d6maca@dtek.chalmers.se (02/11/90)
Hi! I have read a short article about the new 68040 from Motorola. It should be a real beast. The average instruction takes 3.4 cycles on the 68030 and 1.3 cycles on the 68040. It looks like they are doing everything at the same time in this chip. This is a very big step in performance, it's 2.5 times as fast as the '030. I want to know more about the 68040, has anybody heard anything more? flame on The article says that the 68040 is 25% faster than the 80486 at the same clockspeed. flame off - Martin Carlberg - Chalmers University of Technology, Gothenburg, Sweden
halam@umnd-cpe-cola.d.umn.edu (haseen alam) (02/11/90)
There is an article on BYTE magazine on 68040. Its there latest issue. Haseen.
carlo@merlin.cvs.rochester.edu (Carlo Tiana) (02/13/90)
In article <1990Feb10.233412.15585@mathrt0.math.chalmers.se> d6maca@dtek.chalmers.se (Martin Carlberg) writes: >... >The article says that the 68040 is 25% faster than the 80486 at the same >clockspeed. >... While on the subject of "cross-sultural" comparisons (and with the caveat that I know just about nil about the internals of anything more complicated than a ball bearing): why is it that people are always quoting the number of wait states in 80xxx chips and never for 68xxx chips? This probably a silly thing to do, but how does my IIci rate compared to a say, an 80386 machine runnng at 25Mhz with n wait states? On a similar note (equally naive question): why does a 25 Mhz 68030 in my above-mentioned ci feel like it is nowhere near as fast as the 25Mhz 68030 in the news server that allows me to post these silly questions? I certainly would not want other people logged into my Mac reading news while I was typing away at a Word document, because performance would just die (it's bad enough as it is -- though I guess I am spoiled). carlo@cvs.rochester.edu
minich@a.cs.okstate.edu (MINICH ROBERT JOHN) (02/14/90)
From article <5221@ur-cc.UUCP>, by carlo@merlin.cvs.rochester.edu (Carlo Tiana): > In article <1990Feb10.233412.15585@mathrt0.math.chalmers.se> d6maca@dtek.chalmers.se (Martin Carlberg) writes: >>... >>The article says that the 68040 is 25% faster than the 80486 at the same >>clockspeed. >>... > > > While on the subject of "cross-sultural" comparisons (and with the caveat > that I know just about nil about the internals of anything more complicated > than a ball bearing): why is it that people are always quoting the number > of wait states in 80xxx chips and never for 68xxx chips? This probably a > silly thing to do, but how does my IIci rate compared to a say, an 80386 > machine runnng at 25Mhz with n wait states? Good question. Answer: because noone else sells Macs, so you can basically have five machines to compare (Plus, SE, II__, IIx + IIcx + SE/30, and the IIci) so who cares about wait states. Here's the low down... Mac Number of Ram speed for x wait states Model Wait States CPU Speed 0 1 2 3 ------- ----------- --------- --- --- --- --- Plus 0 7.8MHz 130 260 390 520 SE 0 7.8 " " " " Portable 0 15.6 60 120 180 240 II__ 1 15.6 " " " " IIx/cx* 1 15.6 " " " " IIci 2 25.0 40 80 120 160 ???? ? 33.0 30 60 90 120 Mr RISC 0 50.0 20 40 60 80 * includes SE/30 Note: The second column is the result of what Apple gave us. The last columns are what speed of RAM would be necessary to build a machine at speed X with the given wait states. + The IIci has a slot for a "cache card" which is basically memory fast enough to operate at 0 wait states, which allows the CPU to run more efficiently. + The Portable is unique in that all of its RAM is of the static variety (SRAM as opposed to normal dynamic RAMs, aka DRAM) so it operates at no wait states. + The problem with these fancy new RISC machines is that they operate at such high clock rates, the design of a cache system to keep them feed is one of the hardest (if not THE hardest) parts to getting the things to exploit the full capabilities of the CPU. When a RISC machine doesn't get memory when it wants it, it has to hold up an entire pipeline of instructions to do it, and consequently clobbers effective performance. > > On a similar note (equally naive question): why does a 25 Mhz 68030 in my > above-mentioned ci feel like it is nowhere near as fast as the 25Mhz 68030 > in the news server that allows me to post these silly questions? I > certainly would not want other people logged into my Mac reading news while > I was typing away at a Word document, because performance would just die > (it's bad enough as it is -- though I guess I am spoiled). > Most likely because your news server doesn't happen to be supporting a graphical interface between the CPU and yourself. A IIci is very much the number cruncher of a computer. The reason why it doesn't always "feel" that way is because doing something as simple as writing a line of text on the screen involves manipulating about 500 bytes of data for 80 columns on a monochrome screen. (That ignores the effort to do all the stuff like finding the font in memory, finding the each character in the font as you draw each character, etc.) A text based system, on the other hand, manipulates 80 bytes of data to achieve the same effect! (Side note: PC type people have bitched at software companies about communication programs that are SLOWER, but are they're just ignorant of the extra overhead involved in drawing text as graphics under PM. :-) The responsiveness of a the classic Mac in this light is nothing short of incredible. Kudos to the original hackers at Apple! Now, when you do the same thing with a color/greyscale monitor set to, say, 4 bits, you have to at least four times the work, plus the OS on the Mac may have to convert a 1 bit deep font to 4 bits, which takes time... So, THAT's why that darn news server feels like greased lightning. I must admit, playing with the latest PCs in text mode feels good, but you switch them to Windows, and it's like a Plus, but uglier. Robert Minich Oklahoma State University minich@a.cs.okstate.edu This school is too worried about other things to worry about little ole me, so I can say what I want. It's that First Ammendment I keep trying to convince that I try to convince the existence of to these guys here of that will get ME into trouble.
sobiloff@agnes.acc.stolaf.edu (Chrome Cboy) (02/16/90)
In article <5378@okstate.UUCP> minich@a.cs.okstate.edu (MINICH ROBERT JOHN) writes: >+ The problem with these fancy new RISC machines is that they operate at such > high clock rates, the design of a cache system to keep them feed is one of > the hardest (if not THE hardest) parts to getting the things to exploit the > full capabilities of the CPU. When a RISC machine doesn't get memory when it > wants it, it has to hold up an entire pipeline of instructions to do it, and > consequently clobbers effective performance. I just read in the Wednesday edition of the Minneapolis Star Tribune that IBM has finally started production line testing of their new 16 meg SIMMs. The nice thing about them is that they already are running at 50ns, and I suppose that after IBM gets all the bugs worked out, 50ns will be the slowest of the 16 meg SIMMs. This certainly should keep those RISC chips happy, huh? Just FYI... -CCb f d d r :-)
landman@hanami.Sun.COM (Howard A. Landman x61391) (02/19/90)
In article <5221@ur-cc.UUCP> carlo@cvs.rochester.edu (Carlo Tiana) writes: >On a similar note (equally naive question): why does a 25 Mhz 68030 in my >above-mentioned ci feel like it is nowhere near as fast as the 25Mhz 68030 >in the news server that allows me to post these silly questions? Well, does your news server have cache memory? Howard A. Landman landman@eng.sun.com -or- sun!landman