Jump to content

68030 @ 16MHz vs. 68000 @ 16Mhz: how much faster?


Recommended Posts

Just what the title says, I'm wondering about the performance differential between 68030 and 68000 running at the same 16MHz clock. Any notion or point in the right direction to find out? My hunch is that it would be significantly faster, if not amazingly faster which I doubt.

Link to post
Share on other sites

Difficult to estimate.

 

68000 to 68020 is about 1.6 more IPC (benchmarks done by Motorola on CPU32 vs 68000. CPU32 is a basically 68020 core with only 68010 instructions and a 68000 like bus (16 bit wide)).

68020 to 68030 is mostly dependant on

- how the software make use of the CPU cache

- how many new instruction it uses (memory access, compute)

 

Out of my A** I guess 2.0 more IPC

 

 

Edited by demik
Link to post
Share on other sites
  • 68kMLA Supporter

Maybe compare a Macintosh Classic II (68030 16Mhz) to a Macintosh Portable (68000 16Mhz) via old benchmarks:

 

https://lowendmac.com/2000/mac-classic-ii-benchmarks/

cache    CPU  graphics disk   math
 32 KB   4.00   3.49   2.30   5.60
 64 KB   4.00   3.45   2.30   5.43
128 KB   4.01   3.49   2.26   5.54
256 KB   4.01   3.49   1.91   5.42

vs.

 

https://lowendmac.com/1998/mac-portable-benchmarks/

cache   CPU  graphics disk   math
 32 KB  2.15   1.82   1.25   1.95
 64 KB  2.15   1.82   1.23   1.95
128 KB  2.15   1.82   1.24   1.95
256 KB  2.15   1.82   1.24   1.94

 

CPU almost twice as fast on the '030.  There are a bunch of other old Macs you could look at the Speedometer 3.06 results for on that site for caparison.

 

 

Link to post
Share on other sites

All these factors contribute:

  • Bus width. 16 bits versus 32 bits, except for LC / LC II and similar models
  • Memory clocking. m68000 takes four cycles. '020 and '030 take three.
  • ALU. m68000 has 16 bit ALU and '020 / '030 have 32 bit ALU
  • Instruction cache of 256 bytes on the '020 and '030. m68000 has none.
  • Data cache of 256 bytes on the '030. Both caches make a big difference, particularly on a 16 bit data bus.
  • Instruction timing. Lots of instructions take less clocks on '020 and '030 than m68000.

I'd generally expect a 16 MHz m68020 or m68030 to be generally twice as fast or more than a 16 MHz m68000.

Link to post
Share on other sites
4 hours ago, Fizzbinn said:

Maybe compare a Macintosh Classic II (68030 16Mhz) to a Macintosh Portable (68000 16Mhz) via old benchmarks:

It's a good idea to compare to 16-bit buses such as the LC II, Color Classic, and Classic II. This helps temper expectations but also illustrates the vastly increased performance of the '030 core over the '000, which is significant.

 

The 68000, though a pretty well obsolete core by '89, was used in the Portable and PB 100 in CMOS form (as the 68HC000, probably for "High Cost") because it was extremely power efficient and, in at least the PB 100, an extremely tiny chip. I don't believe the '030 was ever made in a CMOS process except perhaps as an EC variant (no MMU), which should be fine for Portable use but would preclude your use of VM. If you plan to build a Portable accelerator with a full '030 your battery life will likely be noticeably impacted.

Link to post
Share on other sites
27 minutes ago, Franklinstein said:

I don't believe the '030 was ever made in a CMOS process except perhaps as an EC variant (no MMU)

According to the datasheet the '030 was fabricated in the same HCMOS process as the 68HC000. Apparently the 68020 is as well. The difference in max power draw between the 68HC000 and the 68030 is still about an order of magnitude; the datasheet says it's .26 watt at 16.67mhz for the 'HC000 while the '030 manual says it can dissipate up to 2.6 watts, doesn't say what speed that applies to; presumably 40mhz, that's the highest rating in the datasheet I'm looking at.

 

Interestingly enough this *does* mean that an '030 doesn't draw a lot more power than an original HMOS 68000.

Link to post
Share on other sites

That's probably what the "HC" in 68HC000 stands for, is HCMOS. High Cost is probably also accurate though. They seem to have split the datasheets for the '030 and 'EC030, which is probably the source of my inaccuracy. After further reading of the apparently EC-specific data sheet, it appears the EC variant dissipates the same 2.6W at 40MHz, so there's likely not much to be gained there by using it rather than a full '030.

Link to post
Share on other sites
7 hours ago, Gorgonops said:
7 hours ago, Trash80toHP_Mini said:

Portable is the target machine, so the bus remains 8bit.

The Portable has a 16 bit wide bus.

DuOh! My bad there. Can't forget the 4-5 GALs in terms of power budget. Haven't looked into compatibility yet, but Freescale is still turning out 68030 for embedded applications. Hoping to end up with a very modern process 68030 and the GALs lumped into a single more modern component that takes less power. Could have voltage problems, dunno. The embedded procs aren't same/full featured, but there's no need of MMU in this application.

