Performance of the 8088 on PC, PCjr and Tandy 1000 | Hacker News Reader

Performance of the 8088 on PC, PCjr and Tandy 1000

retro.moe

69 pointsWoodenChair37 comments

For context, when these machines were being sold, nobody even thought about performance in these terms. It is only later that we are benchmarking old machines to understand the performance gaps, and if you weren't there, you might think this was relevant at the time.

Generally, in the mid 80s, you needed base performance (8088/8086) or "3 times as fast" performance (80286) when the AT came about. Compaq (now HP) became the real competitor to IBM PC/XT because the end user could understand that they ran at 8 MHz vs the PC 4.77 Mhz. This is as much performance issues that the users would want to know. Better performance without needing to buy an AT. Then the clones came about, and everybody started to put a "turbo" button on front of the clone PC.

My first degree was in finance and accounting, but I fell in love with technology, so I sold PCs out of the biggest gray marketer in Seattle when the PC AT (second real computer for business) was coming out. I have first hand knowledge of the buying behavior because I was on the front lines of computer sales, and made a ton of good money selling $2-5K computers to people walking in off the street.

The advent of the PC revolution was all about apps. I'm not saying that there weren't benchmarks, and maybe you'd find a savvy buyer that subscribed to Byte magazine, but these were far and few between. Tandy simply didn't have Lotus 1-2-3. The PC Junior wasn't a performance issue, but the term "Junior" just killed the thing out of the gate. The investment in a PC was a lot of money, and you weren't going to get a Junior, you were going to get a real PC. (For instance, the keyboard was much more impactful than any performance thoughts.)

Finally, for those not around, storage was an incredible change agent. A lot of people think an SSD in a client platform is pretty cool. However, you cannot image the step up that a hard drive provided versus a floppy drive. The PC Junior and Tandy had no real hard drive path. While IBM had hard drive specific model (XT), the effort to install a hard drive in a PC was pretty small because of the accessibility. An open architecture, which any home builder takes for granted today, was pretty revolutionary at the time, and IBM's willingness to allow plug-ins (where they followed Woz) caused it to cremate the competition with a peripheral market.

> I'm not saying that there weren't benchmarks, and maybe you'd find a savvy buyer that subscribed to Byte magazine, but these were far and few between.

In 1982 Byte magazine had 420,000 subscribers. They might not have been the sort who bought from you and your store.

Let's use 1985-1986 because that's when I was active at the store.

There were about 8M computers sold, and assuming Byte continued to sell at around 400K copies, this mean that there were 5% subscribers to every PC bought, or 1 out of 20. My guess is that sounds approximately correct for my hit rate. There were more pubs, but I bet that there were a lot of overlap for magazine subscriptions. If you subscribed to Byte, you probably had a good hit rate for also being a subscriber to PC Mag, so the spread probably wasn't a ton wider.

What I do remember was that Consumer Reports came out with a recommended PC. They loved the Leading Edge. Consumer Reports was a bit of cult back then and I bet their subscription base was in the millions (where does one find the subscribers for magazines in 1985?). When Consumer Reports suggested something, the word got around.

I remember people coming in asking for LE, but Ballard Computer had the franchise and was back ordered like 3 months.

Now that was power to influence the end buyer.

> There were about 8M computers sold, and assuming Byte continued to sell at around 400K copies, this mean that there were 5% subscribers to every PC bought, or 1 out of 20.

Well, that's deceptive. How many of those computers were purchased by a buyer for a business? (again, a population of buyers you might have had little or no exposure to) The ratio of subscribers to PCs purchased would certainly be higher.

It's not like there's anything hugely wrong with anything you're saying, but I was a kid who read Byte magazine and discussed benchmarking with adults, and I don't think we were all total freaks of nature or something. (that said, a benchmark measuring the execution time of a specific operation definitely represents the far side of wonkiness)

1. The market for PCs were virtually 100% business buyers in 1985. The non-business buyer would buy Apple IIs, which I didn't sell. I had a lot of small business and/or individual businesses. Large businesses like Boeing would obviously go through central procurement unless desperate for a PC , which absolutely would happen from time to time.

