wirehead@oxy.edu (David J. Harr) (11/02/90)
Ok, I am finally taking the big step and upgrading my Mac II to a IIfx. Since I have to buy all new memory anyway, I am considering getting 4 4Meg SIMMs and 4 1Meg SIMMs instead of 8 1Meg SIMMs. This is a signifigantly more expensive proposition, so I want to ask all the people out there in A/UX land: Is it worth it? Using 20 MBytes of memory, do you see any major improvement while running A/UX as compared to a stock (stock? I remember when I lusted after a machine with a whole *2 Megs* in it. Ah, the good old days...) 8 Meg machine? I don't foresee any really large compiles or anything under A/UX although I might try out my new shiny MPW 3.2 and see if it really will compile in the background under A/UX while running MultiFinder as the front process. Anyway, I would like some guidance on this (Actually, I'm probably really just looking for enough people to say "GO FOR IT" that I will fell justified in maxing out yet *ANOTHER* credit card doing this, but that's another story...) to help me make an informed decision. The preceding opinion was another fine product from the fevered brain of wirehead@oxy.edu "When you want wacko opinions, get only the very best."
abm@alan.aux.apple.com (Alan Mimms) (11/02/90)
In article <121667@tiger.oxy.edu>, wirehead@oxy.edu (David J. Harr) writes: |> Ok, I am finally taking the big step and upgrading my Mac II to a IIfx. |> Since I have to buy all new memory anyway, I am considering getting 4 4Meg |> SIMMs and 4 1Meg SIMMs instead of 8 1Meg SIMMs. This is a signifigantly more |> expensive proposition, so I want to ask all the people out there in A/UX |> land: Is it worth it? Using 20 MBytes of memory, do you see any major |> improvement while running A/UX as compared to a stock (stock? I remember |> when I lusted after a machine with a whole *2 Megs* in it. Ah, the good old |> days...) 8 Meg machine? If you run a large Macintosh world or if you run a lot of large processes, you WILL see a significant improvement. You cannot get 10 kilos of manure into a 5 kilo bag. :-> Particularly, X11 client and server take a lot of space (we're working on this), so if you use X, you might consider going for the 16MB or more. I have a 20MB FX here and the difference is like night and day for my 16MB Macintosh development environment and MacX world and all of the X11 clients I run all day long. You also see multiple processes running on the machine (say, a background compile and the Macintosh world) as "fighting" less, subjectively. The real reason is the reduced paging, but the result is that you seem to get more from your CPU in terms of real work per nanosecond. |> I don't foresee any really large compiles or |> anything under A/UX although I might try out my new shiny MPW 3.2 and see |> if it really will compile in the background under A/UX while running |> MultiFinder as the front process. Anyway, I would like some guidance on this |> (Actually, I'm probably really just looking for enough people to say "GO FOR |> IT" that I will fell justified in maxing out yet *ANOTHER* credit card doing |> this, but that's another story...) to help me make an informed decision. I recommend it. Software only tends to get larger in the long run. GO FOR IT! |> |> The preceding opinion was another fine product from the fevered brain of |> |> wirehead@oxy.edu |> |> "When you want wacko opinions, get only the very best." Mine are about as wacko as you can get. Luck. Alan Mimms (alan@apple.com, ...!apple!alan) | My opinions are generally A/UX X group | pretty worthless, but Apple Computer | they *are* my own... "Laugha whila you can, monkey boy..." -- John Whorfin in Buckaroo Bonzai "Never rub another man's rhubarb" -- The Joker in BatMan
alexis@panix.uucp (Alexis Rosen) (11/03/90)
Lesse now.
8 1MB SIMMs: 8 x 40 = $320
4 4MB SIMMs: 4 x 205 = $820
So for $500, you can double your memory from 8 to 16 MB. I can't speak for
you, but I personally wouldn't consider any other alternative. A/UX by itself
is plenty happy in 8MB (_not_ 4, though), but if you want to do lots of Mac,
X, or disk-intensive stuff, you'll page fairly heavily. A 16MB system with
4 or 6 MB or buffers makes for an excellent personal workstation, though...
