Question Can a CPU bottleneck a drive's speed?

[Elia1995]

Hello, I'm on my quest of building retrocomputers and in a Pentium II 233MHz I put an SSD with a SATA to IDE adapter and installed Windows 98SE on it.
The difference in loading times is noticeable compared to a SCSI HDD I was previously using in that build (now I put OS/2 on that drive), but I also noticed that copying files or even installing games from within the SSD is a lot slower than how it should be, so I was wondering, is it possible that the CPU can affect the writing/reading speeds of the SSD, or the adapter kinda limits it?
Of course an SSD in a retro build isn't the ideal way to go and it's "unfitting", but I like experimenting with hardware, especially making retro-modern Frankensteins like that 😅

 

[Roland00Address]

Yes it occurred in the Core 2 Duo era. Your cpu is much slower.

Also that os does not have trim functionality so certain tasks will always be running the ssd at not ideal write speeds.

 

[Elia1995]

On which CPU should I expect to start seeing a bit of improvement in speed with that SSD and Windows 98SE?

 

[Tech Junky]

This is just a waste of resources.

Retro these days is coffee lake not a PII 233mhz. That's just archaic. Your OS choices as well aren't a good idea as they're unpatched and vulnerable to attack if attached to the internet. Kids and idiots are constantly scanning for devices like this to attack.

Difference between IDE and SATA - GeeksforGeeks

A Computer Science portal for geeks. It contains well written, well thought and well explained computer science and programming articles, quizzes and practice/competitive programming/company interview Questions.

www.geeksforgeeks.org

.

I've done a similar drive upgrade on my car's MMI from a spinner to mSATA and it was more for peace of mind to not have an aging HDD in the dash but, to get a snappier response from the apps running on the system. The limitation though is the host interface being capped at 100MB/s but the drive itself can hit 500-600MB/s using SATA over IDE.

Now, as to how the OS handles things it shouldn't matter much since you're OS sees the drive as IDE.

Win98 is probably older than you and has been EOL for quite awhile. It's a crappy OS compared to where we are today. It's a PITA to deal with in the least. Even W7 was a POS to deal with compared to W8-W11.

You're making your system a target for attack. If you insist on playing with this sort of thing make sure your machine is isolated to a separate VLAN / network to make sure you don't expose your other gear to vulnerabilities.

 

[Elia1995]

Well, technically Windows 98 is 2 years younger than me.

Also, of course I don't plug it to the internet directly, all of my retro builds are set up in LAN, from where I can just locally transfer the files from one machine to the other.
Some of the oldest ones I got them in a parallel port local network, so I'm not even using Ethernet (I found it extremely hard to find Ethernet adapters for 286 machines and the ISA cards I have are just 56k modems which aren't good for a local network, so I just went full parallel network on them)

This specific machine, I have it in local network with my Pentium III, so I can transfer files directly to it from my most stable Win98 build.

Btw it's extremely hard, if not impossible, to find viruses for 98 these days, as hackers are mostly focused on XP and modern systems (proof is my PIII which I connected to internet at some point to test Quake 3's multiplayer on that beast, tho the system is still stable and perfectly healthy)

For me, building retrocomputers is a lot better and fun than using virtual machines, especially when it comes to Win98 builds, it's full childhood blast (not to mention that I put an AWE32 sound card in this PII, which VMware or virtual box can't even emulate properly)

 

[DAPUNISHER]

Well, technically Windows 98 is 2 years younger than me.

Also, of course I don't plug it to the internet directly, all of my retro builds are set up in LAN, from where I can just locally transfer the files from one machine to the other.
Some of the oldest ones I got them in a parallel port local network, so I'm not even using Ethernet (I found it extremely hard to find Ethernet adapters for 286 machines and the ISA cards I have are just 56k modems which aren't good for a local network, so I just went full parallel network on them)

This specific machine, I have it in local network with my Pentium III, so I can transfer files directly to it from my most stable Win98 build.

Btw it's extremely hard, if not impossible, to find viruses for 98 these days, as hackers are mostly focused on XP and modern systems (proof is my PIII which I connected to internet at some point to test Quake 3's multiplayer on that beast, tho the system is still stable and perfectly healthy)

For me, building retrocomputers is a lot better and fun than using virtual machines, especially when it comes to Win98 builds, it's full childhood blast (not to mention that I put an AWE32 sound card in this PII, which VMware or virtual box can't even emulate properly)

You should checkout Philscomputerlab on youtube if you have not already done so. He does some cool stuff with old gear like yours'. His latest vid was about using a SSD with win 98.

