 815 chipset. The ATA100 capability on the IDE controller has caught my eye. Going to have to see if I can scrounge an SSD for it now. I tend to leave XP for bigger machines so the 384MB memory ceiling is plenty for me.

Now a Pentium 4-M with maxed RAM (2GB?), Radeon 9000 mobility 64MB, and ESS Allegro might work for an ultimate WinXP/DOS retrogaming laptop. Online specs say the Gateway 600 YG2 is such a machine. I wonder though if it would work in practice. No doubt getting all that to work would be a headache. But then again it might just be the Allegro OPL. 😉

I need to read up PC sound tech. It's not intuitive to me that laptops with chipsets introduced after ac97 are accidentally legacy soundblaster compatible in DOS without an engineer's attention. Who's going to boot DOS on a laptop that shipped with WinME or WinXP?

I actually have quite a low opinion on the i815 Solano chipset- Intel delibrately crippled the i815 (i.e. made the RAM limit smaller than the 440 series) to push people into buying Pentium 4s with RDRAM for “professional applications”. It was much less tunable than a old 440 and designed to give Intel a cheaper architecture to market while prepping everyone for the more expensive and "professional" P4. It was the same playbook as the Pentium MMX/II transition (positioning the Socket370/P55C as the cheap solution while marketing the Sock1/P6 as the next big exciting thing. Just like Super Socket 7/K6 kept AMD alive as a viable P6 alternative, the Athlons were a much stronger competitor versus P6/Netburst, and people seriously questioned the sanity of going Intel/RDRAM when the Athlons were kicking ass on cheaper SDRAM, and later, DDR SDRAM. Considering that RDRAM didn't get much traction in the market (well, maybe except that 4MB upgrade card in the Nintendo 64) Intel were forced to use DDR SDRAM on its Pentium 4s, and the i820 memory hub debacle didn’t help them much, either. The Almadors were supposed to the Pentium III's last hurrah - support for 512MB SODIMMs, 133MHz SDRAM support, lower voltages, and just a better chip...except Intel cancelled the desktop version and only did the mobile version (partly because they want to push the P4 and partly because the Pentium 3m-ULV processors don't work well with the Solano).

Eh, I would not get too excited about the ATA100 support on the i815 - back then no laptop hard drives took advantage of it (most are 4200rpm/2MB cache units) and even on the later 7200rpm/8MB units they don't really sustain transfers fast enough to justify ATA100. I think the fastest Seagate Momentus mobile PATA drives in my inventory only did about 40-44 MByte/sec sustained, up to 70 burstable on small unfragmented files. Besides, mSATA to PATA/SD to PATA bridge chips fitted on the common AliExpress adapters max out at around 25-27 MBytes/sec even with DMA turned on, so they are not really going to show you anything spectacular. The big benefit to using them with the old gear is cheaper/younger storage media and better random read/writes and latency values.

As for the "ultimate" machine? Eeeeh, I don't favor Pentium 4s - in fact, I avoid the entire range (Williamette to Prescott) like the plague. Hot running pieces of crap with only 2 clock levels on the Pentium 4m, with all the grace and sophistication of a '70s American muscle car. A good Tualatin will match it, and a Pentium M (with much thermals, better clocking and power management) will beat it, as will the Barton core XP-Ms. There's a reason why the Pentium-Ms ended up being Intel's roadmap to the Core 2s, and not the Netbursts.

ESS Maestro (C600/n600c) - well, that doesn't have ESFM hardware onboard so the FM synthesis were done in software. Some hated it, I didn't mind it (it sounds better than a Soundblaster AWE64 at the very least)...but then, I have Crystal FM (T21), ESFM (TP 560E/240) and a YMF724 in a thin client somewhere, so I didn't really give enough of a crap to obsess over it.

As for the accidental carry-over of the DS-XG to Tualatin? Well, it's either serendipity or Yamaha signing a fixed-price contract with Toshiba for a given allotment of those chips, and Toshiba had to use them up before switching to different vendors. The fact that there wasn't that much Yamaha AC-XG usage in later Japanese laptops were probably a sign that Yamaha were pricing their chips above the myriad of AC97 vendors out there (there were definitely quite some new ones like Sigmatel or Analog devices in additonal to Crystal and/or ESS) and got outbid on parts sourcing down the line. Not that many good Coppermine machines got legacy SB hardware with hardware FM. The Compaq Presario 1800s being one, the Thinkpad T2x with Crystal FM being another, and a few rare machines are DS-XG.

