[PeterKG]

reply to:

did you say the 12" aluminum out performs the 1 Ghz Titanium?



Not to give everything away, but, yeah, almost across the board. We're re-testing the 1GHz Titanium with 10.2.3 to see if that makes any difference, but the 12" running 10.2.3 currently scores much higher than the 1GHz running 10.2.1(?) in Speedmark. When we re-tested the 800MHz Titanium in 10.2.3 its scores didn't really improve, so I'm not expecting the 1GHz score to go up either.

Right now the 12" completes our 3D render test two minutes faster than the GHz Titanium; Photoshop is faster by 10 seconds. That's a big difference that I suppose can be explained by the DDR RAM and other improvements, but it's worth verifying with 10.2.3 on a GHz Titanium before we post the numbers to Macworld.com.

But hey, if true it really bodes well for this new Aluminum PowerBook architecture!

Jason Snell, Editor, Macworld

[skyman]

Originally posted by PeterKG:
reply to:

did you say the 12" aluminum out performs the 1 Ghz Titanium?



Not to give everything away, but, yeah, almost across the board. We're re-testing the 1GHz Titanium with 10.2.3 to see if that makes any difference, but the 12" running 10.2.3 currently scores much higher than the 1GHz running 10.2.1(?) in Speedmark. When we re-tested the 800MHz Titanium in 10.2.3 its scores didn't really improve, so I'm not expecting the 1GHz score to go up either.

Right now the 12" completes our 3D render test two minutes faster than the GHz Titanium; Photoshop is faster by 10 seconds. That's a big difference that I suppose can be explained by the DDR RAM and other improvements, but it's worth verifying with 10.2.3 on a GHz Titanium before we post the numbers to Macworld.com.

But hey, if true it really bodes well for this new Aluminum PowerBook architecture!

Jason Snell, Editor, Macworld

That is a bunch of

The 12" has no L3 cache and the 1GHz Ti does. Also, we are talking about 1GHz vs 867MHz. There is NO way the 12" PB is faster than the 15" PB.

[seanyepez]

It feels as if the 15-inch PowerBook's visual effects are much smoother. In my experience, the 12-inch PowerBook is about as fast as my old 500-megahertz PowerBook G4.

[iBorg]

LOL!!!

Each individual spec of the 12" Albook is exceeded, or at least met, by the 15" Ti, and as an aggregate, the 15" blows the 12" away!

If "Speedmark" shows the 12" winning, all it means is that the test is worthless.


LOL!!!



iBorg

[sonoronos]

There is a way that this is possible. This has to do with the use of 266 mhz DDR RAM on the smaller powerbook, as opposed to 133 mhz standard SDRAM on the 15" book.

One big secret that us Computer Architecture guys know is that memory bandwidth and memory speed are the limiting factor for modern processor performance. This is the entire reason why we have caches at all. Therefore, while the 12" powerbook has 7/8th the performance of the gigabook, it also utilizes a memory system with 4x the maximum throughput of the gigabook.

An L3 cache makes a little bit of a difference, but truth be told, on regular matrix traversal benchmarks (such as 3d rendering) caches get trashed a lot (relatively speaking) and therefore memory bandwidth matters.

Question: You say the 12" book was faster by 2 minutes, but how long was the total rendering time? 5 minutes? 50 minutes? 500 minutes? It really makes a difference.

[John123]

Originally posted by sonoronos:
There is a way that this is possible. This has to do with the use of 266 mhz DDR RAM on the smaller powerbook, as opposed to 133 mhz standard SDRAM on the 15" book.

One big secret that us Computer Architecture guys know is that memory bandwidth and memory speed are the limiting factor for modern processor performance. This is the entire reason why we have caches at all. Therefore, while the 12" powerbook has 7/8th the performance of the gigabook, it also utilizes a memory system with 4x the maximum throughput of the gigabook.

An L3 cache makes a little bit of a difference, but truth be told, on regular matrix traversal benchmarks (such as 3d rendering) caches get trashed a lot (relatively speaking) and therefore memory bandwidth matters.