2. By the way, my 5% estimate on Byte saturation for my buyers is probably is too high. Refresh cycles were about 3 years back then, and there were roughly 11M PC bought in 1983 and 1984. There was around 8M PCs bought in 1985. So around 19M users of which 8M were buyers in 1985, which is where Byte was selling around 400K subscriptions. Therefore, the saturation rate of Byte magazine to PC users was roughly 400K/19M = 2.1%.

3. I loved your line "but I was a kid who read Byte magazine and discussed benchmarking with adults, and I don't think we were all total freaks of nature or something." Actually a young kid reading Byte magazine to discuss benchmarks with adults was a freak at the time.

I should know because I was one, probably the same as you.

At the time, price was everything. With the latest and greatest PC often $4000 and up, or in the $8000 and up range when the 386 came along, if you could get a Packard Bell for $1000 then you’d take it.

the effort to install a hard drive in a PC was pretty small

Relative to today, it was non-trivial though physical access tended to be easier. http://www.oldskool.org/guides/tvdog/1kfaq.html#II.D.1

Sure it's harder than today, but you have to remember that at the time, almost no computers were "plug it in and go". Many of the 8-bit era home computers up to that point, if they were upgradable at all, might require actual solder for internal upgrades. There were exceptions, like the Apple II and the old S-100 systems, but it wasn't common. Going to the IBM, where everything was cards and cables and plugs, was a big step.

Also, the Tandy 1000 is bad example there because they were notorious for weird hardware incompatibilities and funky proprietary parts.

Ordinarily, I wouldn't have used the Tandy 1000 as an example. But it's one of the computers in the article, and anyway, the complexity of adding hard disks to a T1000 is only linearly higher than the complexity of adding hard disks to a standard PC of the era. Getting compatible controller cards, setting jumpers and configuring IRQ's was always involved when installing a hard drive (or video card or anything else) up until some time after Windows 95 a decade or so later.

I don’t remember it being that hard - I had a Tandy 1000SX and I bought a 3rd party hardrive that was mounted on an extension card so that it took up two slots.

I can still hear the sounds that disk made (sort of like a spray paint can being shook) when it was seeking and the “clunk!” it made when you ran the park command.

Sorry for not being clear. My comment about difficulty was relative to installing a "hard drive" today. Back in those days, reading manuals and modifying configurations was normal and expected.

In the mid 90's I rescued a PS2 8580 with a huge 20MB disk that sounded like a washing machine...$8500 brand new. Loved the Model M keyboard -- though didn't know what it was at the time -- and kept it well into the 00's.

Maybe this is a bit difficult for me to be non-biased because I started working on putting hard drives into the original IBM PC in our gray market shop, and I've been working (or leading engineering teams) that have integrated hard drives for 30 years. However, in my mind, the original IBM PC hard drive wasn't a lot harder than any home builder upgrading his or her PC today.

The original IBM PC had a 63 watt power supply (I hope I'm remembering correct), so you had to take this out and put in a bigger one, but this was simple. Then you open up your hard drive box and put the paddle card in the slot and screwed in the HDD in one of the bays.

You need to know a few CLI commands like fdisk and format c: /s/v, and you had a c: prompt. The problem is that there was no youtube or internet. If you didn't know the few steps, this might as well have been black magic. (Even though most HDD kits did come with some instructions.)

The Quantum guys (a company I later worked for) came up with a novel idea of putting a hard drive on a card, and you could just plug in the card and that was it. This was the peak of simplicity, but carried a premium, so it was a niche product.

By the way, what most people don't realize is that the 3.5" form factor was actually the first portable hard drive. It was designed by Conner for the Compaq portable. Bill Frank was the technology guru at Western Digital, then a silicon maker, that lead the creation of IDE for Conner, the forerunner of SATA.

So it was tricky because of the lack of knowledge, but the actual work was pretty simple.

At the time, I thought it was relatively simple myself. I guess what I am trying to say is that the bar for "simple" was lower back then. Setting DIP's, using needle nose pliers to move jumpers and sorting out IRQ's was as normal as using a timing light to adjust a Buick's points. That there was a market for those Quantum disks (which I had long since forgot until now) existed because of that complexity.

On the other hand, back then building a PC was probably easier because there were far fewer choices...thanks to a combinatorial explosion of CPU sockets, RAM specifications, MB form factors and such today. And anyone building a PC probably knew someone who had built one, if only the computer store owner.