 

[Elia1995]

Yup, that and Druaga1 are exactly where I got the idea to put Windows 98SE on an SSD in the first place

 

[mikeymikec]

You absolutely want to make sure DMA is enabled in the BIOS and OS.

 

[Roland00Address]

Does trim even work (if you have the proper os) with a sata to ide adapter? I do not remember anymore, used to remember these things over 10 years ago when I thought the ASUS EEE pcs were cool but definitely not for everyone.

Gosh the first ASUS EEE models were 2007, I got mine sometime in the first year so we are saying 14 or 15 years ago. TRIM did not happen in windows till 2009 (only Windows 7 and newer), and Linux till the last week of 2008…

Gosh memory lane here.

 

[thigobr]

What you say about file transfers make sense... I am not sure what version of the SCSI protocol your other HD uses but SCSI was always a faster bus than IDE. Your Slot 1 motherboard probably has an UDMA33 IDE interface that limits transfers to 33MB/s (i440BX era chipsets stopped in UDMA33). While SCSI buses from the same era had around 50MB/s or even 80MB/s maximum transfer rates.

So the SSD has way faster random access times (the snappiness you noticed) but continuous transfers rates will be potentially slower than IDE depending on what version of SCSI you're using.

 

[lakedude]

You absolutely want to make sure DMA is enabled in the BIOS and OS.

OMG the memories...

Yeah DMA was a huge improvement from whatever crap there was before (PIO).

Edit: PIO would load a CPU down because it depended on the CPU. DMA relieved the CPU of this duty. That said the old interface would be a terrible bottleneck to a modern SSD but not because of any CPU loading.

Every once in a while DMA would get turned OFF somehow and a system would be intolerably slow. I think DMA ON eventually became the default but in the early days it was OFF by default and needed to be manually turned on.

 

[VirtualLarry]

It depends 100% on the chipset used for the SATA-IDE bridge adapter, but I set one up with a SATA SSD into an older Pentium IV rig, with a mobo with IDE ports, and... I found that the adapter "locked up" under heavy usage, when plugged into an SSD. I didn't try anything fancy, like disabling DMA (I don't think, I don't recall that well).

Anyways, I had to resort to using a SATA HDD, instead, because the SSD was "too strong", and IMHO would "overrun/over-load" the SATA-to-IDE adapter.

 

[Elia1995]

Another interesting thing I noticed with this SATA to IDE adapter, is that if I plug in another SSD in it, the bios would lock up at the boot when it has to scan it, idk why that happens only with that specific drive and not with this one, which is even bigger 🤔

 

[LightningZ71]

SSDs require a lot of modern data streaming and buffering techniques to achieve their high performance.

SCSI was a drastically more capable data protocol than parallel IDE. The controllers were fully bus mastering, had deep buffers, often had a ram cache, and could actively reorder drive commands to make transfers more efficient. Parallel SCSI interfaces we're much quicker and more efficient than IDE. Native SATA controllers added a lot of those features for performance and removed the use of bus sharing in most situations.

Parallel IDE has no ability to handle an SSDs throughput.

 

[dr1337]

Retro these days is coffee lake

You must be joking

You're making your system a target for attack.

Simply being vulnerable isn't the same as being a target. Now maybe if they DMZ their router with the retro build online... And even then, for a retro build, getting a virus would only help increase immersion for the actual experience. Its almost funny to me how much vulnerabilities are fear mongered these but it wasn't long ago that every windows computer implicitly had a virus on it at one point or another.

 

[Muadib]

now I put OS/2 on that drive

Oh wow the memories....

 

[LightningZ71]

You say memories…

I say nightmares…. I used to have to care for an OS/2 warp connect server. Glad I don’t anymore.

 

[Muadib]

You say memories…

I say nightmares…. I used to have to care for an OS/2 warp connect server. Glad I don’t anymore.

Warp was awesome!!! I bust my internet cherry with Warp. It was great!!!

 

[LightningZ71]

I will give it credit, it was a solid Lotus Notes server platform... It's the rest that was questionable.

 

[DrMrLordX]

You must be joking

Simply being vulnerable isn't the same as being a target. Now maybe if they DMZ their router with the retro build online... And even then, for a retro build, getting a virus would only help increase immersion for the actual experience. Its almost funny to me how much vulnerabilities are fear mongered these but it wasn't long ago that every windows computer implicitly had a virus on it at one point or another.

Not sure how long it takes for a WinXP system (for example) to get pwned if plugged naked onto the Internet, but the last time I heard of anyone trying that it was pretty quick. Lots of ports getting probed from out of nowhere. Behind a consumer router, well . . .