---
Alexis Rosen
Owner/Sysadmin, PANIX Public Access Unix, NY
{cmcl2,apple}!panix!alexissramtrc@windy.dsir.govt.nz (11/05/90)
Another factor to consider when using memory is the size of the disk cache (by default 10% of available memory). Even if you are not running large programs or not doing lots of paging activity the benefits of a larger disk cache can still be had. With a cache of size S MB the last S MB of files that you accessed remain in RAM and so are faster to access next time. So even if you are not a heavy user, commands execute more quickly. If you buy 1MB SIMMS and want to add more than 8MB in the future you will have to remove those SIMMS. But if you buy 4MB SIMMS you won't have to remove them till you go past 32MB. Fewer trashed SIMMS. Tony Cooper sramtrc@albert.dsir.govt.nz
barnett@grymoire.crd.ge.com (Bruce Barnett) (11/05/90)
With all this talk of adding 16 megabytes to Mac's, does anyone know of a NuBus card that can be used to add memory? Here's an interesting thought: I can get as many 256K RAMS as I want for free. Could a NuBus board handle 16 Meg of 256K RAMs? If such a board could handle 256K, 1M, and 4M modules, This would be the way to go. -- Bruce G. Barnett barnett@crd.ge.com uunet!crdgw1!barnett
alexis@panix.uucp (Alexis Rosen) (11/06/90)
Bruce G. Barnett (barnett@crdgw1.ge.com) writes: >With all this talk of adding 16 megabytes to Mac's, does anyone know >of a NuBus card that can be used to add memory? >Here's an interesting thought: I can get as many 256K RAMS as I want >for free. Could a NuBus board handle 16 Meg of 256K RAMs? >If such a board could handle 256K, 1M, and 4M modules, This would be >the way to go. Yes, but no... National Semiconductor has a board that can handle up to 16MB, but I don't know if it can take 256K simms. In any event, you probably don't want to do this. The money you spend on the board could instead go to 4MB SIMMs. Furthermore, access to that RAM would go over NuBus, and would thus be _slow_ compared to regular memory. (I seem to recall a minimun 300ns access time, though I'm not sure. That would make it about 4 times slower, although you might make some of it back on wait states you wouldn't need.) Tech issues aside, the economics of the situation pretty much rule it out, I think... --- Alexis Rosen Owner/Sysadmin, PANIX Public Access Unix {cmcl2,apple}!panix!alexis
barnett@grymoire.crd.ge.com (Bruce Barnett) (11/06/90)
In article <1990Nov6.041551.2038@panix.uucp> alexis@panix.uucp (Alexis Rosen) writes: >In any event, you probably don't want to do this. The money you spend on the >board could instead go to 4MB SIMMs. Furthermore, access to that RAM would go >over NuBus, and would thus be _slow_ compared to regular memory. (I seem to >recall a minimun 300ns access time, though I'm not sure. That would make it >about 4 times slower, although you might make some of it back on wait states >you wouldn't need.) I didn't realize this. I knew the NuBus was slow, but this is ridiculous. I guess I have been spoiled by *real* workstations. :-) > Tech issues aside, the economics of the situation pretty much rule it out, > I think... Well, I thought if I could get a bare board for $200-$300, and get 16 Megs of 256K SIMMS for free, I could come out ahead. I guess not. -- Bruce G. Barnett barnett@crd.ge.com uunet!crdgw1!barnett
alexis@panix.uucp (Alexis Rosen) (11/07/90)
