03/09/98 - Monday's News:
- ByteMarks Tests: Awhile back I ran some tests using the Bytemark (Motorola version) benchmark. The results showed my PowerBook G3 (no extensions loaded) has better Integer and Floating Point performance than my Pentium II 300mhz system. With the PII running at 75mhz bus, 338mhz CPU speeds it just barely edged out the PowerBook G3 in Floating Point, but still trailed in Integer performance by a good margin.
System
(Bus/Cache/CPU speed) | Integer
Index | Floating-Pt
Index |
PowerBook G3
(50/100/250) | 6.998572
| 5.228775 |
Pentium II 300
(66/150/300) | 4.226743
| 4.829937 |
PowerBook G3
(50/166/250) | 7.071710
| 5.396240 |
Pentium II 338
(75/169/338) | 4.754942
| 5.398357 |
Both systems had 128MB of Ram and 512k of 2:1 cache. The lower two
scores were taken with the PowerBook G3 cache running at 166mhz (using PowerLogix's G3 Cache Utility), Libmoto and Speed Douber 8 active. The PII 300 used a ABIT LX6 Motherboard set to 75mhz bus, 4.5X ratio (337.5mhz, 1/2 speed 512K cache).
I'll be running the same test in the future on all CPU cards and systems as part of the standard review benchmarks. The BYTEMARK benchmark is available at Byte's download page.
- More Apple G3 Speed Bumps: I've gotten several reports of successful
Speed Bumping of the 233 Mhz G3 systems to 83mhz bus/290mhz CPU speed. Here is an example from Sam writes:
"Finally wrangled a G3 for use at work out of my boss. 233DT. I remembered someone saying that the limit to higher bus speeds was the RAM and you had to use non-Apple RAM? [That is true in some cases, but not always.-Mike]
Well I did the "two jumper" trick on the G3 (take it to 290Mhz/83Mhz) about three weeks ago now and the thing's been running like a train ever since with the
standard Apple RAM. No crashes. I also put my hand on the heat sink and
it's practically cool to the touch even at that speed. MacBench of 957 and
730 floating point with Virtual Memory, Appletalk and Filesharing ON. I
wonder, can most of these G3s (even 233s like mine) take this jump to
290/83?"e;
I'm beginning to wonder - as this was the 3rd report this weekend.
Bill sends in the results of his MT 266 boosting efforts:"
Thank you for the countless hours you spend keeping xlr8yourmac on
the web. It is the one site I visit _every_ day. [Hey, I couldn't cut that out-Mike ;-] I've looked through the Forum, FAQs, etc. as well as the links on your
page but I still need some advice. Using Imai's settings, I produced the
following results on my G3 266 MT with 96 Mb RAM:
CPU
Speed | Bus
Speed | PCI
Speed | MacBench
CPU Score |
266
(stock) | 66 | 33 | 897 |
300
| 66 | 33 | 1001 |
292
| 83 | 33 | 970 |
315
| 70 | 35 | 1043 |
Is the MacBench CPU score the bottom line in overall acceleration? I
can leave the Bus at 66MHz and clock up the CPU to 300MHz for 1001. Or,
should I key on the bus speed, not the CPU speed? A 70MHz Bus at 4.5
multiple produces an apparently stable 315MHz CPU for a MB score of
1043. With an 83MHz Bus, 3.5 is the highest multiple I can get (higher
multiples are crashes or no-starts). This yields a 292MHz CPU with a
less than 1000 MB score.
Assuming clock chipping can be harmful; which is potentially _less_
harmful over the long run, higher CPU or Bus speeds? Which produces the
overall *faster* machine? BTW, what is the definition of fast? And what is the
meaning of life?"
"Fast" is always a relative term, and no computer ever seems to be "Fast Enough" for long it speems, as we quickly become accustomed to speed. That's why I keep my stock 8500/120 CPU card around. Reinstalling it is a good way to regain my perspective ;-).
In my opinion the high CPU speeds are the biggest stress/risk. The duty-cycle (or "on time") is very high on a 315 Mhz part - that increases heat of course, and is the primary concern in this scenario. The relatively low PCI and Bus speeds are less of a concern, and usually if there is a PCI device or SDRAM module that can't tolerate that speed you'll see it fairly soon. The main concern is how much do we shorten (if any) the life of the CPU? The G3/PPC 750 has not been in use long enought to really tell. I've asked for long term feedback on this to see what the effects are, but realize that you are taking a risk. I've run a 250 Mhz PPC750 chip at 292mhz since October with no ill effects so far, but that's not a guarantee that another chip will do the same. Some people in the industry say the IBM CPU chips run faster and longer when overclocked that the Motorola versions but I'm not sure anybody really knows (outside of the IBM or Motorola labs anyway).