Being a owner of a Toshiba 2805-s402 i can say it is far from perfect. The compromises it makes are low max memory, less powerful geforce 2 go variant. Far from perfect LCD that only does 16m colors even tho the gpu does 32m. weak build quality that effects the entire series such as the hinges. Not the strongest PIII for the time it released.

That being said, compared to laptops of that vinage the GPU is fantastic for retro gaming and even light 3d gaming I say light becasue its performance is probably on par with a tnt2 but with the compatibility of dx7. And games scale correctly.

The audio is fantastic, and while its not ISA or going to play every game. Between windows dos box support, and pure dos support with TSR combined with a excellent wave table in windows it. It has acceptable compromises.

For just running win 98se the ram is adequate, and yes hooking up a SSD to the 815e ide controller helps the laptop considerably

Trade offs are still way less than other laptops of that vintage, which feature simple 2d accelerators plagued with bugs, and window scaling problems, probably slower CPUs and isa sound is the only great thing they may have going for them. But they wont have a GS synth and not all of them even have real OPL3 like the Toshiba. they don't also have a ps2 k/m port that you can use a Y cable and hook a ps2 mouse and keyboard up to the laptop at same time or dual monitor support let alone VGA out. or built in floppy drives.

The laptop has its quirks as well, which are all of its undocumented features. one of which is hardware support for CPU throttling,requires a specific bios setting so you can mess with multipliers. allowed me to downclock far enough to run wrath of earth. Unofficial pure dos support which took me awhile to get working. Abandoned gpu drivers,which require me to hunt down drivers and mod them, cuz last driver is like 7.xx, and nvidia never publicly release geforce 2 go drivers. so that take a long time for me.

TLDR: What you end up with is a laptop that check all the boxes, but does so in a way that makes it less than perfect but more than adequate as long as you work within the constraints. It will go from about pentium 100 speeds to Pentium 800 and probably play 99% of the games released between those time frames, plus waymore.

Warlord wrote on 2020-12-21, 08:29:

The laptop has its quirks as well, which are all of its undocumented features. one of which is hardware support for CPU throttling,requires a specific bios setting so you can mess with multipliers. allowed me to downclock far enough to run wrath of earth. Unofficial pure dos support which took me awhile to get working. Abandoned gpu drivers,which require me to hunt down drivers and mod them, cuz last driver is like 7.xx, and nvidia never publicly release geforce 2 go drivers. so that take a long time for me.

Bump!

You know... I changed my mind 😀 Could you please upload everything that you have on your GeForce machine 😀 I'll download it and keep it on my pc so I won't have to run and search when the machine comes.

i'll dig it out tomorrow mate

ragefury32 wrote on 2020-12-21, 05:05:

Both use PC100 SDRAM

I have 2x256MB PC133 SDRAM installed and it is reported as 132.8MHz in CPU-Z. I also did memory benchmark:

ragefury32 wrote on 2020-12-21, 05:05:

As for whether the Radeon M9 is 64 or 128 bit, that depends on how the video daughter board is constructed

In my model it only uses external memory, 4 BGA chips (HYB25D128323C-3.6 128 Mbit DDR SGRAM) soldered to the video board (not on video chip)

ragefury32 wrote on 2020-12-21, 06:03:

Besides, mSATA to PATA/SD to PATA bridge chips fitted on the common AliExpress adapters max out at around 25-27 MBytes/sec even with DMA turned on, so they are not really going to show you anything spectacular.

I get between 47MB/s and 73MB/s in SiSoftware Sandra 2001

I also did 3DMark2001SE benchmark, here are the results if anyone is curious

ragefury32 wrote on 2020-12-21, 06:03:

snickersnack wrote on 2020-12-21, 00:59:

I'm surprised you're down on the mobile 815 chipset. The ATA100 capability on the IDE controller has caught my eye. Going to have to see if I can scrounge an SSD for it now. I tend to leave XP for bigger machines so the 384MB memory ceiling is plenty for me.