The analogy I'd use is that it was simple back then to the degree Linux is simple today: reasonably logical and well documented. But Arch ain't iOS.

Where I worked (a digital electronics simulation company in the late 1980's), we said that PC's had "user interface MIPs" but not "CPU MIPs".

A VAX 11/780 (size of a refrigerator) had more CPU power, but access was via RS232 link to a VT100 terminal. These could be pretty slow: it wasn't unusual to type a command line and wait up to 10 seconds (or a minute, in really bad times) to get a response. And a graph could easily take a minute to fully display on a VT240 graphics terminal.

A PC on the other hand had "user interface MIPs" and could display a graph instantly. What the PC couldn't do is run our very large, 32-bit required simulation code.

(As far as how easy it was to add a hard drive: at the time, to add a new disk to a sun station, you had to recompile the kernel)

Sounds like you where using an overloaded VAX then the PR!ME 550 (similar tech and a competitor to VAX) we had at my fist job never had such sluggish performance.

And Vt240 was very much noddy graphics high end then was insanely expensive vector systems from HP os Tectronix

Twenty programmers on one vax and actually doing the simulations is CPU intensive. The VT240 is what our customers had (and they are expensive enough all by themselves; a single VT340 in 1991 was $3600.

20 people all running intensive programs well I am not surprised sounds like you needed a bigger vax or more vixen.

Our vax cost $500,000. It's not something you just get more of :-)

My dad brought a Tektronix 4051 home from work when I was a teenager, so that was one of the computers I grew up with back in the 1980s. I didn't realize at the time how expensive those were.

> so I sold PCs out of the biggest gray marketer in Seattle when the PC AT (second real computer for business) was coming out. I have first hand knowledge of the buying behavior because I was on the front lines of computer sales, and made a ton of good money selling $2-5K computers to people walking in off the street.

Which store? I'm wondering if my first PC may have come from you. :) I remember it was a place in Georgetown, a 386 that came with DR DOS.

Computer and Applications in Bellevue on NE 8th, by the old John Dantz theater. I was there from 1985 to 1986, then I went back to the UW to get my EE degree. I had all types of people, and the the occasional Microsofter, as customers.

We sold IBM and Compaq gray market. We tried to go legit with Acorn (GEM) and Multitech, which turned into Acer. However, as I understand it, the owner was selling so much gray IBM offered him a dealership after I left.

I got my EE degree, already had a degree in accounting and finance, and eventually went to work for the IBM PC company, which was very strange considering I started off selling them gray market.

I started before the 386 and the PS-2, so I was probably back in school when you were buying.

When the press said it had a "chiclet" keyboard, the perjorative stuck and doomed the PCjr. (It was probably doomed anyway. Nobody wanted a crippled PC.)

Mostly it stuck because it was appropriate. Not a keyboard anyone would want to use for actual work.

I had a "computer module" for the Intellivision game console, with a similar keyboard. It was bad for that, and I barely used the thing it was so useless.

IBM really ought to have offered a normal keyboard option, with the chiclet as a deeply-cut-price accessory "for the kids to use for games".

Yes, the keyboard was insulting. Ironically, today's keyboards are mostly chiclet keyboards, but they hide it with a normal keyboard facade.

I think the switch action is somewhat better now, though. Probably an improvement in the materials used. Scissor switches might help a little, even though the underlying electrical contact mechanism is similar. At least with scissors the feel isn't purely based on the behavior of rubber.

I find stuff like this fascinating; I'm old enough to remember computers from this era, and—at least for me—I have a certain nostalgia for them. The first computer I was exposed to was an IBM PCjr, so seeing stuff like this 35 years after-the-fact is pretty mind-blowing, that people still care.

I know that a lot of it is just my age, and I remember my father growing up used to work on/rebuild the engine on his '64 VW beetle. I think for very similar reasons: nostalgia, limited complexity, the tangibility of the "API". For me, POKEing a value in memory and seeing the screen change or a sound play is tangible. The opcodes of an 8088 are _just_ enough that I can feel like I understand it. Modern computers feel several orders of magnitude more complex and—admittedly—more powerful, but at a _cost_. I'm sure my dad thought the same about electric fuel-injected engines and the ever increasing complexity of "modern" cars.