Getting back to the question of the best setting to use, remember that the MacBench CPU score is affected primarily by two things: Clock speed (CPU/Cache/Bus) and Cache size/speed. For the same system configuration usually the fastest CPU speed will produce the highest CPU score. However look at all the other results as well -FPU (usually scales directly with CPU speed), Video and Disk test scores as a whole to determine which setting is the best overall performer. In general most applications that are CPU/FPU intensive (i.e. Photoshop filters) will benefit most from the faster CPU speeds. It's best to test the combinations in your favorite applications doing timed tests to see which really performs best. I
suggest taking two settings - one with the max CPU speed and another with the max
Bus speed and compare timed tests in the applications you use most to determine which is the best performer. From the chart above, I'd guess the lower setting would be the fastest off the top of my head, but it's also the
one that stresses (overclocks) the CPU the most.
- Apple Posts Power Computing Manuals Online: Billy Pelzel wrote to let me know that Apple has posted a lot of Power Computing manuals at their FTP site:
"
Hey Mike! I thought I would pass this info on to you about the manuals
section of Apple support ftp site. You already mentioned this
prieviously but they have added a section on Power Computing! The
section has everything from manuals for every Power machine made, to a
little history of PCC, to HD/scsi jumper settings, RAID info,
troublshooting manuals and a bunch more! As a PC user this is great info
for troublshooting etc, and the history of PCC is interesting too. I
know you have a PTP or two so I thought you might want to pass this info
on and possibly put it in your Power Computing section of your
homepage.....by the way your page is truly incredible! Your efforts are
truly appreciated, keep up the good work!
Go to ftp://ftp.info.apple.com/Apple.Support.Area/Manuals/
"
Thanks Billy! I've added this link to the PowerCC section of my Links Page as well.
- More Initio Miles SCSI card Feedback: Jeff H. Yip wrote in to report
problems with video capture with the Initio SCSI card and Cheetah drive. I've commented in the past a lack of control software for the Initio card (no video tuner capability) would make other SCSI cards with this feature a better choice for video/audio capture. Jeff wrote:
"
I recently went for the G3/initio/cheetah upgrade, and have discovered a
weird problem. Wondering if you have ever heard of it before or have any
thoughts. Here's the email I sent MacGurus:
Dear MacGurus,
I just purchased a Initio UltraWide Card and 4 gig wide Cheetah from you
guys and have come across a fairly serious problem I cannot resolve (and
I have spent considerable time). I am wondering if you have ever come
across this problem yourself or had any reports from other customers.
I bought the drive to do some home video on my 8500 (which has been
upgraded with a NewerTech G3 250/512/125, running system 8.1). My problem
is the the Cheetah does not play back quicktime movies correctly. It's
hard to tell, but I think it's playing them too fast! (or just skipping).
Video and sound are choppy and sped up. MoviePlayer shows frame rates
going above 30 when the movie is only recorded ~ 30fps. Recording seems
to be ok because if I record on the cheetah and copy the movie over to
the stock 1 gig drive, it plays correctly (though it drops frames cuz it
can't handle the data rate, but sound is fine). Finally, if I just copy a
demo movie from a cd (e.g. MacBench), it will not play correctly from the
cheetah, but fine from the stock drive (even movies with low data rate).
Is this weird or have I totally missed something?
Here are the test I performed to try to resolve/isolate:
Works:
G3, motherboard scsi, stock HD
604/120, motherboard scsi, stock HD
G3, initio card, stock HD on 50 pin internal
Doesn't work:
G3, initio, cheetah, FWB low level formatted (optimized mode page)
G3, initio, cheetah, FWB formatted (digit video mode page)
G3, initio, cheetah, Conley driver
G3, initio, cheetah, Conley low level formatted
604/120, initio, cheetah, conley
G3, initio, cheetah, Drive Setup 1.4 low level formatted, system 8.1
G3, initio, cheetah, Drive Setup, system 7.6
Please help, this is a serious problem for me as video production is why
I bought the card/drive combo. It would also seem like a common setup for
other people so I am wondering if anyone else has had this problem.
Thank you,
Jeff H. Yip
(510) 848-8160 (daytime)
P.S. The other annoying thing is the buzzt sound it makes 1/2 second
after every read or write. Other people on usenet have noticed this but
there is no concensus as to why and how to stop it/stop it from happening
so frequently. Does this happen for every Cheetah drive?"
I've had similar reports (posted here) from 2940UW owners recently regarding Audio capture, which were solved by changing the Video Tuner options (PowerDomain Control Panel menu option) to "cache line" settting according to one reader. Also note that the Bravado 1000 video capture card documentation recommends against using a PCI SCSI controlled disk for the same reason. If you're planning on capturing video/audio to a PCI SCSI controller, you're better off buying one that has a Video Tuner capability (2940UW, ATTO, etc.). The Jackhammer has a special extension for use with video capture cards for instance. If you can get by with less audio quality, setting the audio to 22Khz vs 44Khz may also help (cuts the sampling rate/data size in half)
Ken Dunnington wrote:
"I picked up an Initio card a few weeks ago, and although I have no complaints, since I don't do anything that really requires an extremely
fast I/O and I was running out of SCSI ID's anyway, I was kind of
dissapointed after some of the tests I did. Here are FWB Benchtest scores
for my internal 2.1 GB SCSI-2 drive; the first test is on the internal bus,
along with the stock Apple 1.2 GB without the card installed; the "before"
test. The second is both drives on the card, and the third is the 2.1GB
alone on the card. I've got a SCSI-2 external drive that I really want to
put on the card, but I'm waiting for a 68-pin adapter to arrive, so I don't
know how the card will impact that drive. Apologies if the formatting of
the FWB results are messy - they don't paste well into Eudora.