I actually have quite a low opinion on the i815 Solano chipset- Intel delibrately crippled the i815 (i.e. made the RAM limit smaller than the 440 series) to push people into buying Pentium 4s with RDRAM for “professional applications”. It was much less tunable than a old 440 and designed to give Intel a cheaper architecture to market while prepping everyone for the more expensive and "professional" P4. It was the same playbook as the Pentium MMX/II transition (positioning the Socket370/P55C as the cheap solution while marketing the Sock1/P6 as the next big exciting thing. Just like Super Socket 7/K6 kept AMD alive as a viable P6 alternative, the Athlons were a much stronger competitor versus P6/Netburst, and people seriously questioned the sanity of going Intel/RDRAM when the Athlons were kicking ass on cheaper SDRAM, and later, DDR SDRAM. Considering that RDRAM didn't get much traction in the market (well, maybe except that 4MB upgrade card in the Nintendo 64) Intel were forced to use DDR SDRAM on its Pentium 4s, and the i820 memory hub debacle didn’t help them much, either. The Almadors were supposed to the Pentium III's last hurrah - support for 512MB SODIMMs, 133MHz SDRAM support, lower voltages, and just a better chip...except Intel cancelled the desktop version and only did the mobile version (partly because they want to push the P4 and partly because the Pentium 3m-ULV processors don't work well with the Solano).

Oh, you're disappointed 815 didn't have Almador's 1GB max capacity with affordable memory. I can understand that. But if you had a nice 440gx board, wouldn't you miss the 2GB and multiprocessor support?

I think sometimes when you live through an era, you can become captive to your first impressions. I bought my first gaming PC in 1995, a pentium 100 on 430fx. I kept that for a long time but followed new developments pretty closely for a few years after. When you say Pentium Pro, the first words that come to my mind are "expensive" and "sucks at 16bit". When you say Pentium II, I'm thinking "hot" and "looks like a candy bar"*. S3 virge is the renowned "world 's first 3d decelerator". The highest spec 486 gear was budget tier when it was released.

Fortunately we live in the future where it's affordable to play with a lot of this stuff. Lurking on this board has taught me a lot. I now see unlocked Deschutes PII as being pretty cool for overclocking or time machine tinkering. Geforce FX is a terrible D3D 9 GPU family but can be quite nice for running older stuff in high quality. Virge remains a joke for 3D but is now a favorite for 2D DOS games due to its high compatibility. When I hear intel 815, I'm thinking boring, stable, and 512MB. An ideal Win9x platform. The same machine sounds painful for a patched up XP SP3 box however**.

* The first PII Klamaths used up to 42W vs the 10W of my already toasty Pentium 100. I'd never seen heatsinks that big and the candy bar SECC cartridge with holographic sticker struck me as ridiculous. Now the haswell quad core in my laptop uses 47W and I don't care. 🤣.

** My mother's intel 815 Dell dimension 4100 with WinXP was quite a horror when it was finally retired about 2009. And that's after I cleaned it. 🤣

Drivers off my 2805-s402 dumped in thread.
Re: Toshiba 2805-s402 Ulimate retro gaming laptop

ragefury32 wrote on 2020-12-21, 06:03:

snickersnack wrote on 2020-12-21, 00:59:

The ATA100 capability on the IDE controller has caught my eye. Going to have to see if I can scrounge an SSD for it now. I tend to leave XP for bigger machines so the 384MB memory ceiling is plenty for me.

Now a Pentium 4-M with maxed RAM (2GB?), Radeon 9000 mobility 64MB, and ESS Allegro might work for an ultimate WinXP/DOS retrogaming laptop. Online specs say the Gateway 600 YG2 is such a machine. I wonder though if it would work in practice. No doubt getting all that to work would be a headache. But then again it might just be the Allegro OPL. 😉

I need to read up PC sound tech. It's not intuitive to me that laptops with chipsets introduced after ac97 are accidentally legacy soundblaster compatible in DOS without an engineer's attention. Who's going to boot DOS on a laptop that shipped with WinME or WinXP?

Eh, I would not get too excited about the ATA100 support on the i815 - back then no laptop hard drives took advantage of it (most are 4200rpm/2MB cache units) and even on the later 7200rpm/8MB units they don't really sustain transfers fast enough to justify ATA100. I think the fastest Seagate Momentus mobile PATA drives in my inventory only did about 40-44 MByte/sec sustained, up to 70 burstable on small unfragmented files. Besides, mSATA to PATA/SD to PATA bridge chips fitted on the common AliExpress adapters max out at around 25-27 MBytes/sec even with DMA turned on, so they are not really going to show you anything spectacular. The big benefit to using them with the old gear is cheaper/younger storage media and better random read/writes and latency values.

