PDA

View Full Version : G5 Specs


MacRumors
Oct 18, 2001, 06:54 PM
The Register posted (http://www.theregister.co.uk/content/39/22328.html) an article remarkable similar to Mac OS Rumors (http://www.macosrumors.com) regarding the upcoming G5's:

Two major problems have been hampering the company's efforts to get the chip into upcoming Power Macs: issues with the clock multiplier and cache coherency glitches. The former has been fixed with the G5's latest revision, version 0.5, we hear.


Despite this, performance tests seem to be very promising...

dougie
Oct 18, 2001, 08:45 PM
What the heck is SpecInt 2000? The register posts P4
tests as if I cared! If the G5 1.6 scores in the 1000's, what we want to know: What does the fastest G4 (867) score?! Then we can kind of gauge the progress here...
possibly significant compared to the Real World Barely faster dual 800 "improvement" over the dual 533?

zaustin
Oct 19, 2001, 01:27 AM
SPEC (www.spec.org) is an industry standard benchmark for CPU Integer and Floating-Point performance. Strangely, Motorola doesn't have ANY of their chips in the SPECCPU2000 database, and with a check to Motorola's chip website (www.mot-sps.com), it seems that they use MIPS as a benchmark for the PPC 7450. For comparison, the PPC 8540 (should be loosely similar to the G5, if rumors are correct) scores a MIPS score of 1385 @ 600 Mhz, whereas the 7450 scores a 1324 @ 733 Mhz. If you scale the scores, the 8540 would have a 1692 score at 733 Mhz. Doesn't tell you THAT much since this probably isn't the exact same chip design that Apple will use (the 8540 is an embedded chip), but it does look promising. Motorola SPEC scores are available for the older benchmark SpecCPU95, and for reference, Motorola's SPEC scores have NEVER been as high as Intel's.

Zach

Pants
Oct 19, 2001, 03:37 AM
Originally posted by dougie
What the heck is SpecInt 2000? The register posts P4
tests as if I cared! If the G5 1.6 scores in the 1000's, what we want to know: What does the fastest G4 (867) score?! Then we can kind of gauge the progress here...
possibly significant compared to the Real World Barely faster dual 800 "improvement" over the dual 533?


hmm - well there was a link from the registry page to the SPEC site that included a handy FAQ. Im sure clicking that link and reading it would have saved you valuable typing effort. As for why motorola dont release spec marks for current machines and only for this 'future' chip, I suspect its because current chips perform pretty poorly in comparison.

toofeu (France)
Oct 19, 2001, 03:38 AM
Well what I really care is the shipping date of those babies.
I'm actually a PeeCee User and I wait the forthcoming of the new CPU to buy a PowerMac. I was ready to buy one in July but I was so disapointed by the configuration presented a New-York that I decided to wait another 6 months to buy my babie..., I just hope that the G5 will be ready for january.

jefhatfield
Oct 19, 2001, 07:29 AM
thanks for the info zaustin

joey j
Oct 19, 2001, 09:46 AM
jesus. so much crap and misinformation. and why the hell has this 4 NEGHITS? macrumors.com: DISPOSE NOW of the +/-hits system, it sucks almost as much ass as some of the posts in this story.

so anyway, to the meat:

dougie>What the heck is SpecInt 2000?

You've never heard of SPEC? HAHAHAH.

dougie> The register posts P4 tests as if I cared!

Well, genius, to WHAT do you plan to COMPARE the scores, hrmm?

dougie>If the G5 1.6 scores in the 1000's, what we want to know: What does the fastest G4 (867) score?!

If moto wasn't so fuxored with spec results, i could answer that. however... they do distribute some... see below somewhere, when i start on zach.

dougie>possibly significant compared to the Real World Barely faster dual 800 "improvement" over the dual 533?

um, you must be smoking crack.


zaustin>Strangely, Motorola doesn't have ANY of their chips in the SPECCPU2000 database

only intel do, the workstation _vendors_ are meant to submit specbenches, not the _proc_ vendors, intel is the odd one out, (and sun too i guess, but they do the sparc inhouse.)

zaustin>Motorola SPEC scores are available for the older benchmark SpecCPU95

Presumably http://e-www.motorola.com/collateral/PPCCPUSUMM.pdf


zaustin>Motorola's SPEC scores have NEVER been as high as Intel's.

You must be smoking crack as well. Try normalizing the specmarks in PPCCPUSUMM against the specint/fp95 results at spec.org. Or if that's too difficult just match the clockspeeds and see. Then come back and repeat yourself.


toofeu>Well what I really care is the shipping date of those babies.

AMEN!

toofeu> I just hope that the G5 will be ready for january.

yeah, **** yeah! finally, someone with grey matter adds their tuppence. the story looks pretty legit, the reg have decent sources.

GPTurismo
Oct 19, 2001, 10:54 AM
The only thing that will hold back the G5 is yeild numbers.

