• Please review our updated Terms and Rules here

10MHz battle: 286 vs NEC V30

twolazy said:
You can use a zip100 on a 8086 no problem... Just cant use the iomega drivers, have to use palmzip.

Click to expand...
Yes, thanks. The "and such" was to include other software like network drivers, Windows 95 version of EDIT.COM (my go to light weight editor), etc. The point being there are additional reasons, like ChuckG mentioned, to go with a V-series CPU other than just a (small) performance boost.
 
Yep. It's a sad fact that a not-insignificant body of 1990's vintage (and up) DOS software is compiled to use the "enhanced" real-mode x86 ISA introduced with the 186/286 even if it doesn't otherwise need a post-8088/8086 CPU. Having a V20/V30 solves a real problem if you're not just using your XT-class machine to run 80's game software.

Ironically I never really noticed this problem back in the actual 1990's because the only thing I was "using" XTs for back then was either spare parts or slapping them together for poor people to use as word processors. (And for that use case they'd end up with a pirated copy of PFS:Write or something.) This is pretty strictly the kind of issue that comes up when you're trying to push these things further than they were ever really meant to be pushed. Which is half the fun, right?
 
When I bought my first XT clone as a poor college student in 198(cough), I purposely ordered it with a V-20 for the CP/M code compatibility. One of my microprocessor labs used the 8080. Using the CP/M version of ASM, I could write and test my code right on my XT. When arriving at the lab, my code ran correctly the first time, every time. And significantly smaller than everyone else's. Thanks NEC V-20 and whatever emulation code I was using (22NICE?).
 
Probably. I recall working with a guy at NEC Natick on certain issues with the V20. I may even still have a MicroNOTE or two somewhere in my files. Most 8080 code will run, but oddly, not JRT Pascal--there was a bug in the emulation. As I recall, the JRT calling sequence set SP very close to the entry point of a routine; the instruction prefetch in the V20 messed things up.
 
=
Chuck(G) said:
If you want a close-technology match, an 8086-to-V30 would seem to be a better match. Matching a V30 with an 80286 is not an even race--completely different generations.
Click to expand...

resman said:
The point being there are additional reasons, like ChuckG mentioned, to go with a V-series CPU other than just a (small) performance boost.
Click to expand...

I think jasa1063's real goal here was more to compare a "10 mhz XT" to "10 mhz XT" where the CPU would have been the major selling point: "Hey, don't buy that one, buy the 286. It's better!" While the V30 is an 8086 clone, here it's running at the "same speed" as the 286. So, while the two CPU's are different generations, the machines are the same class (XT) and roughly equivalent in their other components. So I think the idea is to see how a V30 XT matches up to a 286 XT. Am I right jasa1063?

resman said:
You can see a difference in benchmarks, but doesn't *feel* particularly faster.
Click to expand...

This is an interesting point for this "10mhz battle" benchmarking. Do jasa1063's two machines *feel* about the same or does one feel faster?

jasa1063 said:
If anyone would like additional benchmarks, I will be glad to run them. Please let me know.
Click to expand...

How about running the "human benchmark" on the 2 machines. What are their jasa1063 ratings?
 
Last edited:
acgs said:
=



I think jasa1063's real goal here was more to compare a "10 mhz XT" to "10 mhz XT" where the CPU would have been the major selling point: "Hey, don't buy that one, buy the 286. It's better!" While the V30 is an 8086 clone, here it's running at the "same speed" as the 286. So, while the two CPU's are different generations, the machines are the same class (XT) and roughly equivalent in their other components. So I think the idea is to see how a V30 XT matches up to a 286 XT. Am I right jasa1063?



This is an interesting point for this "10mhz battle" benchmarking. Do jasa1063's two machines *feel* about the same or does one feel faster?



How about running the "human benchmark" on the 2 machines. What are their jasa1063 ratings?
Click to expand...
It was my point to compare two 10MHz XT type computers that were each at the top end of the XT class to see how they would stack up against each other. I knew the 286 was going to win, but the question was by how much. As to the feel between the two computers, the Tandy 1000 TL/3 definitely feels faster in just about every respect. The human benchmark goes with the Tandy 1000 TL/3 in my case.
 
seaken said:
I've never heard of the 286 referred as an "XT Class" before. I thought 286 was "AT Class". No?