Not quite correct. Bandwidth matters a lot, yes, but you are confusing the theoretical benefits of DDR-RAM with their applied benefits in this case. The architecture of the PowerBook 12/17, as well as the Power Macs for that matter, does not take advantage of DDR-RAM. (The entire premise behind DDR-RAM, for those that don't know, is that it transmits data at both the peaks and the valleys of the cycles...by contrast, regular SDRAM works only at the peaks. Hence the acronym DDR, or "double data rate").

[skyman]

Originally posted by John123:
Not quite correct. Bandwidth matters a lot, yes, but you are confusing the theoretical benefits of DDR-RAM with their applied benefits in this case. The architecture of the PowerBook 12/17, as well as the Power Macs for that matter, does not take advantage of DDR-RAM. (The entire premise behind DDR-RAM, for those that don't know, is that it transmits data at both the peaks and the valleys of the cycles...by contrast, regular SDRAM works only at the peaks. Hence the acronym DDR, or "double data rate").

EXACTLY!

DDR-RAM makes absolutely NO difference in performance with the current G4 memory BUS limitations.

[cgreer00]

no way in hell the 12" is faster. no way. EVERY single benchmark I have seen pegs the 12" as significantly slower than the 1ghz.

[tkmd]

So if I understand you correctly when the 15.4 ver of the Al book comes out, probably with the GF440 in it, the current Ti 1 Ghz with the ati video card would actually be faster than the newer model that it got replaced by?

Assuming that the same marketing technique will be used in the next PB line as was with the latest imacs line - cripple the lower end without any upgrade options and push all the go-go goodies to the more $ top-end...

[StiZeven]

Originally posted by seanyepez:
It feels as if the 15-inch PowerBook's visual effects are much smoother. In my experience, the 12-inch PowerBook is about as fast as my old 500-megahertz PowerBook G4.

That's odd since the 12" PowerBook is 367MHz faster, has a faster HD, and a presumably faster/better memory card than your 'old 500MHz PoweBook G4'.

I've played with 3 12" PowerBooks at different locations and they've all felt almost as snappy as my 1GHz TiBook. The difference was very slight (as I expected since it's only 133MHz slower). Real time usage may tell another story - but lets not get carried away.

Anyway, back on topic, are these guys who are posting this info from Macworld? Do they work there, write for them or so on? What's the deal and how are they coming up with these figures?

[Eug]

Yeah, where is this posted. If not a fake post, then that test is seriously worthless.

[PeterKG]

It's not a fake post. It is a question with an answer from Jason at Macworld. I lifted it from the Macworld forums.
Obviously Macworld is testing the new powerbook. I will look forward to the results.

Originally posted by Eug:
Yeah, where is this posted. If not a fake post, then that test is seriously worthless.

[sonoronos]

Even if Apple's proprietary memory bus does not support DDR, its memory bus still operates at twice the rate of the original TiBook. This compared to a slower clock speed by 1/8 means that in the long run the slower CPU could (and would) defeat a slightly faster CPU with half the speed of memory.

If you still doubt me, read any conference paper by David Patterson in the last 6 years.

-ed

[John123]

Originally posted by sonoronos:
Even if Apple's proprietary memory bus does not support DDR, its memory bus still operates at twice the rate of the original TiBook. This compared to a slower clock speed by 1/8 means that in the long run the slower CPU could (and would) defeat a slightly faster CPU with half the speed of memory.

If you still doubt me, read any conference paper by David Patterson in the last 6 years.

-ed

Soronos, semantic discussions about memory aside, we already saw the results of DDR-RAM with the Power Macs. When they got DDR-RAM, their speeds didn't really improve. In some cases there was marginal improvement, and in most, there was none.

If you still doubt me, read any of the benchmarks done in the last 1 year.

[sonoronos]

John,

Thanks. Actually, I was going to delete that post, but the board didn't let me. I just realized after I posted that really I was arguing with myself. (It wasn't like anyone was disagreeing with me. I realized that)

The speed at which you replied was amazing! Right after I posted...wow.

[John123]

Originally posted by sonoronos:
John,

Thanks. Actually, I was going to delete that post, but the board didn't let me. I just realized after I posted that really I was arguing with myself. (It wasn't like anyone was disagreeing with me. I realized that)

The speed at which you replied was amazing! Right after I posted...wow.