Oh, I'm disappointed to hear those cheap MSATA SSD IDE adapters are so slow. I really like the passive compact flash IDE adapters in my pentium 1 boxes and figured those would be even better. Satellite 2805 aside, there's a Pentium M notebook with IDE on my project shelf that's begging for a SSD. You've got a tricked out thinkpad x31 don't you? Do you have any recommendations for storage?

ragefury32 wrote:

As for the "ultimate" machine? Eeeeh, I don't favor Pentium 4s - in fact, I avoid the entire range (Williamette to Prescott) like the plague. Hot running pieces of crap with only 2 clock levels on the Pentium 4m, with all the grace and sophistication of a '70s American muscle car. A good Tualatin will match it, and a Pentium M (with much thermals, better clocking and power management) will beat it, as will the Barton core XP-Ms. There's a reason why the Pentium-Ms ended up being Intel's roadmap to the Core 2s, and not the Netbursts.

A Tualatin mobility matching the top Pentium 4-M? That I would have to see. Maybe on battery. Pentium 4-M's are Northwoods and max out at 2.6GHz and 35W vs the Tualatin mobility's 1.333GHz and 22W. The 20 stage+ pipeline aside, that quad pumped 100MHz FSB should be tough to beat. I've got a 1.8GHz p4-m in my Inspiron 8200 and found it almost usable on the web in 2019 with appropriate blockers. The CPU being too weak to do video in software was ultimately the deal breaker. The p4-m are kind of cute for Pentium 4's, unlike the higher clocked mobile Pentium 4's (curse Intel's naming scheme) which topped out at 88w!!

I've don't think I've ever actually used a Tualatin. None of my slot 1/ socket 370 desktop boards support it without hacks and I foolishly threw away my fancy active slocket in a purge about 15 years ago. 🙁

Pentium 4 gets no respect. 😉 The Pentium M is a legend. I don't really have an opinion on the mobile Athlons. I've got one in my Presario 900 but its a low end machine. I've never seen one in a nice config.

Speaking of that Presario 900 , it has that ALi southbridge that's supposed to have a Trident 4DWave in it like the SIS630/ 730. That would give some legacy DOS sound support. Unfortunately, unlike the SiS630/ 730 I've never heard of anyone ever getting it to work. I was tinkering with it some early this year before the world turned upside down but no joy. The ATI 320M IGP scales the LCD nicely. Based off the Radeon M6 I think but slower. The Compaq NC4000/ 4010 is also very similar but Pentium M. A good candidate for "fastest but lame" Dos/ Win9x laptop if that ALi southbridge can be convinced to cooperate.

ragefury32 wrote:

ESS Maestro (C600/n600c) - well, that doesn't have ESFM hardware onboard so the FM synthesis were done in software. Some hated it, I didn't mind it (it sounds better than a Soundblaster AWE64 at the very least)...but then, I have Crystal FM (T21), ESFM (TP 560E/240) and a YMF724 in a thin client somewhere, so I didn't really give enough of a crap to obsess over it.

I'm actually okay with how the AWE64 FM sounds, I just wanted to highlight the Allegro's weakness. My first wavetable card, an Ensoniq Vivo 90 simulated it with it's wave bank and I thought it was improvement over my authentic sbpro clone until some wise people on the internet helped me realize it was a POS many years later. Hehe, ignorance is bliss.

I'm curious about that ESS software FM. Does it still work if the machine is heavily throttled?

snickersnack wrote on 2020-12-22, 03:11:

Oh, you're disappointed 815 didn't have Almador's 1GB max capacity with affordable memory. I can understand that. But if you had […]
Show full quote

ragefury32 wrote on 2020-12-21, 06:03:

snickersnack wrote on 2020-12-21, 00:59:

I'm surprised you're down on the mobile 815 chipset. The ATA100 capability on the IDE controller has caught my eye. Going to have to see if I can scrounge an SSD for it now. I tend to leave XP for bigger machines so the 384MB memory ceiling is plenty for me.

