home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.mac.hardware
- Path: sparky!uunet!nwnexus!kanefsky
- From: kanefsky@halcyon.com (Steve Kanefsky)
- Subject: Re: My $.02 about RAM Caches...
- Message-ID: <1992Nov17.174408.16999@nwnexus.WA.COM>
- Sender: sso@nwnexus.WA.COM (System Security Officer)
- Organization: The 23:00 News and Mail Service
- References: <1992Nov16.215823.207@physc1.byu.edu>
- Date: Tue, 17 Nov 1992 17:44:08 GMT
- Lines: 66
-
- In article <1992Nov16.215823.207@physc1.byu.edu> seth@physc1.byu.edu writes:
- >Hi. I wanna share a couple of things I have found about RAM caches. This is
- >particularly in regard to the posting by Chuck Chung entitled something like
- >"68040 33 MHz -> 40 MHz...", where he asked about cache cards. I don't have a
- >quadra, so I cannot comment specifically on that issue, BUT, I have messed
- >about a bit with caches. Here goes: The 68030 chip has two caches (like the
- >68040, but they don't work quite the same way), which are a 256 BYTE cache and
- >a 256 BYTE instruction cache. Now, this doesn't seem like a lot, but IT IS. I
- >found a control somewhere (I forgot where, but I could probably find it again
- >if someone wants it. it is probably to found at sumex-aim). Anyways, I found
- >that can turn off the caches in the 68030. I did some tests. I turned these
- >256 BYTE caches off, and ran Speedometer. I found out that these caches just
- >about DOUBLE the speed of my IIsi (still at 20 MHz...). I also have a 64 K
- >external cache from Logica Research. Now, I understand that an internal cache
- >is faster and all, but this 64K cache still only brings my machine up maybe
- >25% over the performance I get with JUST the 2 internal 256 BYTE caches. Now,
- >if you were to draw a graph of performance enhancement versus size of cache,
- >you would probably see that the graph rises VERY fast at first, as you go from
- >having no cache at all to having ANYTHING, but then it quickly flattens out.
- >We see that the first 512 bytes of cache memory doubled the computer's speed,
- >and the next 64000 bytes could only bring it up another 25%. You can see that
- >the difference between a 32 K cache and a 64 K cache is not going to be that
- >much, and the difference between a 64K cache and a 128 K cache even less. Now,
- >the 8 K worth of cache memory of the 68040 is internal cache, which is faster
- >(at least from what I have read; I could be wrong) than external cache. With
- >8 K cache, you have probably gotten a sizeable chunk of the possible increase
- >over 512 bytes, so I would guess that spending the $400 for the DayStar cache
- >may not bring much in the way of extra speed. I don't know this as a fact, but
- >I would definitely check out with somebody who did spend the money and has one.
- >If you do, by all means run Speedometer and post the results for us all to see.
- >I too have seen the DayStar ads, and I have to laugh when I see the line
- >"Up to 40% speed increase." That's what the Logica Research ad said too. Sure,
- >8-bit graphics performance went up 40%, and THAT IS IT. Everything else was
- >down around 25%.
-
- I think the speed gains from a RAM cache are dependent on the difference
- between the speed of the CPU and RAM access time. For example, if you had
- a 50mhz CPU, everything else being equal, the CPU could process a lot more
- data in the same amount of time, and could thus benefit from a larger cache.
- I actually do have a 50mhz 68030 CPU in my LCII (DayStar PowerCache), and
- turning the cache on approximately doubles my performance. Running the
- 50mhz 68030 without the cache on is only a slight (about 50%) improvement over
- the 16mhz '030. If you'd like some Speedometer results, I can send them
- (but I don't have the non-cache results handy right now).
-
- How well the DayStar FastCache Quadra performs will depend on how much of
- a bottleneck the rest of the system is relative to the CPU. I must admit
- I have seen a couple people on CompuServe complain that they didn't get as
- big of a speed boost as they expected. Luckily, you can buy one with a
- 30-day money-back guarantee from most Mac mail-order places, so you can
- try one out and see if it makes a difference. I *can* assure you that
- DayStar doesn't want to sell *anything* to anyone that isn't going to be
- happy with it. They have an excellent reputation for customer service.
-
- One thing that just occured to me: I wonder if the FastCache will work
- with programs that can't handle the '040's internal caches. The caches
- in the PowerCaches are certainly compatable with everything under
- the sun. In this case, you could run at nearly full speed with programs
- where you'd normally have to disable the '040's internal caches and slow your
- machine down to a crawl. I'll check with Daystar on this and find out.
-
-
- --
- Steve Kanefsky
-
-
-