What's that song by Beck -- loser?

Yeah, that's me.

[ymmit]

the mentioned thread from Macworld forum: <http://www.macworld.com/forums/ubbth...rue#Post112906>

[PeterKG]

Here is another post at the Macworld forums by Jason........


In reply to:

Sorry to disagree with your words, especially that you are Macworld editor but I don't believe DDR SDRAM has anything to do with the increased overall speed.
As we could see on page 75 of November 2002 Macworld, there is almost no difference in CPU-centered benchmarks between systems with SDRAM and DDR SDRAM.



Ah yes, but there we're talking about desktop models. These are totally different systems, with a new architecture, and it may be (just speculating here) that this architecture is more efficient and takes better advantage of the DDR SDRAM. Or perhaps it's not the SDRAM at all, but some other factor in the new architecture that's affecting our scores.

I'm not going to join the "DDR is useless" debate, since that's been fought before and I personally don't agree with those who claim that the Power Macs can't take advantage of the DDR SDRAM.

As for XBench, we don't use it here at Macworld. We're interested in taking a look at what it does, when we have the time and resources, but for our testing we have a serious of tests that we've standardized on that let us test real-world uses with shipping apps and come up with a judgment of real-world performance gains. This is not to say that XBench might be useful -- there's no way for me to endorse it or attack it right now since I haven't analyzed what it does and how it works -- but it's a synthetic test (a bit like the old MacBench was) and therefore may or may not map to real-world performance.

Look for our lab data in the next couple of days. We'll see if the numbers hold up when we re-test a GHz Titanium with the new OS version.

-jason

Jason Snell, Editor, Macworld

[proux]

Well at least there will be no more doubt when the 17'' will get benchmarked vs the current 15''.

[John123]

Originally I ripped into Macworld...but for now, I'll suspend the verdict until I see final results.

[Eug]

(It wasn't like anyone was disagreeing with me. I realized that)

I disagree with you.

A PowerMac CAN use the increased bandwidth that DDR offers. However, what can use it is essentially everything else EXCEPT the CPU (eg. hard drives, video cards, etc.). The G4 7455 CPU is simply incapable of it no matter how you design the motherboard. AFAIK, the 12" PowerBook is the 7455.

The only way around this is to design a new CPU. Indeed, the G4 7457-RM is supposedly capable of this, but unfortunately, that chip doesn't yet exist.

So anything that the CPU grabs from memory is going to be single-pumped, even if there is double-pumped DDR memory there. And thus, any CPU-centric test is basically going to be nearly the same speed regardless if DDR or SDR used. This is NOT true for PCs, because PC CPUs are specifically designed to take advantage of the increased bandwidth.

Now, it's quite possible that the new architecture IS faster than the TiBooks. However, the contention that an L3 cache-less 867 CPU is faster than a 15% faster MHz chip with 1 MB L3 makes no sense, unless the test is specifically designed to favour the DDR architecture.

[sonoronos]

Eug,

The 12.1" Powerbook clocks the memory twice as fast as the 15" powerbook.

Bottom line - it CAN be faster, forget the DDR!

My original question still stands. How long was the total rendering time?

[John123]

Originally posted by sonoronos:
Eug,

The 12.1" Powerbook clocks the memory twice as fast as the 15" powerbook.

Bottom line - it CAN be faster, forget the DDR!

My original question still stands. How long was the total rendering time?

Haha, dude, you can't "forget the DDR" -- that *IS* why it "clocks the memory twice as fast."

[sonoronos]

John,

I'm sorry, but 266 mhz memory is twice as fast 133 mhz memory.

[NeoEsper]

Doesn't the 12inch powerbook only have 233mhz ram, while the 17inch has the 266mhz??

[jamez bond]

I'll second sonoronos...how long was total rendering tmie, and also, what app was this?
Have you used 3d apps a lot on this machine, and if so, how is the overall performance?

[dialo]

It's not faster. Every other benchmark done on it shows this, including the one done by torifile in the "i ran logic..." thread here in the powerbook forum. Don't trust this stuff from Macworld. In fact, don't ever trust anything from that magazine. Macworld's revenues come directly from making apple look good.

