MacBidouille Rumors Followup

[NicoMan]

Re: Timeframe:

Originally posted by Kai
Even though quite a few things outside MacBd hint towards Apple using the P970 (suspicious quotes from IBM, Altivec, certain logic) remember that we still don't have real PROOF that Apple will be using it!

I for my part certainly hope they will though!...

Absolutely. But in a way it wouldn't make sense for Apple to announce new chips for their Macs (i.e. making a statement about the 970) and not having the stores ready to take orders on those machines. The policy is to say: here is our new processor, you can place order right now on the Apple store. If they didn't have anything ready to accept order, they would be in trouble. Customers wouldn't want to buy machines soon to be obsolete and wouldn't be able to buy the new stuff... hence big problem.

Like you, I'm really hoping we will get the 970 announced at WWDC or soon after, so that we don't have to wait too long until those puppies ship (I know, there is a good chance we will have to wait until Panther is available anyway)...

NicoMan

 

[anonmac]

ppc970 benchmarks...the scoop?

as for photoshop....

who can say what the benefit of 2x450 mhz busses are? (most say 900mhz bus, but it is actually 2x450, each unidirectional.) i would not want to be the one who said 'there is no way this is possible because of sp vs. mp processors test in photoshop' - the bus speeds, at 6.4GB/sec, are about 3-4x what the fsb is in the g4. so the mobo optimizations could be such that the 'sp' 970 COULD be 2x as high as the 'sp' G4. now i am not saying these ARE accurate, just that from my reading on the 970, i am open to the possibility that they MAY be accurate.

all you 'chip-producing insiders' who are quick to discount these 'rumours' as false, and made-up.... i am certainly glad i am not that closed minded. or perhaps you have found articles on the power4, power4+, power5 and ppc970 that i have not. if so, please reference so i can read what has 'put you in the know' because i have not only read all the press releases, pdf presentations by sandon, et al., (clayton and wang have EXCELLENT google postings) but followed lots of technical forum info. (from mac, linux, and ibm forums) as well. if i have missed something, please alert me- seriously. thanks.

 

[Kai]

Anonmac:

I understand you want these numbers to be true. We all do. However, they're not.

We didn't mean the comparison G4/P970 but the comparison Single-P970 vs Dual P970!
Same Bus, same CPU, so there!..

And no, i don't have to be a chip designer myself to see that these Benchmark-"Results" are fake and made up... It's really just simple logic and a bit of common background knowledge, nothing more! It's got nothing to do with a closed mind, just with logic!..

P970 will be a great CPU, no doubt, but these are not true benchmark-results of it, that's all we're saying...

In case you haven't read the Arstechnica-Article i quoted earlier, i would really recommend that if you're such a P970-fan! ;-)

 

[ktlx]

Re: ppc970 benchmarks...the scoop?

Originally posted by anonmac
as for photoshop....

who can say what the benefit of 2x450 mhz busses are? (most say 900mhz bus, but it is actually 2x450, each unidirectional.) i would not want to be the one who said 'there is no way this is possible because of sp vs. mp processors test in photoshop' - the bus speeds, at 6.4GB/sec, are about 3-4x what the fsb is in the g4. so the mobo optimizations could be such that the 'sp' 970 COULD be 2x as high as the 'sp' G4. now i am not saying these ARE accurate, just that from my reading on the 970, i am open to the possibility that they MAY be accurate.

Your argument is difficult to make.

The numbers from BareFeats that were posted were for the SP Action filters. These numbers have shown that a second processor does not help performance, regardless of memory I/O available. However, the MacB numbers posted show that the difference between a 1.4Ghz PowerPC 970 and a dual 1.8Ghz PowerPC 970 is over a 100% increase in speed. That implies that either some other tests were run than the SP Action filters or that a new version of Photoshop was used that allows these filters to benefit by a second processor. You cannot get a 100% performance increase going from 1.4Ghz to 1.8Ghz on a single processor alone (regardless of the application).