(also I know we're more talking Win98/WinME than XP, just throwing out the XP experience as an example. Old Win95/98/ME machines might be too old/obscure to be a significant threat target these days)

 

[Elia1995]

Viruses for 98SE are almost non existent, you really have to look hard for them in order to get your retrorig with 98 infected, and even then, some of those oldschool viruses were amusing, such as the one that scrambled the screen, or spammed nonsense error message boxes.
But, regardless that they're almost impossible to find, nobody would put important data on a retrorig beside just oldschool games to make them run on their native hardware, so there isn't a big problem in going online with those rigs anyway... Lil infection? Quick format and ready to play again

 

[DrMrLordX]

Viruses for 98SE are almost non existent, you really have to look hard for them in order to get your retrorig with 98 infected, and even then, some of those oldschool viruses were amusing, such as the one that scrambled the screen, or spammed nonsense error message boxes.
But, regardless that they're almost impossible to find, nobody would put important data on a retrorig beside just oldschool games to make them run on their native hardware, so there isn't a big problem in going online with those rigs anyway... Lil infection? Quick format and ready to play again

Sounds like a fun experiment. Put up a fresh Win98 install and bypass the router. See how long it takes before it becomes unresponsive or starts downloading payloads.

In 2004, The Register ran this article:

Infected in 20 minutes

Stepping in and taking charge

www.theregister.com

It apparently took ~20 minutes for a naked WinXP machine to be compromised, and that was without actively downloading any malware.

 

[BonzaiDuck]

I've posted my frustrations about the CPU exclusions of Win 11. This thread just makes me depressed . . . I could borrow one of my mother's "evening pills" it's so bad . . .

I know! This is almost like that South Park episode of "Goo-backs from the fu-ooo-ture gonna shtake our jobs!" In the end, they all get naked and pile on each other for a sex-romp so that no children are born for the future, eliminating any immigrants coming back through the Time Portal.

So here's what I say. Let's all resurrect our Sinclair ZX-81 or Timex-Sinclair-1000 computers and use them exclusively. The hackers and malware-programmers probably never thought about that!

 

[Shmee]

Hello, I'm on my quest of building retrocomputers and in a Pentium II 233MHz I put an SSD with a SATA to IDE adapter and installed Windows 98SE on it.
The difference in loading times is noticeable compared to a SCSI HDD I was previously using in that build (now I put OS/2 on that drive), but I also noticed that copying files or even installing games from within the SSD is a lot slower than how it should be, so I was wondering, is it possible that the CPU can affect the writing/reading speeds of the SSD, or the adapter kinda limits it?
Of course an SSD in a retro build isn't the ideal way to go and it's "unfitting", but I like experimenting with hardware, especially making retro-modern Frankensteins like that 😅

You might want to take a look at this thread.

https://www.reddit.com/r/windows98/comments/oxxrha

Looks like some Socket 939 boards will work, which opens you up to Athlon 64 X2 CPUs, and possibly SATA. Apparently VIA preferred over nforce4. Time to go down the rabbit hole...IDK if AHCI would be supported for the SSD, but you may still be able to use it on a SATA port, in IDE mode.

 

[mikeymikec]

nforce4 in my experience had an NCQ bug that meant you had to disable it if you didn't want your system to hang regularly. I can't remember whether a bug I experienced whereby I had to disable SATA 6Gbs with a jumper change on WD drives was with an ALi or Via chipset.

The kind of hackery being suggested in this thread reminds me of the workarounds and knowledge in the Amiga community of the hardware, but one huge advantage they had was that they were talking about a much more limited range of hardware to test on. In this situation you could have a dodgy implementation at the board level, the chipset, the drivers, if you're using some cheap adapter, and each of these potential dodgy limitations could just affect one revision of any of those layers, just one model of board or even multiple bugs. Some hardware designer is told to cut costs wherever they can, so some bit of the specs that doesn't seem very important because say SSDs were barely dreamt of could ruin your day. It's really messy. It's one of those projects that I might be inclined to throw some spare time at it, but I would draw a line at it being a waste of time very fast if things go squirrelly.

IMO it's a much better idea to pour one's energy into getting a working VM or use say dosbox to achieve what you want. Ancient hardware that's compatible with Win9x isn't getting any younger, and would likely have fallen foul of such issues as the capacitor plague, let alone the idea of trying to get an SSD working it. Software solutions aren't going to age half as badly, and chances are you can take what you learnt from one solution to its successor.