I actually have quite a low opinion on the i815 Solano chipset- Intel delibrately crippled the i815 (i.e. made the RAM limit smaller than the 440 series) to push people into buying Pentium 4s with RDRAM for “professional applications”. It was much less tunable than a old 440 and designed to give Intel a cheaper architecture to market while prepping everyone for the more expensive and "professional" P4. It was the same playbook as the Pentium MMX/II transition (positioning the Socket370/P55C as the cheap solution while marketing the Sock1/P6 as the next big exciting thing. Just like Super Socket 7/K6 kept AMD alive as a viable P6 alternative, the Athlons were a much stronger competitor versus P6/Netburst, and people seriously questioned the sanity of going Intel/RDRAM when the Athlons were kicking ass on cheaper SDRAM, and later, DDR SDRAM. Considering that RDRAM didn't get much traction in the market (well, maybe except that 4MB upgrade card in the Nintendo 64) Intel were forced to use DDR SDRAM on its Pentium 4s, and the i820 memory hub debacle didn’t help them much, either. The Almadors were supposed to the Pentium III's last hurrah - support for 512MB SODIMMs, 133MHz SDRAM support, lower voltages, and just a better chip...except Intel cancelled the desktop version and only did the mobile version (partly because they want to push the P4 and partly because the Pentium 3m-ULV processors don't work well with the Solano).

Oh, you're disappointed 815 didn't have Almador's 1GB max capacity with affordable memory. I can understand that. But if you had a nice 440gx board, wouldn't you miss the 2GB and multiprocessor support?

I think sometimes when you live through an era, you can become captive to your first impressions. I bought my first gaming PC in 1995, a pentium 100 on 430fx. I kept that for a long time but followed new developments pretty closely for a few years after. When you say Pentium Pro, the first words that come to my mind are "expensive" and "sucks at 16bit". When you say Pentium II, I'm thinking "hot" and "looks like a candy bar"*. S3 virge is the renowned "world 's first 3d decelerator". The highest spec 486 gear was budget tier when it was released.

Fortunately we live in the future where it's affordable to play with a lot of this stuff. Lurking on this board has taught me a lot. I now see unlocked Deschutes PII as being pretty cool for overclocking or time machine tinkering. Geforce FX is a terrible D3D 9 GPU family but can be quite nice for running older stuff in high quality. Virge remains a joke for 3D but is now a favorite for 2D DOS games due to its high compatibility. When I hear intel 815, I'm thinking boring, stable, and 512MB. An ideal Win9x platform. The same machine sounds painful for a patched up XP SP3 box however**.

* The first PII Klamaths used up to 42W vs the 10W of my already toasty Pentium 100. I'd never seen heatsinks that big and the candy bar SECC cartridge with holographic sticker struck me as ridiculous. Now the haswell quad core in my laptop uses 47W and I don't care. 🤣.

** My mother's intel 815 Dell dimension 4100 with WinXP was quite a horror when it was finally retired about 2009. And that's after I cleaned it. 🤣

Well, yeah. hindsight is always 20/20, and people do look back at the old eras with rose tinted glasses.
my first computer paid for with my own money was based on some spare parts I got for cheap back then. An Abit BX6r2 (440BX based machine), first a Mendocino 300A, then a Deschutes 350 with a Matrox G450/TNT2 m64 (The G450 were great for 2D while the TNT2 ran half-life better). That machine was eventually upgraded to a Coppermine P3 (800MHz)? the Deschutes were good, but the Mendocino Celerons were amazing for their time. The Coppermines felt okay on the 440 but I wanted an old K7 Athlon back then - the KT133s were a good little chipset back then.

Perhaps it was my job as an IT guy working for a computer lab at a non-profit back then, but their Coppermine i815 based machines were all...kinda “meh”. Not sure if it’s just merely the RAM limit, or it’s just that it was going up against similar Athlons and barely reaching parity. I really didn’t expect to see memory speeds and latency stay roughly the same (or maybe slightly better) on the 815s compared to the 440s, which were designed for the Pentium IIs. I really should whip out Benchmarking tools and do some comparison between the 440ZX laptops that I have (TP240, T21 and the C600), this n600c Almador machine and the Banias equipped X31 just for a side-to-side comparison.

Don’t get me wrong, I love the Tualatin P3 cores - I just didn’t like what the heck Intel did with it. Instead of improving it, they went for Netburst (until it hit a dead end and they bet big behind the Pentium-M). Does anyone remember some of Intel’s claims about a 10Ghz Tejas P4 by 2011, which would require a cooler dealing with a TDP of close to 200w?
That was a dystopian future of times past.

