|
|
|
|
| Welcome, Guest | Home | Search | Login | Register | |
| Author | MiniBench Scores (Read 101330 times) | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Deer Steak
16 MB  Posts: 25 
|
Reply #30 on: October 07, 2006, 21:10
Macintosh LC III - 0.2 I think my two Powerbook results are interesting. They are completely the same software-wise, aside from the 3400 having to use the PB-specific release of the software, vs. the 1400 having a vanilla install. They have the same 512k disk cache, the same 24MB virtual memory setting, the same amount of RAM, both run with no open apps, the same versions of all software, so the difference is in the memory bus speed (33MHz on the 1400 vs 40MHz on the 3400) and in CPU speed/cache. Powerbook 1400cs 117 running 7.6.1 with 16MB RAM (603e cacheless) - 34.6 Powerbook 3400c 180 running 7.6.1 with 16MB RAM (603e 512k L2) - 53.7 Estimated 180MHz 603e with no cache in 7.6.1 (calculated using the 117MHz) - 53.2 L2 cache speedup - almost nill. Maybe even more interesting in the Power Macintosh 5500 results below, because now it's the exact same piece of hardware, the difference being the L2 cache being either installed or removed. FWIW, 7.6.1 is super snappy with it either way, though web browsing enjoys the extra cache. Power Macintosh 5500/225 with 32MB running 7.6.1 (603e, 256k L2) - 67.3 Power Macintosh 5500/225 with 32MB running 7.6.1 (603e, L2 cache removed) - 67.0 This is the nail in the coffin for L2 cache having an effect on PowerPC, IMO. The 68k's have a much smaller L1 cache - the Powerbook 540c listed above and my LC III both have a measly 8k L1 cache, so an L2 might have a more profound effect. Unfortunately, I'm unwilling to spend the money on a Mac that can use a cache card - and the cache card itself. LOL |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
dpaanlka
|
1024 MB Posts: 1646
Reply #31 on: October 09, 2006, 02:37
|
I bet it has to do with the bus speed between the processor and memory.
|
madmax_2069
|
16 MB Posts: 28
Reply #32 on: October 09, 2006, 07:54
|
yea that is my outake on it. i was figuring it was limited by the memory bus AKA system bus
|
Deer Steak
|
16 MB Posts: 25
Reply #33 on: October 09, 2006, 16:30
|
I don't think it's that, either. Dan's 400MHz G3 in an 8600 (so that's what, a 40MHz bus) scored a 138. My 333MHz Beige G3 scored 113. That's strictly by-the-books there. If you set yourself up a little cross-multiplying table there 400/333 = 138/? you get 114.9. That's less than a 1% difference in what I got, something I consider within the margin of error. But the bus is like 35% faster on the Beige, so if it was RAM enough to saturate the memory bus, I'd think the beige would score higher than linear. It's certainly not hitting the hard drive, so my process of elimination says that it's small enough to fit in the 32k L1 on the 603e, and the 64k L1 on the G3 is positively spacious by comparison. And that's excellent news for PowerPC - Dan's program is calculating one variable over and over, re-using the same memory address. That's exactly what cache was built for.
|
madmax_2069
|
16 MB Posts: 28
Reply #34 on: October 10, 2006, 02:00
|
well yea also when you take into account that emac G4 800 mhz only scoring 176.5 to my Beige G3 with a G3 466mhz cpu score of 160.5. all in all this is all very interesting
|
Deer Steak
|
16 MB Posts: 25
Reply #35 on: October 10, 2006, 03:50
|
I think I figured out that issue earlier when I talk about my iMac. http://arstechnica.com/articles/paedia/cpu/ppc-1.ars/1 The PowerPC used a 4-stage pipeline from the 601 through the original G4 (7400/7410). Then Motorola broke with that tradition, almost doubling the length of the processing pipeline with the G4e, making it 7 stages. We're talking about a 75% lengthening, which also means that, all else being equal, it takes a 75% faster CPU to match a 7400/7410 G4. That means a 500MHz G4 7410 is roughly as fast as a 775MHz G4e. The G3 and G4 variants are covered here: http://arstechnica.com/articles/paedia/cpu/ppc-2.ars/1 I really don't see any other compelling arguements as to why an 800MHz machine is nearly outperformed by a 466MHz machine. The G4 was on both ends of the Megahertz Myth. Had Motorola merely applied some processing tweaks and shrunk the feature size of the G4e, we'd have barely seen 1GHz G4's, but they'd be every bit as fast as 1.42GHz Mac Minis.
|
Deer Steak
|
16 MB Posts: 25
Reply #36 on: October 10, 2006, 03:53
|
I'd be really interested in seeing a 2GHz G5 probably only hit about 200 or so on this benchmark, because one operation depends on the one before it. But there's your 21-stage pipeline at work. It'd be about as fast at this as a 6-year-old high-end B&W G3. Unfortunately for Dan, it'd also kill the usefulness of the MiniBench. OTOH, I'm not capable of doing what he did; I'm no programmer.
|
dpaanlka
|
1024 MB Posts: 1646
Reply #37 on: October 10, 2006, 03:56
|
How many of you are running this under OS X's classic mode?
|
madmax_2069
|
16 MB Posts: 28
Reply #38 on: October 10, 2006, 12:22
|
yea i have 10.4.8 tiger on this Beige G3 (runs great i might add, i use tiger as my main OS now), but i booted directly into OS 9.2.2 with extensions off to do this test. i can how ever test in classic mode and test with extensions on if you need me to.
|
Deer Steak
|
16 MB Posts: 25
Reply #39 on: October 10, 2006, 13:56
|
I've done it in Classic and 9.2.2 on that iMac and my score was within one or two points, so it's not having adverse effects on this machine. The eMac 1.25 I posted earlier was in Classic as well, because it's X only.
|
madmax_2069
|
16 MB Posts: 28
Reply #40 on: October 14, 2006, 04:32
|
wow this thread died awful fast
|
dpaanlka
|
1024 MB Posts: 1646
Reply #41 on: October 14, 2006, 07:10
|
yeah mid terms are happening now then starts term papers and finals for the next month or so i'll be pretty busy
|
Deer Steak
|
16 MB Posts: 25
Reply #42 on: October 14, 2006, 16:01
|
Basilisk II on an Intel Mini using the 68040 JIT compiler
|
Deer Steak
|
16 MB Posts: 25
Reply #43 on: October 15, 2006, 05:52
|
The same machine gets a 124 on OS 8.6 running SheepShaver. Pretty impressive, given my iMac 800's 176.
|
Brooklyn
|
16 MB Posts: 21
Reply #44 on: December 11, 2006, 09:57
|
My Powerbook 1400c gets 80.0 with the Nupowr 233mhz G3 and 32mb ram. I have to run the computer with extensions on though, so the Nupowr drivers load on startup.
|
|
Pages: 1 2 [3] 4
|
| |||||||||||||||
|
© 2021 System7Today.com. |