In article <BARNETT.90Nov6083827@grymoire.crd.ge.com> barnett@crdgw1.ge.com writes: > [in response to my note about the Nat. Semi. RAM board] >I didn't realize this. I knew the NuBus was slow, but this is ridiculous. >I guess I have been spoiled by *real* workstations. :-) Well, let's not be _too_ critical. It's a lot more powerful than the Sbus. And it's not so far behind VMEbus. But today's high-speed processors demand a specialized high-speed connection to memory. This includes all the Sparc machines and the RS-6000s (current speed champs?) as well as the Mac. They all use SIMM slots of some sort. (The IBM has an intermediary board which does all sorts of slick stuff, but it's irrelevant in this context). In fact, on the IBM, even with the direct access to memory, they need such a high bandwidth that they interleave four ways (eight?). So the Mac is not really any different from the current crop of workstations. Like all of them, it wants special connections for RAM. The only difference between it and other workstations is that it can't deal with more than 32MB (unless, possibly, you have 16MB SIMMs). --- Alexis Rosen Owner/Sysadmin, PANIX Public Access Unix {cmcl2,apple}!panix!alexis
esmith@goofy.apple.com (Eric Smith) (11/08/90)
In article <1990Nov6.041551.2038@panix.uucp> alexis@panix.uucp (Alexis Rosen) writes: > In any event, you probably don't want to do this. The money you spend on the > board could instead go to 4MB SIMMs. Furthermore, access to that RAM would go > over NuBus, and would thus be _slow_ compared to regular memory. (I seem to > recall a minimun 300ns access time, though I'm not sure. That would make it > about 4 times slower, although you might make some of it back on wait states > you wouldn't need.) 300 nS is the minimum CYCLE time. On a Macintosh IIx, the RAM on the logic board is run with 255.3 nS cycles, which is not that much faster than the maximum NuBus speed. Then again, there is a synchronization delay to access NuBus, and many (most?) NuBus cards don't run a maximum theoretical speed. I don't know exactly how fast the Macintosh IIfx runs DRAM cycles, but it can't be much faster than 200 nS (most 80 nS RAMs are spec'd for about 190 nS cycles). I would expect that main RAM access on the IIfx would be about twice the speed of a fast NuBus card. -- Eric L. Smith Opinions expressed herein do not necessarily reflect those esmith@apple.com of my employer, friends, family, computer, or even me! :-)
barnett@grymoire.crd.ge.com (Bruce Barnett) (11/08/90)
In article <1990Nov7.033951.6420@panix.uucp> alexis@panix.uucp (Alexis Rosen) writes: > >I didn't realize this. I knew the NuBus was slow, but this is ridiculous. > >I guess I have been spoiled by *real* workstations. :-) > > Well, let's not be _too_ critical. It's a lot more powerful than the Sbus. The SBus has a few features: Cheap. Fast. Multiple Masters. The SBus graphics Accelerator (GX) has the ability to scale with the speed of the CPU. The same Sbus device supports the 27 MIP Sparc 2 and will also support a 50 MIPS Sun. If the NuBus is better, send me e-mail off line and we can continue this discussion without boring people. > And it's not so far behind VMEbus. True, Sun added an extra set of connectors on the VME bus to handle fast memory. But the Sun is a workstation company who thinks about DMA, virtual memory, etc. External memory boards worked fine on the Sun 3/Sun4 VME family. > So the Mac is not really any different from the current crop of workstations. > Like all of them, it wants special connections for RAM. The reason why some workstations don't have external memory boards is that they can put enough directly on the main CPU board. Some workstations will need 256MBytes (or more) of memory, and these support a bus that allows access at memory speeds. I guess I was looking for a cheap way to add more memory. Sigh. -- Bruce G. Barnett barnett@crd.ge.com uunet!crdgw1!barnett
aib@stc06.ornl.gov (BLAND A S) (11/08/90)
The National Semiconductor NuBus memory board does not work with A/UX 2.0. When I called, their Tech rep said they had no plans to update the A/UX driver to make it work. Don't buy one for use with A/UX unless you are willing to write your own driver! Buy the 4Meg SIMMs. This is my own opinion. Not my employer's. Buddy Bland Oak Ridge National Lab aib@ornl.gov