In either case, the comparison of the numbers between the G4, P4 and PowerPC 970 is invalid. I am not saying the PowerPC 970 numbers are fake, because I have no idea. I am saying to achieve those results, you need to change a variable that renders the comparison useless. Either different tests were run or they were run on a different version of the software. The numbers may all be correct but useless for comparison because they were gathered differently.

Unless Apple is implementing some exotic memory access that no one else is using, the dual processors are not able to get any more memory I/O bandwidth than a single processor. The processors would still be memory I/O starved--just at a higher rate. The highest bandwidth implemented today is dual channel DDR400 which provides around 6.4GB/s. I am sure they could go to quad channel, I guess, but then you are requiring a minimum configuration of 1GB (four DIMMs at 256MB) memory just to use a system. They could go with something like DDRII or the new memory technologies ATI and nVidia are using for graphics cards but then how do you buy memory upgrades?

 

[anonmac]

ok kai....you da man! so there!

('so there'-- kind of childish huh? i thought this was an intellectual discussion)

how about if the 'low-end' ppc 970 uses single channel ram and the dual processor uses two banks, as was said? that would account for some of the differences, and keep the low-end price down.

WE DON'T KNOW.
what i am saying is to simply discount these tests as bogus on the information given is no more educated of a decision that to simply believe them.

there are tons of possibilities...what are the backside cache levels- is it a similar thing to todays, where the high-end has 2x what the low end has (i.e. so really 4x in a dual?)? do the dual processors have 4x450 busses? do the duals share cache memory, like the power 4? what if all of these are true?

i don't know, and don't claim to. but they could certainly have huge advantages- especially based on the limited amount of info we have been given regarding the tests.

yup, read the ars article -- i also read the 'rebuttal' at david e's igeek.com, (it fleshed out a lot of things left unresolved or ambiguous by ars tech) and the recent 64-bit article at ars,. (ars is rarely as unbiased as they 'purport' to be, however still a good source to cull.)

 

[ktlx]

Re: ok kai....you da man! so there!

Originally posted by anonmac
how about if the 'low-end' ppc 970 uses single channel ram and the dual processor uses two banks, as was said? that would account for some of the differences, and keep the low-end price down.

While that is theoretically possible, having two different types of memory controllers for single vs. dual processor would require a lot of engineering and testing expense, for very little end user savings. Apple would have to design, engineer, manufacture and test two different memory controllers when the difference in price for the complete implementation is less than $100 (chipset, memory slot and additional DIMM). Not even Intel, which has much higher volumes, is doing new work on single channel memory controllers for desktops and workstations. They are just using what has already been designed, engineered and tested.

Since the 970 is completely new, Apple would have to develop two new ones (one single channel and one dual channel) from scratch. That combined with their relatively small volume, make the economics uncertain at best. I find it hard to believe that Apple would burn profit on crippling the memory I/O on single processor systems just because they could.

there are tons of possibilities...what are the backside cache levels- is it a similar thing to todays, where the high-end has 2x what the low end has (i.e. so really 4x in a dual?)? do the dual processors have 4x450 busses? do the duals share cache memory, like the power 4? what if all of these are true?

According to everything I have read, the PowerPC 970 is not equipped to support L3 cache. It only supports L1 and L2 like the G3s, Intels and AMDs and the sizes are all the same across the product line.

The problem I see with people wanting the comparisons to be valid so badly is the arguments have to resort to economically or logically infeasible situations in order to claim the comparison is valid. The far simpler conclusions either are they are fake or that the PowerPC 970 numbers are valid but use tests or software that makes comparison to the BareFeats numbers useless.

For me, I don't care which one is true because either case renders the numbers useless. I would expect to find a set of Adobe Photoshop tests that would nearly double in speed between a single 1.4Ghz and a dual 1.8Ghz so the comparisons within the PowerPC 970 family don't give me any new information. If the PowerPC 970 tests are not made up, then they must use either different filters or different software in order to get those results. In that case, I still cannot use the comparisons because the BareFeats numbers were gathered differently.