I think Jobs will announce in january, but do an available later deal as usual.

fragiledreams
Oct 19, 2001, 12:14 PM
Er.. looking beyond the G5... what about some other things about the new PowerMaC???
What do you think about AGPX8, USB2, DDRAM and Firewire 2...
Basicaly I only care about the AGPX8 AND the new memory, and since the G5 bus will be 400MHZ an AGPX8 should be perfect!!!!

gandalf55
Oct 19, 2001, 01:46 PM
G5... this is what i have been waiting for. And if it needs a fan, i certainly will not mind.

Someday we'll have qMacs (quantum) and won't life be cherry!

degs
Oct 19, 2001, 02:45 PM
>tests as if I cared! If the G5 1.6 scores in the 1000's, >what we want to know: What does the fastest G4 (867)

One of the things often forgotten is that PowerPC is a core spec and not a bus spec, and that's why you have to MIPS to extrapolate a performance, but that's misleading as well.
Even the SPEC specs couldn't give you a very good feel for the speed increase because the comparisons are fundimentally different.
For example, an IBM PowerPC with a wider bus and a slower clock rate would do few theMIPS than MOT version, but would have higher data through put.

Regardless of the standards, benchmarks of little practical use except in marketing. I could make a processor that just does NOPs at 10GHz, but it wouldn't be very useful. :)

IMA
Oct 19, 2001, 04:17 PM
People,

these SPEC2K scores are clearly a hoax. They are almost the same (or a little faster) than the just released IBM Power4 - faster than a1GHz alpha!!!

Forget them. They are too high by a factor of 2.

Sorry.

ThlayliTheFierce
Oct 19, 2001, 04:34 PM
I think we'll see the chips at least announced in January, if not shipping. Apple has lost ground speedwise to Intel, and they need to get it back. They have to be pushing this thing really hard, they need it bad. That said, I'd really like to see, at the very least, DDR ram. I still can't think of why they only run at 133 mhz. I'd never buy one if it didn't have DDR.

Luckster
Oct 19, 2001, 05:54 PM
One of the considerations for not using DDR was the cost and availibility. While ture, these generally aren't Apple virtues, they sometimes do try to minimize cost where possible. Supposedly, they reengineered the memory managing chip so that using DDR would only be slightly better then PC133, and thus does justify the additional cost. Also, you simply cannot get DDR ram in a large a block as you can PC133. For memory intensive applications, a slightly slower, but larger ram block is better than a faster, but smaller ram block. If you want justification, just at Intel. The Rambus ram was simply too expensive and not as readily available. The P4 now uses PC133.

dougie
Oct 19, 2001, 07:23 PM
"what the heck is spec..." in other words, will the G5 really give us a performance boost? Look at 2-pop tests of the dual 800 vs 533 with FCP, look @ Barefeats tests of PhotoShop...the new machine is typically 5-20% faster: whoopee! This is crack speed? Comparing it to Wintel is helpful if you're gonna switch platforms, but my guess is that Mac Rumors types are wondering when to upgrade, and the $ wasn't and still isn't worth it for the 800DP (it should have been a DP 867) over the 533 DP. As per replies in this thread, knowing what Spec is, seems a waste of time.

joey j
Oct 19, 2001, 10:41 PM
>"what the heck is spec..." in other words,

Then you should have said what you _meant_. duh.


>will the G5 really give us a performance boost?

Nah. What a stupid idea. YA THINK? Naw.


>Look at 2-pop tests of the dual 800 vs 533 with FCP, look @ Barefeats tests of PhotoShop...the new machine is typically 5-20% faster: whoopee! This is crack speed?

Now if I didn't know better, i'd say that these weren't G5 machines (which is understandable if you tend to amble in conversation aimlessly from one point to another). Second, the realworld results with FCP are irrelevant; it's up to adobe, apple etc to ship binaries (in os x, a .app could probably choose the optimized code segment automatically, at the price of a larger executable) optimized for increased pipeline lengths, extra ops units etc.


> Comparing it to Wintel is helpful if you're gonna switch platforms,

in which case you're probably an amateur (nothing wrong w/that, i have friends which photoshop on the side on their PCs, webdesign et al, but not serious work) or smoking crack.


> but my guess is that Mac Rumors types are wondering when to upgrade, and the $ wasn't and still isn't worth it

>for the 800DP (it should have been a DP 867)

you think apple wanted an 800dp? apple's hands are tied by moto's yields. go bitch to them.


> over the 533 DP. As per replies in this thread, knowing what Spec is, seems a waste of time.

agreed, a p4 getting scores comparable to an alpha is a load of toss.

namu
Nov 8, 2001, 07:00 AM
Just a wild guess: Motorola lets all this info leak out because they have a good product and want everyone to know it, especially Apple (rumors about dumping Motorola for AMD, and IBM making GHz G3s)

About the "can't be faster than a 1GHz Power 4!": wait and see.

jefhatfield
Nov 8, 2001, 08:28 AM
...but due to a lot of the pretty accurate predictions i have seen here and on other sites, it does seem like apple has a few leaks here and there and i sometimes wonder if maybe one or two of them are intentional to create a buzz...which never has been a bad thing in this business