[Eug]

Originally posted by sonoronos:
Eug,

The 12.1" Powerbook clocks the memory twice as fast as the 15" powerbook.

Bottom line - it CAN be faster, forget the DDR!

You just don't get it do you? DDR is good, but it's not magic. It simply won't do much if the system hasn't been designed to make use of it.

The bus clock is actually 133 MHz, but it's DDR memory that allows it to be double-pumped to 266 MHz.

However, the 7455 G4 CPU's design is NOT capable of using a DDR bus at all. Any access to the memory by the CPU will always be at single-pumped SDR speeds. This is a design limitation of the CPU and pairing DDR with it isn't going to make this design limitation disappear.

Thus, any speed benefit from DDR must come from actions that are independent of the CPU. In real life apps (except for very specfic ones), this means that DDR memory in a machine with a non-DDR capable CPU (like the 7455 G4) will mean at best a marginal (a couple of percent) speed increase.

Why do you think people have wanted to have the 7457-RM so badly? Well, not only is it faster, but it is designed with DDR in mind. ie. It can actually use double-pumped memory, whereas the 7455 cannot. Having a CPU capable of DDR is where the big speed increase will come.

Go read the PowerMac threads if you want additional info.

[John123]

Originally posted by sonoronos:
John,

I'm sorry, but 266 mhz memory is twice as fast 133 mhz memory.

On a 133 Mhz bus.

D....D.....R......

[sonoronos]

If what you are saying is that the 12" Powerbook cannot use 266 mhz SDRAM because its system bus is limited to 133 mhz, then I can understand that. However, what does DDR have to do with anything? Forget DDR please!

If indeed Apple maintains a 133 mhz system bus that impedes the potential performance of the 266 mhz RAM, then why even use 266 mhz RAM in the first place? Why not just use 133 mhz DIMMs?

BTW:

266 mhz DDR SDRAM - 4.256 GB/s
266 mhz SDRAM - 2.128 GB/s <= Powerbook G4 12"
133 mhz DDR SDRAM - 2.128 GB/s
133 mhz SDRAM - 1.064 GB/s <= Powerbook G4 15"

I don't know how else to explain this to you.

[AlbertWu]

the 12 inch is faster than my old 550, and that's all i care about

[Pierre B.]

Originally posted by Eug:


Thus, any speed benefit from DDR must come from actions that are independent of the CPU. In real life apps (except for very specfic ones), this means that DDR memory in a machine with a non-DDR capable CPU (like the 7455 G4) will mean at best a marginal (a couple of percent) speed increase.

Just because I think I understand what is said here about DDR RAM, I don't understand why Apple uses this type of memory in the power-macs-books. Is there some sort of marketing reasons?

[Hozie]

Now I'm no expert on these matters, but here's how I understand the DDR-SDR difference:

SDR can only communicate with the processor once per cycle. DDR can communicate twice per cycle, given the same bandwith. Therefore, if the motherboard architecture allows for a 133Mhz bus, the DDR memory will pass twice the amount of information that the SDR would (Effectively at an artificial rate of 266Mhz). In fact, in the PC world (where I come from), the front side bus for so-called 266Mhz DDR is actually 133Mhz*2Cycles. So, according to Eug's post, the only reason a Mac wouldn't make use of this is due to the CPU. However, I'll venture out to say that the main use of Ram is that the CPU can access data more quickly. Therefore, if the CPU can't really tap into the additional bandwith that DDR provides, a machine with 133Mhz SDR should perform almost identically. Thus, DDR in the current crop of Macs is useless. Q.E.D.

[mousehouse]

the DDR used in current powermac's should e far from useless...

fact 1:
CPU, memory and several other devices in the system are connected to the controller chip on the system board.
fact 2:
CPU-bus and memory-bus are two different busses.

1) as the bandwidth available to the CPU as not increased it will not be able to perform calculations faster than an equally (or higher) clocked CPU of the same type, when the calculations involved saturate the CPU-bus. for that matter, with calculations that saturate the CPU-bus will perform faster on a system with a large L3 cache, as this offloads the CPU-bus (and therefore also offloads the memory bus).