 

[anonmac]

ram-dual vs. single channel...and the debate!

as i undestand it, the ram architecture can be EITHER dual or single channel, but the dual requires interleaving. so a 'better' memory controller- if that were the only one developed, it would certainly still be capable of supporting the single channel architecture. that doesn't cost too much more does it? (kind of like the USB 2.0 chipset rumors in new macs-- they probably just couldn't get 1.1 controllers anymore (if it isn't true today- it will be. 2.0 sets will be just as cheap as 1.1-if 1.1 is available at all. why not use it?)

there are all kinds of things outside of what we know at this point that could influence these. let me reiterate for those who havent read. I DO NOT SAY THESE ARE TRUE- JUST THAT THERE IS THE POSSIBILITY. probability- no. but i always keep an open mind until someone presents me with ironclad proof ('the sun is the center of the universe' was a scientific fact until 1600's) that it could not be. so i say, argue on my friends perhaps we will come to a meeting of the minds, and i like the debate. thanks.

 

[macrumors12345]

Re: ok kai....you da man! so there!

Originally posted by anonmac
how about if the 'low-end' ppc 970 uses single channel ram and the dual processor uses two banks, as was said? that would account for some of the differences, and keep the low-end price down.

Still no go. If the test were limited by the bandwidth to main memory (unlikely...the MP and Altivec stuff is more likely to be limited than the SP stuff because a lot of data is being moved), then the P4 would have absolutely slaughtered the DP G4, since it more than triple the memory bandwidth. But it didn't (it was faster, but not THAT much faster). So your explanation does not work, even if you assume that the high end is using dual channel and the low end isn't (unlikely).

There is only one way to explain these benchmarks: the 970 results are faked, OR the 970 is running an ENTIRELY different test suite than the G4/P4. Either way, these results are completely worthless.

 

[fpnc]

Reality check only suggests problems

The benchmarks and PPC970 timeline that MacBidouille has publish are highly suspect for one simple and unavoidable reality. There will never again be a single-processor Macintosh that will be able to outperform a top-of-the-line WINTEL/AMD PC (over a broad range of tasks). In fact, it will be a miraculous accomplishment even if they come anywhere near to parity. About the best we can hope for is that on any given task a dual-processor PPC970 will be somewhat faster than any single-processor PC. And by "faster," I'm talking about the ability to complete a broad range of processor-intensive tasks in a given amount of time (i.e. it's not clock speed).

Interestingly, I think this "reality" alone will pretty much determine how the PPC970-based Macs will perform in relationship to PCs, and it also places a limit on when they will be introduced (because Apple could not produce a single-processor PPC970 that outperforms any existing P4, eventually there will be a 970 that outperforms a 3.2 GHz P4, but by that time the PCs will be faster still). The reason I say this is that it seems highly unlikely that Apple will ever be able to overcome the massive advantages and lead that PCs have in hardware and software R&D and investment. So, we all need a reality check here, whenever the PPC970-based Macs appear they will not be faster than any then existing PC. These new Macs will certainly introduce a new phase in the processor performance "wars," but in my opinion there is no way that Apple is going to leapfrog over the entire PC industry. I'm not saying that the next generation Macs are going to be disappointing, they just aren't going to be world beaters.

So, that's the "bad" news (I guess). The good news is that the PPC970 will move us much nearer to performance parity with PCs (let's face it, today, desktop Macs are pretty far behind). The only remaining questions are exactly when are we going to see these machines and how much will they cost? My guess, not very soon and they will probably be pretty expensive. Any outcome other than this would probably mean a highly compromised offering (i.e. just shoehorn the 970 into the existing PowerMac hardware architecture).

 

[anonmac]

fpnc....clueless?

the tests at barefeats (latest) show that for concurrent processing the mac already beats the pc.

run tasks in succession and the pc beats the mac.
(by taking a '13%' beating- when processor speed is over 100% 'faster'.)

these are not my conclusions- SEE barefeats for yourself. then tell me about how far behind the mac is. i have been very polite because most of the people posting have been well informed and articulate. you are neither. READ about what you are saying BEFORE you say it and you may seem a little smarter.

 

[ktlx]

Re: ram-dual vs. single channel...and the debate!