Currently, on my work-bench in the garage is a half-completed [Micro8088](https://github.com/skiselev/micro_8088) and 8-bit ISA back-plane; still waiting for a few parts to come in. Its been a lot of fun to work on, and the intent is to get back into doing some retro-programming. (Hopefully, I have a better understanding of the underlying architecture now than I did when I spent a summer when I was 14 trying to figure out the relationship between AH/AL and AX without a concrete understanding of binary/hexadecimal math)

The lower 128k on the PCjr is shared with the video controller. Contention caused by this dual access is what slows it down.

I expected the Tandy 1000 to have similar performance - IIRC, their video modes are very similar.

Most of the Tandy 1000s included a DMA controller which the PCjr did not. This eliminated the shared memory issues the PCjr suffered from.

Although similar Tandy graphics could utilize different memory blocks for video. This means that the Tandy was compatible with software written for the PCjr graphics system but some software written for the Tandy graphics system is not compatible with the PCjr.

The DMA issue was only a factor for the disk drive. While the PCjr graphics memory was shared with the main CPU memory, the Tandy graphics memory was separate in the same way that the EGA graphics memory was separate---it was never used for program code.

In fact, the Tandy graphics was probably more compatible with the normal PC graphics than with the PCjr graphics. Yes, you could still access CGA graphics on the PCjr via $B800:0000 but the graphics chip rerouted the accesses to low memory were it was actually stored.

But to use the 320x200x16 graphics mode (PCjr and Tandy supported this, CGA did not) on the PCjr, you needed to not only set the mode, but know the physical address (which could be anywhere in the lower 128K of RAM). When I wrote programs that used this mode, I had to use a custom MS-DOS configuration that set aside a block of low RAM for this mode. On this Tandy, you could avoid this and use the RAM set aside for graphics (starting at $A000:0000 if I recall correctly).

"the Tandy graphics memory was separate in the same way that the EGA graphics memory was separate---it was never used for program code."

This is incorrect. The Tandy graphics adapter did share ram with the cpu. The service manual for the Tandy 1000 can be found at http://www.thealmightyguru.com/Wiki/images/7/7c/Tandy_1000_-.... Starting on page 40 is a good description of how this worked.

I know on our PCjr, the first 128k was built in, and above that only worked when we had the memory expander card attached, so there is a location difference. The cynic in me says it was an administrative decision to show the jr as a lower cost lower performance option.

I had a base PCjr with 128K of RAM, one 360K floppy disk drive, and no hard drive. Copying disks took FOREVER. I was so jealous of my friend who had two disk drives on his PC.

I remember the PCjr felt considerably slower than the 5150 (PC) and also had a terrible keyboard, however it was innovative to use infrared to make it wireless. It also had cartridge slots and a small catalog of (mostly) games that came on them, but in practice there was no benefit to that.

The biggest problem I saw with Tandy was partial incompatibility with certain popular software.

"but in practice there was no benefit to that."

Apparently the cartridge ROMs weren't afflicted by the wait states that applied to the first 128k of RAM, so software on the cartridges could actually run faster than software on disk.

On these machines performance was very consistent. They has memory without any cache, with memory parts with standard timings connected to a memory bus with the same timing as the CPU clock. Differences were in additional wait states introduced by other parts (such as the video on the PCjr) of the machine-specific architecture.

It'd be nice to run a similar set of benchmarks on "almost compatible" machines such as the Sanyo MBC-555, the Zenith Z-100 or the Mindset.

Having owned all three of these; I grew up in an IBM house hold but later on I had two different Tandy 1000 series PCs (EX and the bastardized TX which had an 80286 on a XT board). We had a PC from early days complete with the large hard bound manuals, single sided diskettes, and text only.

Now the regardless of processor the PCjr was seriously impacted in that that the keyboard interrupt was not there but actually had to be handled by the processor. See https://www.brutman.com/PCjr/pcjr_keyboard_handling.html for a good description of what this incurred.

(now for the rambling part) The Tandy was great for their 16 color hack and better sound which made games so much more enjoyable. (star flight being my favorite game ever and it used two 360k diskettes for a persistent environment)

fwiw, even before dedicated graphics we had cool games. Even Wizardry which I had on my Apple was available for the PC and I played it with the ASCII character set being used to create the halls in 3D (think plus sign, bars, and minus signs). All the text based games many reminisce about are from this era

Reply on news.ycombinator.com