Seaken
Click to expand...
The Tandy 1000 TX/TL/TL2/TL3 286 computers were all XT Class computers because:

- They only supported 640K for DOS and no extended memory
- The BIOS did contain the ABIOS like the IBM AT to support protected mode
- The ISA expansion slots were 8-bit only

You can think of these computers like an IBM PC/XT with 286 accelerator card in them. No advanced AT like features just a faster CPU.
 
Last edited:
jasa1063 said:
I currently have my Tandy 1000 TL/3 and Epson Equity Ie computers right next to each[ other]...
Click to expand...
I'm interested in details of the Equity 1e such as a BIOS dump and doing comparisons to a Model 30 (I've got a V30-8MHz in my main unit) - sending my information in a DM.
 
The Tandy 1000 286s also have no high IRQs, no high DMAs, an XT keyboard interface, and an 8-bit IDE-XT hard drive interface in the TL/2, TL/3, and RLX.
 
Chuck(G) said:
Has anyone experimented with putting the 1000 CPU into protected mode? I assume that the thing is limited to 640K?
Click to expand...

I think the hard part would be getting it out of protected mode. So far as I know its missing all the hardware and bios moving parts to recover quickly/non-destructively from a reset, so it’d be a one way trip.
 
Eudimorphodon said:
I think the hard part would be getting it out of protected mode. So far as I know its missing all the hardware and bios moving parts to recover quickly/non-destructively from a reset, so it’d be a one way trip.
Click to expand...
Sounds like it could be a fun machine to play with Minix 2.x on, although for unsupported hard drives and potentially other parts I believe it makes BIOS calls (plus the protected mode kernel might need more than 640k anyway)
 
vwestlife said:
The Tandy 1000 286s also have no high IRQs, no high DMAs, an XT keyboard interface, and an 8-bit IDE-XT hard drive interface in the TL/2, TL/3, and RLX.
Click to expand...
Thanks for adding those clarifications on why it was an XT class computer. I completely missed including that information.
 
The Tandy 1000s as well as a few other systems that used the 80286 in an XT mode illustrates how Intel really dropped the ball back then. It took Intel a non-insignificant time to get the Xenix kernel running (they were doing the work for SCO); initial releases were a bit buggy, but the bottom line was that if you were a computer vendor and wanted something Intel faster than an 8086, you had only two real choices: the 80186, which had compatibility issues (integrated peripherals) or the 80286 run in real mode.

The latter was attractive because you wound up with a PC that was 8086-compatible and could use all of the 8086 support chips and you didn't rely on Intel getting the protected mode bugs worked out.
 
For fun, I did a quick real-world test comparing my Deskpro 8086 with NEC V-30 to my Compaq Portable with an Orchid TinyTurbo 286, as they sit right next to each other. You can see the Top Bench scores earlier in this thread. Both computers are equipped with an XT-IDE connected to a DOM. The TinyTurbo 286 runs at the same clock speed as the V-30: 7.16 MHz. The 286 is handicapped by having to interface to a slower 8 bit memory bus, but it gets a cache.

My test consists of compiling and linking one of my LORES test programs using MSC 5.1, since that's what I have set up at the moment. Timing is based on my reaction time using a stopwatch, and you can guess how amazingly accurate that is. Took the average of a few runs, there was about a 2/10 second range due to a few factors, not the least was my reaction time ;-)

But the final result was that both machines were just over 31.5 seconds, on average. Before you say it was an all I/O bound test, MSC 5.1 spends a lot of time doing who-knows-what without I/O. I did tests with Borland C++ 2.0 which was almost 10 seconds faster - definitely more I/O bound but with almost identical results.

And the gut-feel test says these machines are pretty close as well.
 
FWIW, the 286 Tandy 1000s, despite being “XT class” machines architecturally-speaking, are likely faster than most XT 286 accelerator cards because at least on the local bus between CPU and RAM they’re fully “native”. The 1000 TX, for instance, uses a C&T 82A205 to handle bus transactions between the “286 zone” and the 8-bit Tandy 1000 parts. This is a chip borrowed from a regular AT chipset.
 
With the topic of the Epson Equity Ie coming up with this post, I have finally started looked at the 2nd one I got awhile back. It has the original box and owner's manual. I may be looking at parting with that one since I already have a fully functional one. If someone is interested in taking it off of my hands, send me a PM.
 
The 1000RLX is closer to a "real" 286, not only because it includes VGA and a high-density floppy drive, but also because it supports 1 MB of RAM (640K base + 384K extended), can load HIMEM.SYS (with the /MACHINE:2 switch), and can run Windows 3.1 in standard mode (slowly).
 
You must log in or register to reply here.
Back
Top