Originally posted by anonmac
as i undestand it, the ram architecture can be EITHER dual or single channel, but the dual requires interleaving. so a 'better' memory controller- if that were the only one developed, it would certainly still be capable of supporting the single channel architecture. that doesn't cost too much more does it?

Memory controllers do not work that way. You either have one that supports single channels (as in all SDR SDRAM and earlier and most DDR SDRAM implementations) or it supports two channels (as in the new DDR SDRAM and RDRAM implementations). The chips to handle the memory controller would have to be two different chips requiring two different manufacturing steps and two sets of system testing.

I am sure you could develop some sort of controller that senses whether it can run as a single channel or can run as dual channels (i.e. an even number and a matched set of DIMMs is inserted) but imagine how costly and complicated that would be. With 512MB PC2700 DIMMs going for $50, that would be a stupid route to take.

This is especially true since the rumors only have the PowerPC 970 going in PowerMacs at this time. The total system cost is high enough to support the additional price of dual channel DDR but the volumes are small enough to not justify multiple solutions.

Although it sounds like the discussion is moot at this point. New rumors from MacB say the motherboards use DDR400 with no mention of dual channels for anything. That would be too bad. It makes no sense having this great I/O link and then choking it down to DDR400.

But that may have been the price to get production started quickly. The jump from DDR333 to DDR400 is trivial compared to the jump from DDR333 to dual channel DDR400. None of the other motherboard features are large jumps either.

 

[fpnc]

Re: fpnc....clueless?

Originally posted by anonmac
the tests at barefeats (latest) show that for concurrent processing the mac already beats the pc.

run tasks in succession and the pc beats the mac.
(by taking a '13%' beating- when processor speed is over 100% 'faster'.)

these are not my conclusions- SEE barefeats for yourself. then tell me about how far behind the mac is. i have been very polite because most of the people posting have been well informed and articulate. you are neither. READ about what you are saying BEFORE you say it and you may seem a little smarter.

Well, anonmac , I think you actually need to read what I said. First, I said clearly and in several places that I was talking about performance in a broad range of tasks. Never did I mention Photoshop. Second, numerous, independent benchmarks have been published over the last six months that show that the Mac is dramatically behind the PC in performance, even in areas where the Mac was once considered strong (for example, in video production and, yes, even in Photoshop). The fact that a few isolated Photoshop actions are competitive with PCs doesn't stand very well against the fact that even Adobe now recommends PCs over PowerMacs when considering workflow (performance) under Photoshop.

I'm completely amazed that anyone would try to suggest that the current PowerMacs are faster than PCs. Yes, a few examples can be found where PowerMacs do well, but I was talking about __reality__ not a few marginal cases which have practically zero impact on the vast majority of computer tasks.

By the way, I am not a Mac hater or PC troll. I like Macs, I use Macs, I own Macs, but if you need the highest level of processor performance you aren't going to find it in the current Mac lineup, at least not in comparison to what is available on PCs.

 

[anonmac]

GO TO BAREFEATS.

DID YOU GO TO BAREFEATS OR NOT????????????????????

YOU LOOK FOOLISH.


*****NOT****** PHOTOSHOP RESULTS...CINEMA 4D, DATABASE, ETC.

 

[anonmac]

'INDEPENDANT TESTS'

the 'independant' tests you speak of are digitalvideoediting.coms.

i saw the site three times, and everytime a 3.066ghz hyperthreading ad came up.

they are VERY unbiased and 'independant' (not to mention the kickback given when someone buys a dell with a dvediting 'clickthrough' link.)

if you have others why not post them. ill refute you one by one, and explain to you in simple terms why the 'supposed' results are inaccurate.

lets start now
DVediting: tested after effects--- little boost from multiple processors. graphs were wrong. you ate it up cause thats what you believe. good for you.

do some reading instead of just posting all the time. i can back all my assertions up-- WITH LINKS if need be.can you?

again i say GO READ THE BAREFEATS article.

i use macs, solaris, linux, aix, etc. does this make me smart? (how does it relate to our 'discussion'?)

as for your invalidation of my argument without seeing the site it is based on, what else could i expect from one who posits as you do.

 

[ktlx]

Re: Re: fpnc....clueless?