I got so fed up with Intel on the desktop that my next machine was a Barton core AthlonXP 2600+, which was combined with a Radon 9800 Pro - first in a semi-crappy case (with a Chinese PSU that failed after 3 days) then with a Shuttle case that featured an nForce2. I also bought a Pentium 4 laptop (Inspiron 4150) and swapped it out for a Pentium-M Banias machine (Inspiron 600m) within 5 weeks, mostly due to how bad battery runtime was on the P4ms and how much better/tunable the Banias were in comparison. I’ll always consider the Athlon XPs and the Pentium-Ms to be ideal WinXP machines, especially when used with low latency disk/SSDs and as much RAM as you can stuff on it, while the Pentium MMX/II/3 and S7/SS7 machines (like the K6-2+) to be similarly excellent Win98SE/2K machines. Even the old P54C machines runs Windows 95 well, or 98 if you manage to shove enough RAM into it.

Well, yeah, some of the old hardware were initially cheap, but not really. You are still dealing with a 20+ year old machine in a field where most consumer electronics are not expected to last beyond its obsolescence curve - it’s like watching the Mars Rovers exceed their expectations by orders of magnitude - sure, it’s great, but it’s on borrowed time, and keeping them running would get expensive.

Guys, help. I'm again worried about the thing I'm waiting for.
I bought Toshiba Satellite 2805-S603. According to specs (https://www.dropbox.com/s/95yb3qpkc375ci2/sat … 503603.pdf?dl=0) it will have Yamaha Sound chip; AKM4543 Codec Chip. No idea what that is. But the driver page has files for YMF754 so I thought I'm good. And snickersnack confirmed that all is good (Re: The quest for the perfect retro laptop: a saga).

But then I checked dedicated thread for drivers (thx Warlord) and saw Prez post (viewtopiqc.php?p=911869#p911869) that AKM4543 is crap.

It will take 2 weeks to receive Toshiba Satellite 2805-S603, but I want to understand what I'm getting? That's the laptop with the same YMF754 as 2805-S402 or I horribly failed?

So 402 states Yamaha YMF754B-R sound chip while 603 just mentioned Yamaha Sound chip. But 603 features MIDI Support, while 402 have no such line...

Guys, what is Yamaha YMF754B-R, and what is AKM4543 🙁

might be the same thing only difference I see is your has 1ghz cpu so its a refresh.

I've installed my standard DOS/Win31/Win9x triple boot system on so many laptops, I have it down to a science and it goes pretty quickly now.

I'll post pictures of my new Thinkpad 240 when I get the chance. But the long and the short of it is: it's a really fun little machine, works perfectly, clearly one of the better Classic Thinkpads, but in terms of compatibility it's the exact same as the 600E but less powerful. Should make a great ultraportable (edit: I spoke a little too soon without thinking. The sound chip is better, an ESS Solo-1 rather than the crap-o-phone Crystal chip in the 600E). More powerful than the 560X, which I have Linux on anyway.

I also found a Thinkpad 770. That will be interesting to try out. More on that after Christmas.

snickersnack wrote on 2020-12-22, 06:37:

Oh, I'm disappointed to hear those cheap MSATA SSD IDE adapters are so slow. I really like the passive compact flash IDE adapter […]
Show full quote

Oh, I'm disappointed to hear those cheap MSATA SSD IDE adapters are so slow. I really like the passive compact flash IDE adapters in my pentium 1 boxes and figured those would be even better. Satellite 2805 aside, there's a Pentium M notebook with IDE on my project shelf that's begging for a SSD. You've got a tricked out thinkpad x31 don't you? Do you have any recommendations for storage?

A Tualatin mobility matching the top Pentium 4-M? That I would have to see. Maybe on battery. Pentium 4-M's are Northwoods and max out at 2.6GHz and 35W vs the Tualatin mobility's 1.333GHz and 22W. The 20 stage+ pipeline aside, that quad pumped 100MHz FSB should be tough to beat. I've got a 1.8GHz p4-m in my Inspiron 8200 and found it almost usable on the web in 2019 with appropriate blockers. The CPU being too weak to do video in software was ultimately the deal breaker. The p4-m are kind of cute for Pentium 4's, unlike the higher clocked mobile Pentium 4's (curse Intel's naming scheme) which topped out at 88w!!