2) calculations that do not saturate the CPU bus will run faster on an equally clocked CPU with L3 cache.

3) the CPU is not the only device using the memory bus. all DMA capable devices will be able to utilize the "spare" capacity that the DDR memory-bus has to offer compared to the SDR capacity.

Therefore, the 867 w/DDR _could_ be faster in real life performance compared to a 1G/SDR when a combination of calculations and DMA capable memory accesses occur in parralel. But because of the 1M L3 and 15% faster clock rate the balance will probably tip to the 1G/SDR in most cases.

(tip: get CAS2 mem for your SDR Mac )

[Eug]

Originally posted by sonoronos:
If what you are saying is that the 12" Powerbook cannot use 266 mhz SDRAM because its system bus is limited to 133 mhz, then I can understand that. However, what does DDR have to do with anything? Forget DDR please!

If indeed Apple maintains a 133 mhz system bus that impedes the potential performance of the 266 mhz RAM, then why even use 266 mhz RAM in the first place? Why not just use 133 mhz DIMMs?

BTW:

266 mhz DDR SDRAM - 4.256 GB/s
266 mhz SDRAM - 2.128 GB/s <= Powerbook G4 12"
133 mhz DDR SDRAM - 2.128 GB/s
133 mhz SDRAM - 1.064 GB/s <= Powerbook G4 15"

I don't know how else to explain this to you.

As John already said, the 266 MHz RAM used is 266 only because of DDR.

By your definition:

266 mhz DDR SDRAM - Does not exist
266 mhz SDR SDRAM - Does not exist
133 mhz DDR SDRAM - What the PB 12" uses, but it functions at half the speed with CPU dependent functions
133 mhz SDR SDRAM - What the PB 15" uses.

Then why call it 266 MHz DDR? Because 133xDDR = 266. It might be a bit misleading to some, but that's the way it is.

Thus, the RAM in the 12" AluBook is 133x2. However, with CPU dependent functions, it's actually running 133x1, exactly the same speed as the SDR SDRAM in the TiBook.

Now, as mousehouse has said, there are many parts of the computer that can make use of the extra bandwidth. However, in benches it has been shown time and again that for system performance, the overall speed increase from DDR will be marginal if the CPU can't make use of it. Indeed, having a faster CPU will be much more important most of the time, except for certain specific applications.

Why use DDR in the first place?

1) Marketing. Obviously, judging on the posts here it has worked on many people.
2) It indeed does have some benefit. The benefit would be much more on something like the Xserve I'm guessing than a PowerBook though. But even then it's not going to be a humungous benefit.
3) Perhaps Apple's designers developed this motherboard partially in anticipation of the 7457-RM chips. If such a chip appeared they could just modify the current motherboard a bit and then drop in the 7457-RM immediately. Instant significant boost. Too bad the 7457-RM is vapourware so far.

[wulf]

Just to put people's minds at rest, Jason Snell has retracted the '12" is faster' claim (see here).

Just a follow-up for all concerned. We've done some re-testing and it turns out that, as most of us suspected, the 12" is _not_ really faster than the 15" 1GHz model.

[John123]

Eug, your explanation was eloquent and flawless. Kudos!

[Eug]

Originally posted by wulf:
Just to put people's minds at rest, Jason Snell has retracted the '12" is faster' claim (see here).

About time.

Just a follow-up for all concerned. We've done some re-testing and it turns out that, as most of us suspected, the 12" is _not_ really faster than the 15" 1GHz model.

I'm surprised he made such a statement in the first place.

[John123]

Well when I called the results into question over there, he got really mad at me! Granted, I guess my tone and language wasn't the best in the world, but:

http://www.macworld.com/forums/ubbth...37&Search=true

(Scroll toward the bottom of the thread)

It's kind of funny to me that in Jason's comment about the new results for the 12", he says:
Just a follow-up for all concerned. We've done some re-testing and it turns out that, as most of us suspected, the 12" is _not_ really faster than the 15" 1GHz model.
Whew! All is right with the world. The laws of physics still apply.
More details when our Lab has finalized the results.
-jason


I think the Macworld Editor hates me!

