My First Quadra

[EvilCapitalist]

And really, I think all of this is what makes the 630, 6200, and 5200 so perfectly emblematic of belaguered, directionless '90s Apple. There were too many products and no good way to know which one you should get, because any of them would work for almost anything, and many of them were literally identical hardware, but at different price points (650/800, 475/605) or extremely similar configurations with different price points (475/575/605/610/660). In the midst of making the push for the new PowerPC platform, Apple didn't bother to discontinue a few '030s still on sale (CC, LCIII, 520, 550, TV, Duo230, PB145B PB150 which was introduced after the 6100) and couldn't even be bothered to stop introducing new '040s (630, Duo280, PB500 series, PB190 in 1995)

It seems like Intel may be falling into the same trap that Apple did during that time period.  We now have the following Intel processors (roughly from low to high end):

Atom

Celeron

Pentium

Core m3

Core m5

Core m7

Core i3

Core i5

Core i7

Core i9

Xeon e3

Xeon e5 / Xeon Gold

Xeon e7 / Xeon Platinum

Back in the days of the clock speed wars, it was pretty easy to tell which system was faster:

- 386DX/33 > 386DX/25

- 486DX2/66 > 486DX/25

- Pentium MMX @ 200MHz > Pentium 133

- Pentium II/400 > Pentium II/333

Today, with Intel's crazy number of different options you have Pentiums that can be on par with i3s, i5s that can outclass i7s, m7s that are slower than i3s but can sometimes be faster, and people who are used to the old way of "higher number is better" getting confused that an m7 can be slower than an i3/i5. 

Couple that with the fact that most new processor generations have only had marginal improvements, the old generations are still on sale for years after they're introduced (Micro Center only recently stopped selling socket 1150 i5/i7 Haswell processors and they're still selling Pentium/Xeon ones), and some of the previous generation can outclass the new generation and it gets even more confusing. 

 

[Themk]

Couple that with the fact that most new processor generations have only had marginal improvements, the old generations are still on sale for years after they're introduced (Micro Center only recently stopped selling socket 1150 i5/i7 Haswell processors and they're still selling Pentium/Xeon ones), and some of the previous generation can outclass the new generation and it gets even more confusing.

Indeed. I bought an Intel Core i5 3570K, way back in 2012. Now, five years later, it is still powering my main workstation. After mildly overclocking, I have performance comparable to the i5-6600K, and likely the 7600K too. The most improvements are in the platform IMO. I have no pressing need to replace my computer. How many years longer I will keep it, I don't know. Five years is a good service life, but, I still think it has plenty more. Maybe I will keep it another 5 years ??. Likely not, but I could realistically see it carrying me forward 2-3 more years probably.

Since this is the 'My First Quadra' thread, for fun I decided to open up the intel ark page linked earlier in my post on the machine [] ]'>.

 

[Themk]

Computer prices dropped like a rock from 1990 to 2000 so having to many price points would have been a major sales problem (and inventory problem). Having a Centris and Quadra sticker on the same machines was stupid.

During the 68k era they pretty much just released whole new machine designs every few years, unlike the G4 era where they kept the hardware kind of the same with tweeks and just had multiple processor updates.

Why would you build a 630 and also a Quadra 800 and 650? The 630 took away sales from the probably more expensive 650 and made the 800 look too expensive.

I agree that the 90s were certinally a decade of change in the computer market. I know 2000-2010 were also a pretty big decade, but I have to argue that the 90s probably changed more between 1990-2000.