(Results are for a Quantum Fireball 2.1 GB int, formatted with FWB 2.0.6)
Congifuration: Sus Read (kb/sec) Sus Write (kb/sec) Average
------------- Low Peak Avg Low Peak Avg. Access (ms)
Internal Bus 3020 5577 4405 2723 5783 3869 20.8
2.1 & 1.2 GB 2119 4114 3150 2188 4366 3115 21.1
2.1 GB Only 2120 4314 3104 2034 4738 3109 20.8
Average Reads Writes
Seek(ms)(sec) (sec)
Internal Bus 9.1 309 286
2.1 & 1.2 GB 9.1 319 300
2.1 GB Only 8.9 321 300
My disk cache is set to 7680k, and the machine is an 8500 with a 250
MHz/1mb L2 cache G3 from Newer installed. I have 256 MB RAM, running OS
8.0 with Speed Doubler 8.0 and a typical set of extensions. I have the
card in slot A1, the closest fo the CPU, and I have an Xclaim VR RagePro in
slot C1.
Another odd note about the Initio - it woudln't work properly if the SCSI
chain was terminated correctly. Even though my stock Apple drive was
terminated, when I hooked up the card, the Apple drive, and the 2.1GB
drive, I had to put the 2.1GB as the last device, otherwise the Apple
wouldn't show up. With the 2.1GB drive by itself, I have to leave it
unterminated, or it won't show up! Everything works fine, though - no r/w
errors, no lockups or anything. Very bizarre. I'm going to eventually try
reformatting all drives with the Conley SoftRAID that came with the card
(it's supposed to have a good driver, even for HFS volumes).
Keep up the fantastic site! We all really appreciate what you're doing for
the Mac community."
Thanks Ken!
- MacWorld NY (Summer) I've been asked if I'd be interested in speaking at MacWorld (NY) this summer on the topic of "Saving Your Mac from Obsolescence or 101 Things to do with a Dead Mac". I'm very excited about this, and unlike last year I'm making plans now to have the time off to attend. Hope to see you there!
- VillageTronic sent this press release announcing their new "3D Village Partner Program". It's too long to list here, but all the details are in the press release.
- Mike McConnell wrote with an update on his FWB HDT 2.5 experiences:
"
Yep, I'm the guy that was defending FWB a couple of weeks ago. I've since
found a problem with HDT 2.5.
Problem: it can't partition my drives. I came to find this out after
installing a Initio ultrawide scsi card and plugging a new Cheetah onto it.
HDT 2.5 formatted the drive fine(low level). But after successful
formatting, if wouldn't let me partition the drive. Said I didn't have
enough free space, although 4.2 GBs was available. I increased memory for
HDT, same results. I also tried to create new volumes on drives that were
already HDT 2.5 formatted and partitioned, same result.
Solution: I installed the Conley SoftRaid that came with the Initio card.
Re-initialized the new disc. It allowed me to partition. No problem. Fast.
Problem solved.
Changing of the guard?
Mike McConnell
Glows in the Dark Studios"
- PowerForce 275/275 Feedback: David Badovinac wrote to say he's very happy with his new card, and that it's lived up to the claim of 300/300 operation so far:
"
I wrote you earlier about having purchased the PowerForce 275/275 from
Bottomline (thanks to the advise on your site). It ran at 300/300 right out
of the box on my PowerMac 7300.
I'm running MacOS 8.1 and I have 80 mb.s of ram. I have Libmoto installed
and I use 2560K of disk cache. All my regular extensions were installed
too. Today, I changed the settings, slightly, to set the bus speed at 45
and the bus ratio to 7:1...and it worked fine...I'm now getting 315/315.
After I wrote you and told you about this I ran MacBench 4.0 on my new set-up.
The processor score went up from 1159 (at 300 mhz) to 1218 (at 315mhz) and
the FPU score went from 972 to 1020. I was amazed!!!
" I've got a PF 275/275 here to test, but I've got to finish
the 3D work for my employer first, and it's not going well unfortunately.
- This weekend has been a very frustrating one, with most of my time spend in Windows 95 3D programs. I've seen dozens of GPFs - losing work and corrupting files. It has been the most frustrating weekend in recent memory. I'm about ready to throw in the towel but am going to work on it again today.
- More news later tonight when I take a break from WinDozing...
Yesterday's News
Please keep this site alive by supporting my
Sponsors
|