Originally posted by fpnc
The fact that a few isolated Photoshop actions are competitive with PCs doesn't stand very well against the fact that even Adobe now recommends PCs over PowerMacs when considering workflow (performance) under Photoshop.

That was with Adobe Premiere and not Adobe Photoshop. Many people believe that was driven by nothing more than Premiere getting its butt kicked by FCP.

 

[fpnc]

ktlx, I think you may be correct, the pcpreferred link that Adobe published may have specifically mentioned Premiere. However, I still stand by general belief that for the majority of tasks (including Photoshop and video editing) a top-of-the-line PC will outperform any G4 that Apple currently offers.

anonmac , as far as benchmarks, I've read plenty both on the internet and in print. And I can tell you that the consensus for quite some time has indicated that the PCs lead, sometimes by huge margins, while in only a __few__ cases the Macs are close or maybe a bit ahead. I'm not going to take the time to locate these sources, many were in print, but I'm fully satisfied that PCs are faster in most situations. Does that mean that they are "better," no, not necessarily.

Have you noticed that it has been a very long time (as technology goes) since Steve Jobs has talked about the PPC's "Pentium toasting performance?" When was the last time you saw a live demo from Apple comparing a PC's performance against a Mac? And as I recall even Steve Jobs recently admitted that Apple had some work to do to bring performance parity back to the Mac platform.

And please, anonmac, no more talk about WINTEL conspiracies to make the Mac look bad. Steve Jobs' reality distortion field was fun for a while, but we should all be past that by now.

 

[macrumors12345]

Re: Re: ram-dual vs. single channel...and the debate!

Originally posted by ktlx

Although it sounds like the discussion is moot at this point. New rumors from MacB say the motherboards use DDR400 with no mention of dual channels for anything. That would be too bad. It makes no sense having this great I/O link and then choking it down to DDR400.

Well, given that MacB's benchmarks are fakes, I'm not sure I would put much credibility in anything else they post. Nevertheless, I wouldn't be shocked if the first generation of 970s didn't have dual channel DDR400....sort of like Apple released the Yikes G4 machines initially. Also, unless they use a bit of L3 cache as a buffer between the SDRAM and the 970's FSB, it's not clear that going to dual channel would even help, because dual channel is a bidirectional and the 970 has two unidirectional buses.

 

[ktlx]

Originally posted by fpnc
ktlx, I think you may be correct, the pcpreferred link that Adobe published may have specifically mentioned Premiere. However, I still stand by general belief that for the majority of tasks (including Photoshop and video editing) a top-of-the-line PC will outperform any G4 that Apple currently offers.

I don't disagree with you on that statement. In fact, my own experience confirms this. However that is not what you said. You said that Adobe is telling people the preferred system is a PC for a Photoshop workflow. I simply pointed out that you were mistaken--they said Premiere and not Photoshop. And I also pointed out that many believe Adobe took that stand only because Apple has FCP.

 

[ktlx]

Re: Re: Re: ram-dual vs. single channel...and the debate!

Originally posted by macrumors12345
Also, unless they use a bit of L3 cache as a buffer between the SDRAM and the 970's FSB, it's not clear that going to dual channel would even help, because dual channel is a bidirectional and the 970 has two unidirectional buses.

The PowerPC 970 does not support L3 cache according to IBM. That was one of the design tradeoffs to reduce power and cost.

The two unidirectional buses for CPU memory doesn't have anything to do with it. All SDRAM is bidirectional. That is the purpose of the system controller/north bridge. It converts the CPU's I/O signals into the appropriate memory signals. It will need to have some sort of buffer in order to arbitrate between CPU reads/writes and memory reads/writes but chip designers are able to handle that.

 

[3.1416]

Re: Reality check only suggests problems

Originally posted by fpnc
There will never again be a single-processor Macintosh that will be able to outperform a top-of-the-line WINTEL/AMD PC

What are you basing this on? "Never" is a pretty long time.

 

[macrumors12345]

Re: Re: Re: Re: ram-dual vs. single channel...and the debate!