Why would you build a 630 and a 650/800? 90s Apple being 90s Apple. Look at the timeline of mac models listed in a thread in the Lounge, see the 1994-1995 timeline, up until New World ROM. It's cluttered. I think though that the 630 has some merit, due to the fact that it has built-in CDROM (I know that CD-ROM was optional, as well as available on the 650/800 too), and AV functionality, while sacrificing NuBus expansion, at a lower price point. The case of the 630 as jt mentioned earlier is a nice upgrade from the 800-series case, in the sense that it supports quite a bit of weight, and is easy to work with. The motherboard literally slides right out. Upgrading the HDD is also painless too. The 650 case wasn't too bad though. I still don't understand how the 650 and 800 were sold simultaneously, as they have extremely similar setups, including the same motherboard. (I know it's slightly different between the 650 and 800, but let's face it, it's pretty much identical).

 

[Gorgonops]

So far as I know the motherboard in the 650 and 800 are literally identical, IE, the same replacement part works on either. When you slide the board into the case it interfaces to a little plug in the front and the wiring of that plug changes the Gestalt ID to match the case, but whether that changes *anything* about how the board performs, well, heck if I know. (I vaguely recall reading some article that claimed that when it was in "800 mode" it used slightly more aggressive memory timings to make it infinitesimally faster, but I don't know of any benchmarks that back that up.) Really the only advantage the 800 had over the 650 was the tower case let you fit *slightly* more stuff into it. (A standard configuration of both machines with a CD-ROM drive leaves the 650 maxxed out, while the 800 has *one* open 5.25" drive bay.)

Given it was standard fare in the PC world to sell machines with the same motherboards in both tower and more compact desktop (or "mini-tower") configurations the idea of having two systems with the same motherboard like the 650/800 isn't completely ridiculous, but it does seem a bit odd that the capability spread between them was just so small.

 

[Themk]

When you slide the board into the case it interfaces to a little plug in the front and the wiring of that plug changes the Gestalt ID to match the case, but whether that changes *anything* about how the board performs, well, heck if I know.

Ahh, that's what I was thinking of. I thought some sort of connector was slightly different. I thought they were identical otherwise. Thanks!

Given it was standard fare in the PC world to sell machines with the same motherboards in both tower and more compact desktop (or "mini-tower") configurations the idea of having two systems with the same motherboard like the 650/800 isn't completely ridiculous, but it does seem a bit odd that the capability spread between them was just so small.

You are right, that is standard, and still is standard to a certain degree (i.e. Dell Optiplex systems are available in both Desktop and Tower forms. I *think* the motherboards are different, but as far as standard configuration options goes, they are nearly identical. They might offer a higher-end video card in the tower model, but that's about the only difference I can think of. The most obvious difference is that the 'desktop/mini-tower' has low-profile expansion slots.)  I just still have a hard time figuring out why identical systems were sold at different price-points. I would rather have the 650, I like the IIvx style case better than the 800 case, but, to each their own. But, if you look at the timeline diagram, Apple was swimming in different models, so, why not have two identical computers in different cases and sold as different models!

 

[Gorgonops]

It is a little interesting to note that Everymac.com quotes the RAM speed of the Q650 as 80ns while the 800 is spec'ed at 60ns, so maybe there's something to that RAM timing thing. I believe the pinout of that case plug is floating around, it might be an interesting experiment to make up a wiring loom to let you switch a 650 (or an 800) between different gestalt ID's and see if it actually does run slightly faster as an 800.

 

[Cory5412]

I believe the 800 may support one or two more drives than the 650(1). I don't know or have any references for Quadra 800 pricing handy. I know that the 650 was pretty expensive (at least in 1993 when it was still quite new) compared to everything below it in the lineup.

Given that they were literally the same board, it may have made more sense to sell them as the same system but with different pricing, perhaps call it Quadra 650DT and Quadra 650MT, or something like that.

Then again, all these decisions really play into what a product line is like.

Intel's silicon isn't particularly confusing, but Intel does have a tendency to do bad things to naming.