 

Very early research into possibilities here. But Performer looks like the best approach to getting something faster into the Luggable. I've still not finished buzzing the board out. Started that a couple of very rough years ago, ready to look further into it finally.

 

Original Thread six pages of jibberish, If some research comes together I'll post a new one and get going again.

Link to post
Share on other sites
3 hours ago, Franklinstein said:

High Cost is probably also accurate though

Did it actually cost more? The fact that it appeared in a singularly expensive machine isn't really evidence one way or the other, plenty of ridiculously expensive computers had maybe a $100 bucks worth of silicon inside. "High Speed" CMOS was invented in the late 1970's and started spreading in the early 80's, when the Mac Portable came out there was nothing special about it.

(The main reason for the Portable being *actually* more expensive wasn't CMOS, it was that it used static RAM. Static RAM requires 4-6x the transistors than the same quantity of dynamic RAM. Oddly enough that *still* does save power because it only really consumes power when the contents of those transistors is changing instead of constantly having to cycle.)

 

The 68000 itself was a pretty darn expensive chip when it came out, it's one of the reasons why IBM didn't plunk it into the 5150 and we don't live in whatever alternative universe that would have spawned.

Link to post
Share on other sites

If HCMOS existed in the mid-70s and there was no price difference, why wouldn't they have used it to begin with? Some brief internet sleuthing suggests the original HMOS 68000 was $487 when introduced in 1979, while the HCMOS 68HC000 version was about $800 at introduction in 1985.

 

I'm aware of the Portable's development history. Close to half of the original M5120's $6500 price was down to the Gassee-demanded active-matrix LCD panel. The SRAMs were the next most expensive components, which is why they were dropped for PSRAM in the M5126.

Link to post
Share on other sites
30 minutes ago, Franklinstein said:

If HCMOS existed in the mid-70s and there was no price difference, why wouldn't they have used it to begin with?

I didn’t say it existed in the “mid-70’s”, I said it was invented in the late ‘70s and started becoming mainstream by the early 1980s. (Replacing the older, very slow original CMOS process used in the 4000 series parts.) CMOS versions of CPUs originally released in HMOS were pretty readily available by 1983 or so. (8085, 65c02, etc.) The 80286 was the last NMOS new design Intel CPU, the 386 was also HCMOS and that was in prototype in 1985. Considering the portable didn’t come out until *four years after* your 1985 figure what Motorola thought it was worth when it was brand spanking new is kind of irrelevant.

 

(Motorola also has a long history of price gouging. The whole reason the Apple I was built around a 6502 instead of the 6800 is MOS charged 1/10th as much for it.)

 

TL;DR, HCMOS was old news by 1989. I have a 1986 Tandy computer sitting on my desk that I’m intimately familiar with the guts of, and it is largely made of HCT, which is HC with special sauce added to make it directly mixable with TTL. If HC was still impossibly expensive in 1986 you could lay everything you own on the bet that Tandy would have bought LS instead.

Link to post
Share on other sites
59 minutes ago, Franklinstein said:

Some brief internet sleuthing suggests the original HMOS 68000 was $487 when introduced in 1979, while the HCMOS 68HC000 version was about $800 at introduction in 1985.


Also of note: inflation was Kray-Kray in the late 1970s and early 1980’s. According to the inflation calculator that spread is only about 15% more than the general inflation rate between those two dates. Not saying Motorola wasn’t kidding themselves with that price, but there is that.

Link to post
Share on other sites

I didn't get a ton of semiconductor sales sheets from the 1980s turning up in my 10-minute search, so I don't have accurate figures from launch onward, just a few articles that stated "the 68HC000 will be roughly twice the price of its NMOS counterpart" at introduction in 1985. I'm sure the prices went down with the march of progress and the introduction of successor chips, but even if the difference shrank to only $5 more, it's still a higher cost than the original variant. 

Link to post
Share on other sites

Here’s an announcement from 1988 for availability of new 16mhz versions of both the regular and CMOS 68000s.

 

https://archive.org/details/computer-magazine-1988-05/page/n99/mode/2up/search/68hc000?q=68hc000

 

and, yep, the CMOS one is about exactly twice as expensive... but the difference is $18 vs. $34, quantity 100. So, yes, it’s “high cost” but by no means of a magnitude that meaningfully affects the price of a $6000 laptop.

 

As for why it cost any more at all, it’s probably about a 50-50 mix of “because they can” and there being some capacity premium on foundry space for making high-density newer process chips.

 

... edit: also, don’t forget, the HC000 is a fully static design and therefore actually has more transistors than the NMOS version. So you are actually getting more for your more money.

Edited by Gorgonops
Link to post
Share on other sites
2 hours ago, Gorgonops said:

... edit: also, don’t forget, the HC000 is a fully static design and therefore actually has more transistors than the NMOS version. So you are actually getting more for your more money.

Actually, scratch the fully static thing, I double-checked and the original 68HC000 still had a minimum clock speed. It wasn't until the SEC version that you could completely stop the clock on a 68000.

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...