Originally posted by ktlx
The two unidirectional buses for CPU memory doesn't have anything to do with it. All SDRAM is bidirectional. That is the purpose of the system controller/north bridge. It converts the CPU's I/O signals into the appropriate memory signals. It will need to have some sort of buffer in order to arbitrate between CPU reads/writes and memory reads/writes but chip designers are able to handle that.

Agreed. I was not implying that the PPC 970 itself would need L3 cache - simply that there needs to be some kind of buffer cache between the unidirectional buses and the bidirectional SDRAM.

 

[fpnc]

ktlx, I guess I was unclear in my response. Yes, you were right and I was wrong about the pcpreferred posting.

But, I'd like a clarification from you about your latest statement. You said, "I don't disagree with you on that statement. In fact, my own experience confirms this." Are you saying that in your hands PCs are faster for video and photo editing?

 

[BaghdadBob]

Re: Re: Reality check only suggests problems

...I posted but decided to mute myself on this subject. Don't care to get caught up in the swirl of blanket-statement, short-sighted, status quo ad infinitum, mac-bashing ignorance...

 

[macrumors12345]

Originally posted by Snowy_River
Didn't you, yourself, say that, although you are sure that these benchmarks are fake, you believe that they are right around where real benchmarks will be?

Yes, although I honestly doubt they will be THAT good. I would guess that initially the 970 will be a substantial improvement over the G4, but not quite as impressive as the SPEC scores would indicate (i.e. twice as fast as the G4 at a given clock speed). After all, it is an entirely new core (from the perspective of the Macintosh market), so the current code base is unlikely to run as efficiently on it as it could. But of course that will change as the code base changes to reflect the increasing number of 970 machines on the market.

Explained. (May be a BS explanation, but an explanation has been given.)

Yes. The problem is not that these results are technically impossible. It is that they cannot be explained without highly implausible explanations. While your explanations are technically possible, they are unlikely to be true. And the conjunction of the three events (and three respective explanations) is VERY unlikely to be true.

I don't have an explanation for this. Indeed, what seems most likely is that the tester used some other PhotoShop test, as the likelihood of him/her having the BareFeats proprietary SP test suite seems unlikely.

So then you admit that at the very least, many of these tests are completely worthless, as they are not even running the same test on both machines, no? And by extension the tester is clearly quite stupid (we both seem to agree on that...I mean, even assuming they are faked he is stupid because he didn't even bother to make convincing fakes!), which makes it a little strange that he actually has access to such advanced, pre-production machines...

There is a reasonable explanation that would give the same data as what we have that does not include the data being falsified.

Clearly you have a different definition of reasonable than I do!

If you wanted to truly prove to me that there was falsehood going on here, then find me a copy of Bryce6 beta, and show me that it isn't multithreaded.

Very funny. That could be pretty much impossible to do given that nobody has yet found any credible evidence that this mythical Bryce 6 beta exists (several people have searched Hotline, where the tester claimed it was available - there is no sign at all of a Bryce 6 beta).

For the record, I am a scientist....It is something that is sure to get a research paper shot down in review, and such a paper will never see publication.

Really? In what field? In the journals in my field (economics) we are are rarely able to amass such convincing empirical evidence against a particular hypothesis (though certainly not for lack of trying), but any hypothesis which had as much evidence going against it as these benchmarks do would certainly be dismissed by any academic who was not a complete crank.

 

[fpnc]

Re: Re: Reality check only suggests problems

Originally posted by 3.1416
What are you basing this on? "Never" is a pretty long time.

Yes, "Never" is a long time but I think it's essentially correct because as far as the computer industy is concerned WINTEL is going to be around longer than Apple will, and I don't see any chance of the PCs losing their performance advantages while they are still dominating the business. I'm not saying that Apple will be going out of business, or that Apple won't be able to produce compelling machines, but the single-processor performance advantage will always be tipped towards the PC.

Of course there are other categories of computing hardware and consumer devices, so there is plenty of room for Apple and other players and thus I'm not predicting doom for Apple.

Besides, this is just my opinion, only time will tell. But for anyone who cares, I'll meet you back here in about another year and we'll compare notes again.