I've don't think I've ever actually used a Tualatin. None of my slot 1/ socket 370 desktop boards support it without hacks and I foolishly threw away my fancy active slocket in a purge about 15 years ago. 🙁

Speaking of that Presario 900 , it has that ALi southbridge that's supposed to have a Trident 4DWave in it like the SIS630/ 730. That would give some legacy DOS sound support. Unfortunately, unlike the SiS630/ 730 I've never heard of anyone ever getting it to work. I was tinkering with it some early this year before the world turned upside down but no joy. The ATI 320M IGP scales the LCD nicely. Based off the Radeon M6 I think but slower. The Compaq NC4000/ 4010 is also very similar but Pentium M. A good candidate for "fastest but lame" Dos/ Win9x laptop if that ALi southbridge can be convinced to cooperate.

I'm actually okay with how the AWE64 FM sounds, I just wanted to highlight the Allegro's weakness. My first wavetable card, an Ensoniq Vivo 90 simulated it with it's wave bank and I thought it was improvement over my authentic sbpro clone until some wise people on the internet helped me realize it was a POS many years later. Hehe, ignorance is bliss.

I'm curious about that ESS software FM. Does it still work if the machine is heavily throttled?

Okay, several points in reply.

First, on the disk I/O performance:
a) Not all mSATA bridge chips are like that - typical of what you buy on AliExpress or randos on evilbay, it's a bit of a crapshoot. I've gotten good ones and I've gotten ones that are downright terrible in terms of throughput. SD/CF to IDE though, they really are hamstrung by the 25Mbyte/sec limit off the KTC chipset, but do keep in mind that it's 25MByte/sec sustained. How many multiples of 25MBytes are you going to transfer over in order to sustain that? Most file copies you'll see will probably happen upon a ton of small files (which is typically random reads/writes) and a few large ones. That's why, paradoxically, a few directories less than 100MByte in total size but with a crapload of small files can take more time to copy over than a large, 300MB file (it certainly does on a mechanical drive due to the drive head being made to sweep the disk surface to grab/write to files). It'll be more than adequate and a substantial improvement to the original drives.

b) I use mSATA to PATA ad mSATA to SD card adapters on my hardware not because of speed benefit as much as latency reduction. You have to remember that 9-25ms latency on a hard drive is par-for-the-course back in the old days, while getting it down to a fraction of a milli is considered very good for those oldies. Even on the 240 with a run-of-the-mill SD card, it's less than half a millisecond, even though, as you can see above, the sustained throughput is not that great.

c) The 440 machines are UDMA33, while anything ICH2 and up are UDMA100.

In the T21 example above, latency is actually very good at 0.06ms (since it's on an mSATA SSD drive), but throughput is around 27-30 Mbytes/sec. There are signs it can go faster on bursts, but I don’t expect it often. The fact that the drive controller is misconfigured in Windows 98 can cost it some performance shortfalls (might be polled I/O? CPU intervention seemed high on transfers, so the DMA might not be working properly. In a way, persistent disk usage in those old machine is a sign that something is not configured correctly - not enough RAM (swapping) or the OS is not caching to reduce disk usage - the name of the game is always to reduce disk I/O whenever possible. Since high capacity PC100/133 RAM isn't expensive at all, I max out and make sure that everything is done on RAM. Note that the 440 chipset on the T21 has fairly competitive memory bandwidth figures even for a 100MHz FSB/PC100 machine, an on the DDR2700 equipped X31 the disk latency is a slight bit higher than the 440 model, but that's to be expected. Just remember that even though the southbridge component of the 440 chipset and above can do LBA48, you rarely need more than 64GB of disk space even for a Win98 install.

Speaking of the X31? It's just a plain Jane 1.4GHz, 2672-C2U model. I have an mSATA to PATA bridge/white caddy in there along with the cheapest 128GB DRAM-less SSD known to Newegg, then loaded the machine up with 2GB of RAM (maxing it out). It works just fine - somewhere between 70-85MB/sec on benchmarks. Not nearly as fast as the 400-600MByte/sec typical of the underlying SSD, and certainly nothing compared to the multi-gig numbers I get on modern NVMe drives.

To be honest, I only use the mSATA->PATA route when the caddy (or the lack of one) does not allow the SD->IDE approach. The mSATA bridge/caddy is round 17 USD and an average 128GB SD is around 25-30 USD, while the SD->IDE adapter is around 7 USD on bulk. A 128GB SD card is only, what, 15-20 USD? That's 45 for mSATA SSD and around 25 for SD, and with SD you don't need a caddy or an adapter to work on it with modern machines (provided that the machine has a slot built-in). What you choose...is up to you. I favor SD cards for easy file transfers.