Core m3/m5/m7 processors are a great example of this, primarily because in the Skylake generation, they were renamed to make more sense. Here's how the chips are named over three generations:

• Haswell: Core i5-4200Y
  
• Skylake: Core m5-6Y57
  
• Kaby Lake: Core i5-7Y54
  

I think it gets down to knowing how to read the model numbers. i5-7300U and i5-7300HQ and i5-7300(2) are all going to be pretty different CPUs. Whether or not it's up to Intel to put that suffix up front to differentiate performance/class families, like they did for a while with the m3/m5/m7 (oddly, m3 is still separate) is kind of a personal style preference.

Apple's problem, in part, was that there was no real sensical way to figure out what was what. The 4400, 5400, 5500, 6360, 6400, and 6500 all had pretty similar performance. But, there's weird violations to "bigger numbers are better" all over. Like, the 4400 technically has more expansion than the 6360, and is faster than it. The 7300 is faster than all of them -- except most 6500/5500 configs, which ship with faster processors.

And, at the top end, the numbers signify performance, sort of, but they seem to signifify chassis more, and that's relatively stable within the lineup, except for the transition from the 6200/6300, 6360, and them boom, 6400's a tower. Or, there's also the whole thing about how the 6100 was in effect replaced by the 7200 and the 6200 was a successor more to the 630's performa/lc models than to any Quadra or Power Macintosh.

The whole thing just feels like Apple wasn't thinking very hard about what people would think about its products or what would lead people to a particular conclusion about what to buy.

(1) Regarding RAM speed, I wonder if that's a matter of what RAM Apple put in from the factory. It would do well as a fact to set the tone for the fact that the 8100/100 and 8100/110 existed but the 6100 and 7100 only ever went up to 66 and 80MHz respectively, despite most high-frequency PowerPC chips not being a whole lot more difficult to cool than the low-watt ones. That (and my own soreness over there not being a 7300/350) is a different issue however.

(2) the 7300U and 7300HQ are real processors but the 7300 is not. at the desktop level, i5 starts with 7400T and 7400.

 

[Gorgonops]

(1) Regarding RAM speed, I wonder if that's a matter of what RAM Apple put in from the factory.

Maybe I'm forgetting something, but so far as I know 72 pin RAM doesn't have any mechanism to signal its speed to a motherboard. So in order to actually benefit from from a higher RAM speed you'd need *some* sort of external ability to signal the system to program its memory controller more aggressively. Presumably that would be related to the Gestalt ID selector, given there's no other DIP switches, etc. anywhere.

To really suss out a difference in RAM timing between the two systems (and between different model Quadras in general, given how many ran at the same CPU clocks) you may well have to resort to a "real world" RAM-intensive benchmark that takes a significant amount of time. Many benchmark programs suffer from the problem of the benchmarks essentially fitting entirely into the CPU cache. Assuming that applies to programs like Speedometer that could explain why the CPU scores of all these different models are so consistent despite the use of interleaving, etc.

 

[Gorgonops]

Maybe I'm forgetting something, but so far as I know 72 pin RAM doesn't have any mechanism to signal its speed to a motherboard...

Actually, I stand corrected; it looks like the "Presence Detect" mechanism would allow it. So, scratch the Gestalt thing having any effect, most likely.

 

[Themk]

Indeed so. I should have mentioned that earlier, but, for anyone interested:

http://ohlandl.ipv7.net/config/mempresence.html

http://ps-2.kev009.com/ohlandl/config/mempresence.html#72_pin_SIMM_Identification

Now, what the gestalt *could* affect, however, is how fast the memory controller will run. Someone here with a Wombat board will need to test that idea.

It seems silly though to intentionally cripple the speed of the memory controller with 650 gestalt.

 

[CC_333]

It seems silly though to intentionally cripple the speed of the memory controller with 650 gestalt.