[wataru]

I can say definitively that both the 12" and 15" PowerBooks are exactly the same speed. In fact they're the same speed as every other computer out there, including PCs, regardless of the CPU.

Benchmarks: (Computer / time to fall 30ft)
12" PowerBook G4: 1.37s
15" PowerBook G4: 1.37s
Dell Inspiron: 1.37s
Sony Vaio: 1.37s

30 / 1.37 = 21.9ft/s

Thus all the computers had the same speed, 21.9ft/s.

They all have the same acceleration as well, which I measured to about 32ft/s.

[Eug]

Originally posted by wataru:
I can say definitively that both the 12" and 15" PowerBooks are exactly the same speed. In fact they're the same speed as every other computer out there, including PCs, regardless of the CPU.

Benchmarks: (Computer / time to fall 30ft)
12" PowerBook G4: 1.37s
15" PowerBook G4: 1.37s
Dell Inspiron: 1.37s
Sony Vaio: 1.37s

30 / 1.37 = 21.9ft/s

Thus all the computers had the same speed, 21.9ft/s.

They all have the same acceleration as well, which I measured to about 32ft/s.

Your test is incorrect.

Your test fails to include air resistance. This would depend on the size of the unit and whether its open or not, etc.

And with the 15" TiBook's hyperactive fan, it's liable to take off and go up like a helicopter.

[urrl5201]

Better watch your p's and q's over at MacWorld. WHAT THE???
THEY KILLED KENNY!!!

http://www.macworld.com/forums/ubbth...=&fpart=2&vc=1

And I don't mean the one from South Park!

[wataru]

Originally posted by Eug:
Your test is incorrect.

Your test fails to include air resistance. This would depend on the size of the unit and whether its open or not, etc.

And with the 15" TiBook's hyperactive fan, it's liable to take off and go up like a helicopter.

Air resistance was negligible for all machines tested.

The machines were asleep, so the fans were not on.

My tests are correct.

[fisherKing]

Originally posted by wataru:
Air resistance was negligible for all machines tested.

The machines were asleep, so the fans were not on.

My tests are correct.

keep in mind, tho, that if you run these tests within an hour after the laptops have been on, the 12" will of course be slower as heat rises...

[icruise]

Originally posted by urrl5201:
Better watch your p's and q's over at MacWorld. WHAT THE???
THEY KILLED KENNY!!!

http://www.macworld.com/forums/ubbth...=&fpart=2&vc=1

And I don't mean the one from South Park!

I for one am very glad of that -- he was derailing every thread he participated in.

[kcmac]

I agree iCruise. Some people just get a little too cantankerous. (Phoenetics are great huh?)

[John123]

Good grief.

Okay, Kenny's post was a little, um, pushy maybe. Assuming, probably.

But did it warrant him getting booted?!?!?

[seanyepez]

Originally posted by StiZeven:
That's odd since the 12" PowerBook is 367MHz faster, has a faster HD, and a presumably faster/better memory card than your 'old 500MHz PoweBook G4'.

I've played with 3 12" PowerBooks at different locations and they've all felt almost as snappy as my 1GHz TiBook. The difference was very slight (as I expected since it's only 133MHz slower). Real time usage may tell another story - but lets not get carried away.

Anyway, back on topic, are these guys who are posting this info from Macworld? Do they work there, write for them or so on? What's the deal and how are they coming up with these figures?

I'm sure that my 1-gigahertz TiBook is significantly faster than my mother's 867-megahertz, 12-inch PowerBook.

The 500-megahertz PowerBook G4 included a full megabyte of L2 cache running at 250 megahertz. The 867-megahertz PowerBook has only 256 kilobytes of L2 cache running at full processor speed. While visual effects might not be as smooth on the 500-megahertz PowerBook, the two machines feel very similar in terms of real-world performance.

[MilesWho]

Sean - I thought you were returning your powerbook for an ibook. What happened?

[icruise]

Originally posted by John123:
Good grief.

Okay, Kenny's post was a little, um, pushy maybe. Assuming, probably.

But did it warrant him getting booted?!?!?

Had you read the other threads he was in? Really a very obnoxious person.

And I'm usually a nice guy!