There's a post[1] about optimizing DMA performance for a ISA CompactFlash adapter. Port-mapped IO is quoted at around 170KB/s, up to 420 KB/s for DMA transfers of one sector at a time, to 530KB/s on an 8088 with a CF card that supports multi-sector transfers. This is a but faster than the random read speed of the slowest SATA hard drive I have around (a 160GB 5400RPM 2.5" Seagate pulled from a portable USB/Firewire drive). So there are definitely some 16-bit PCs that are fast enough to measure an improvement from Flash storage, but the original 5150 is probably pretty close to that line.
We got away with overclocking our XT's to 7.66Mhz with no problems, just swap in a different crystal oscillator. Then swap the 8088 with an NEC V20 for I think about another 20% boost.
There are certain implications of performance for "SSD". It's possible for removable flash storage to be responsive, but there are a lot of cards out there that can't even beat a hard drive's IOPS. Get it up to four digits and the word SSD becomes viable.
USB 2.0 should be almost 100 times faster than the original IDE interface, so there would effectively be no difference between using an IDE->Flash adapter versus an IDE->SSD adapter. On the other hand, if you wanted to be clever, I bet you could emulate the whole 5150 faster-than-life on the ARM cores on the SSD, and just use the old hardware as a screen and keyboard.
It's not about the interface. I have two USB 2.0 flash drives sitting here where copying a folder of small files will take seconds on one and minutes on the other. They were bought around the same time and have similar performance for bulk data.
If something is labeled "SSD" then you can have confidence it won't have a completely trash controller chip. Otherwise you can get something that, while solid state, performs worse than even an ancient hard drive.
I think you are doing a "brand split" - i.e. the terminology has been co-opted by branding to mean something, almost, what it used to mean.
SSD as a "consumer label" means really fast upgrade to a modern PC, one could argue.
But solid-state-disks are a subject as old as the hills, and the phrase really doesn't necessarily mean performance, at all. It'd be better to describe it as "no-moving-parts storage", I suppose - less "brand theft" going on.
Though it is true that SOME CF cards implement TrueIDE mode, it is optional.
The second part of what you said is complete nonsense. SATA uses very few differential signaling lanes. SD cards use a single ended bus of a variable width. (1, 4 or 8 bits). Literally nothing like each other
I have an Oric-1/Atmos configuration, with a Cumulus peripheral, so I have another architecture example of this kind of things.
The Oric-1/Atmos is a very small, underpowered little machine from 1982, using a 6502 CPU with 48k of RAM and little else. It wasn't a super successful machine, but yet .. it lives on. (8-bit revival is awesome, btw!)
The Cumulus is a PIC-based peripheral which emulates the Atmos' disk systems - which were extremely difficult to get, if you were an Oric user back in the day (I was) - and as a result of modern hackers keeping the 8bit spirit alive, we have had "SSD on Ancient Computer (LOL)" for some years now.
This is a machine with, at most, 48k of RAM. Attached to a machine more than 8x more powerful than itself, which function is simply to emulate a very archaic storage system.
This storage system provides the little 48K CPU running at 1MHZ with 16 GIGABYTES of storage.
The problem is neither that there are bottlenecks, nor limits to storage, any more.
The problem is, as it has always been, what the hell to do with it, and .. well the answer is, keep running the machines, and keep writing software for it:
23 comments
[ 2.9 ms ] story [ 66.2 ms ] threadWas the ISA bus as fast as the CPU/RAM on the 5150?
[1] https://www.lo-tech.co.uk/xt-cfv3-dma-transfer-mode/
If something is labeled "SSD" then you can have confidence it won't have a completely trash controller chip. Otherwise you can get something that, while solid state, performs worse than even an ancient hard drive.
SSD as a "consumer label" means really fast upgrade to a modern PC, one could argue.
But solid-state-disks are a subject as old as the hills, and the phrase really doesn't necessarily mean performance, at all. It'd be better to describe it as "no-moving-parts storage", I suppose - less "brand theft" going on.
The second part of what you said is complete nonsense. SATA uses very few differential signaling lanes. SD cards use a single ended bus of a variable width. (1, 4 or 8 bits). Literally nothing like each other
The Oric-1/Atmos is a very small, underpowered little machine from 1982, using a 6502 CPU with 48k of RAM and little else. It wasn't a super successful machine, but yet .. it lives on. (8-bit revival is awesome, btw!)
The Cumulus is a PIC-based peripheral which emulates the Atmos' disk systems - which were extremely difficult to get, if you were an Oric user back in the day (I was) - and as a result of modern hackers keeping the 8bit spirit alive, we have had "SSD on Ancient Computer (LOL)" for some years now.
This is a machine with, at most, 48k of RAM. Attached to a machine more than 8x more powerful than itself, which function is simply to emulate a very archaic storage system.
This storage system provides the little 48K CPU running at 1MHZ with 16 GIGABYTES of storage.
The problem is neither that there are bottlenecks, nor limits to storage, any more.
The problem is, as it has always been, what the hell to do with it, and .. well the answer is, keep running the machines, and keep writing software for it:
http://oric.org/software/
See also, this:
http://forum.defence-force.org/viewtopic.php?f=20&t=1273
EDIT: The wonderful Cumulus:
https://retromaster.wordpress.com/cumulus/