At the time, Apple was really into crippling lower-end models which were otherwise identical to the higher end ones so they wouldn't compete, so in that context, this makes total sense to me.

c

 

[Gorgonops]

If someone actually has both a 650 and an 800 handy it might be interesting to look at the markings on the installed RAM (and perhaps more saliently, the markings on the base RAM soldered to the motherboard, since there's a good chance any system owned by someone here will have RAM that was added after purchase, not by Apple) and see if there actually is any difference.  According to the technote, which dates to when the 650 was a "Centris" model and ran at 25mhz, that 80 vs. 60ns thing is "official", but I can't help but wonder if it remained valid after the system became a "Quadra" and was bumped up to the same speed as the 800.

Since the boards *do* have RAM soldered to them it would follow that there would have to be two versions if you were actually imposing a difference; I gather there are several different part numbers on "Wombat" boards; those *could* just be revision numbers, but if you have access to enough different Wombats (preferably including some Centris-label ones) it shouldn't be too hard to figure out if those numbers correlate with different speed installed RAM.(*)

(* I also believe some of the "Centris" models shipped with 4MB instead of 8MB, and weren't there also variants that lacked the AAUI internet connector? That could also explain different board part numbers. According to Wikipedia both the Centris and Quadra label variants could ship with either 4 or 8MB, but *all* Quadras have AAUI, not so on the Centris. Looks like all 800's have 8MB.)

 

[Gorgonops]

... of course, all of the above is more evidence of 90's Apple being silly 90's Apple. A typical 486 board of the time wouldn't bother with soldered on RAM, Ethernet would go in a slot, and you were *starting* to get into the period where motherboards had PLLs working off a fixed master crystal so you wouldn't even have to change crystals to get a different clock speed. All this would let your rotgut cloner use the same board for everything from a bottom-feeding 486SX/25 with 4MB of RAM up to a fully-loaded 486DX2/66 (or even the Rare and Mystical 486DX/50-no-doubling) just by grabbing the appropriate bits off the shelf, with no need to stock different motherboards... :

 

[Themk]

It's worth noting that the RAM speed difference is likely due to the fact that the system bus runs at a different speed in the Centris 650 and the Quadra 800. I wouldn't be surprised if the reason why the 800 was original spec-ed with faster RAM was due to the higher (33MHz) System bus. Once the Quadra 650 came online with its identical speed to the Quadra 800, I wonder if apple switched to the faster RAM for those models too. Someone who has some Wombat boards should check out the soldered-on RAM has Gorgonops mentioned.

 

[Themk]

I'm still slowly plodding away at getting Color X support on NetBSD for this computer. I'll let everyone know my results.

Color X server is a prerequisite for me attempting to try to get Basilisk II to run in Native CPU mode. Granted there is a lot of 1-bit software for the Classic Mac OS, but in that case I would rather run Mini vMac.

 

[CC_333]

What would be fun is a modern BSD package that allows a 68k Mac to do modern things (however, since even a fairly quick G5 seems to be beginning to feel quite a bit of strain trying to access the internet these days, getting a 68k to do that is probably impossible (or, at least, hugely impractical), but still....).

c

 

[jack]

BSD already can do modern things... ;D

And if you want to do modern things on a 68k running *nix, you just stick to the lightweight stuff.

 

[Themk]

Part of me wonders if OpenJDK could be made to work with NetBSD/mac68k

modern BSD package

I'd call NetBSD modern. The version (7.1) I'm using on this machine was released in March 2017. NetBSD is still seeing active development.

 

[Bolle]

So to get back at the RAM timing thing... I just opened up my WGS80 (Q800) to prepare it for getting sold. It has 60ns Texas Instruments RAM chips soldered to the board.

The original RAM SIMM that must have been in it when it was bought has 60ns chips as well. The three remaining SIMMs that must have been added later on are 70ns and 80ns.

The system is working normal with mixed RAM so far. If it really is running a more aggressive timing on the RAM I would expect it to fail with 80ns.

Now if someone could take a look at a Q650 board to check if it has 80ns RAM soldered down.

 

[Themk]

Thank you Bolle for the information! This is very useful. Perhaps a Centris 650 would be a better thing to look at, as that was before the speed bump to 33MHz for the 650 series.