Second - Tualatin versus P4m versus P-M on laptops. Intel had a philosophy during the P4 era of upclocking the crap out of the chip, reduce per-clock efficiency, use fast I/O and just shoot data through a long pipeline as quickly as possible, heat generation and wattage be damned. That only works if you are dealing with code that has easily guessed branch flows and data that can be constantly streamed in and out...kinda like Photoshop filters, compression and transcoding. If your tasks have lots of small branches and your pipeline gets flushed often (like, AI moves in the UT bots) you'll see stalls and dropped performance. The Tualatin has a shorter pipeline, executes more per clock but is only moderately clocked, so in a way, the Tualatin and the P4ms (The 400Mhz FSB Norwoods) are rather evenly matched. A well configured 1.13GHz Tualatin P3m with plenty of RAM will be able to keep pace with a 2GHz Pentium 4m, trading places here and there. The place where you see the P4m benefit is with places where that 400MHz FSB + the larger memory bandwidth can claw back ground from pipeline stalls. The later 533MHz Norwoods with the even higher clocks though? Those will beat the Tualatins.

The Pentium-M however, is a different animal. It has the faster FSB and memory bandwidth of the P4, the efficiency of the Tualatin, and it has a large cache to reduce the cost of a pipeline stall (even if it's a shorter pipeline versus the P4). It has the EIST (Enhanced Intel Speedstep) which can often buy you an extra 30 minutes to an hour on a 4 hour battery. If you are good with nursing your machine (LCD brightness down, Wifi off, undervolt, fix it to 600Mhz), an X31 with the default internal battery can get 6 to 7 hours.
I remember seeing 1.6GHz Banias (400Mhz FSB initial version) Pentium Ms matching the performance of 2.8GHz (533MHz FSB) Norwood machine (Dell Inspiron 600m versus the infamous Inspiron 5100...which is a crappy DeskNote known for...spawning lawsuits). The Dothans (533Mhz Pentium Ms) with the 2MB L2 will even blast through the Mobile P4 HTs. Both the Tualatin and the Pentium 4s just seem...pointless to own - one is an evolutionary side-step, and the other is a space heater. At the end the real way forward were the Pentium Ms.

As for the nc4000/Presario 900 - oddly enough, I actually have a healthy respect for the underdogs. That Presario 900 might be handy as a cheap hobby machine (especially if you like the Athlons), and the nc4000 is just like an X31 but with a more...corporate fisher-price aesthetic (rounded, painted "metal", think 1996 Ford Taurus). Those are solid components, and I am guessing someone at Houston Metroplex (where Compaq was based at the time) probably saw Intel price gouging for Centrino and told them to take a long walk off a short pier. The ATi northbridge is definitely interesting, the weird Ali southbridge (a refinement of a chip used back in the late Super Socket 7/early-mid Coppermine era)
is a little odd, and the fact that they had to add an NEC chip in there for USB2 support (rather than to use an ICH4m) definitely said some things about them keeping up with the Joneses with stuff from the parts bin. Yeah, I mean, if I see an NC4000/4100 on evilBay at a reasonable price I would probably bid on it. It would make for a nice and light little DOS gaming machine, assuming that you can use pciset to poke at the old control registers for the Trident 4DWave NX silicon buried inside that M5451 SPU and it somehow work. It also seem to use the same disk caddy (major plus for me) and AC adapter as the n600c (preferably less of a garbage fire in terms of loose connector).

Eh, regarding the Allegro? Eeeeeeeeh, that's an interesting question. I am not sure if the Coppermines can be throttled back down that much, and even if it can, I don't think if it'll do strange things. The ESFM wavetable emulation is done internally in the Allegro hardware rather than a big fat TSR like the VIA 686C southbridge audio. I know that on my Wyse WT9450E (the thin client with the Via VT8231 and SB Pro emulation that requires a 36k convention RAM hit), once you drop it below the speed, of, say, a 486 SX25, the audio will be odd. VERY odd.

Edit: Yep, and even with a massive clockdown via cpuspd, ESS Allegro works just fine - in fact, better than the Solo-1 on the 240 or the CS4624 on the T21. It's still not an accomplishment though.

I get ~70MB/s with 32GB KCC-REM-SEC-MZ-MPC256D (some kind of OEM/Samsung drive?) and JM20330 mSATA - IDE adapter on Inspiron 8100, so it's not as terrible as you describe