[acadian]

Cant find anything on this yet.

[AssassyN]

The Americans PB's now lack L3 cache...it's strange b/c the UK site says something about including L3 cache, but it's NOT in the new American ones. Bad move Apple!

[cambro]

No kidding! But isn't cache design a Moto problem?

[crispinwilliams]

So what processor is in it.. the 7447 or 7457?

[BrunoBruin]

Unless I'm mistaken, the 7455 in the previous PowerBooks only supported 256MB of L2, so these are probably 7457 which supports 512. My question is, does the increased L2 make up for the lack of L3?

I suspect that the UK site that mentions L3 is incorrect.

[Eug]

Originally posted by cambro:
No kidding! But isn't cache design a Moto problem?

The L2 cache is on-chip. The L3 cache is not.

[gumby5647]

Originally posted by cambro:
No kidding! But isn't cache design a Moto problem?

no...that would be system bus speed.

[pbjudge]

Originally posted by BrunoBruin:
Unless I'm mistaken, the 7455 in the previous PowerBooks only supported 256MB of L2, so these are probably 7457 which supports 512. My question is, does the increased L2 make up for the lack of L3?

I suspect that the UK site that mentions L3 is incorrect.

I think you are right, the Japanese sight agrees with the American site - there is no L3 listed as a part of any of the new PB's.

Does this adversely affect the performance or have they made up for it somehow else?

[BrunoBruin]

Originally posted by Eug:
The L2 cache is on-chip. The L3 cache is not.

Do you think we are taking much of a performance hit here by losing the L3? Or does the increased L2 make up for it, or at least perform well enough to offset the expense of the L3?

We should also take into account the possible clock-for-clock improvements we might see in the 7457. More efficient processors are always a Good Thing.

[crispinwilliams]

I can not believe that the US sites would not state the existance of the L3 cache if it did exist.. so I think it is fair to conclude that there is no L3 cache. Maybe this was part of some compromise forced onto Apple by Motorola... i.e Motorola could make 7447 easier than 7457.. and Apple was forced to accept the 7447 to get something new to update the powerbook line. Certainly there must be a performance hit for losing the L3 .. otherwise why put it in at all.. Hopefully the enlarge L2 will hide the missing L3 somewhat..

[Eug]

Originally posted by BrunoBruin:
Do you think we are taking much of a performance hit here by losing the L3? Or does the increased L2 make up for it, or at least perform well enough to offset the expense of the L3?

We should also take into account the possible clock-for-clock improvements we might see in the 7457. More efficient processors are always a Good Thing.

A bigger L2 makes a huge difference. The doubling of the L2 on the iBooks made for quite a nice speed boost. I'm not sure if it completely compensatees for the loss of a 1 MB L3 or not, but it'd go a long way to doing so. That, with the extra 25% clock rate, would make for a fine upgrade over the previous line.

However, we still don't know if there is an L3 or not so we shouldn't jump to any conclusions yet.

[BrunoBruin]

Originally posted by Eug:
A bigger L2 makes a huge difference. The doubling of the L2 on the iBooks made for quite a nice speed boost.

That's true, I had forgotten about that. I am actually more excited about the new 167MHz bus in the 15; I think THAT's the biggest improvement architecture-wise.

Meh. L3 or not, I'm getting one anyway. I'm just trying to decide if I want to order online or wait until I can get to my friendly Apple Store.

[kikkoman]

With the processor speed bump, bigger L2 cache and no L3 cache, the new Powerbooks should deliver at least equal performance of the previous generation. There maybe a slight performance improvement but I'm being optomistic. That is my unscientific opinion.

[acadian]

Considering how we've been lambasted with the G4's non existent upgrade path, I think knocking out the L3 cache is somewhat of a step backwards.

[Eug]

Originally posted by kikkoman:
With the processor speed bump, bigger L2 cache and no L3 cache, the new Powerbooks should deliver at least equal performance of the previous generation. There maybe a slight performance improvement but I'm being optomistic. That is my unscientific opinion.

I disagree. I think there should be a noticeable speed boost, even without the L3 cache (although it'd be even more of a boost if there were an L3).

However, we still don't know if there is a cache or not.

BTW, if you're looking for OS 9 support remember the old TiBook is the only one that supports it. Otherwise the new 15" looks schweet!

[AssassyN]

Straight from www.pbzone.com...

All three models use the G4 7457 chip.

[Eug]

Originally posted by AssassyN:
Straight from www.pbzone.com...

All three models use the G4 7457 chip.

As far as I can tell, they're just guessing like the rest of us.

[acadian]

bigger L2 cache and no L3 cache, the new Powerbooks should deliver at least equal performance of the previous generation

I think the point here is that we HAD a L3 cache and now we dont (apparantly). I would be interested to find out whether or not there was some sort of dire technical limitation that prevents the incorporation of an L3 cache but I susspect that it is simply a cost compromise on Apple's part, which does not make too much sense considering the relative state of the G4 compared to the intel world. Remember that the 2nd gen G4 powerbook's L2 cache was knocked down to 256k from the first gen's 1 mb and as a result, the 550 model performed slower than the prev gen 500 model, even with the inreased 133 mhz bus.

[Ti X]

Well, simple math tells me 512 vs 256+1MB would make the earlier versions faster. Whether or not the cache is on chip or not. But when combined with the faster processors, I think the new models will be just a hair faster. Remember that just like all previous moto revisions, performance increases are overall very insignificant.

[Simon]

Originally posted by Ti X:
Well, simple math tells me 512 vs 256+1MB would make the earlier versions faster.

Th only problem is simple math is wong in this case.

On-die L2 cache is much faster than slow off-die 133MHz L3. No matter how much.

Clock per clock, the new models are going to be faster. Period.

[BrunoBruin]

Originally posted by acadian:
...as a result, the 550 model performed slower than the prev gen 500 model, even with the inreased 133 mhz bus.

The 550MHz model still had a 100MHz bus; only the 667 in that line had 133. The 667/800 (DVI) line both had 133.

[spalding12]

i like that reasoning
hope it's correct

greg

[Eug]

Originally posted by Simon:
Th only problem is simple math is wong in this case.

On-die L2 cache is much faster than slow off-die 133MHz L3. No matter how much.

Clock per clock, the new models are going to be faster. Period.

Yeah, the L2 cache runs at full clock speed (1.25 to 1.33 GHz in the 15 and 17" respectively), and has doubled in size in the new machines.

However, the L3 on the GHz TiBook ran at 5x133 = 667 MHz. I think the 17" would have been the same probably - 4x167 = 667 MHz. There would be no point in running the L3 cache at the same speed as the bus speed.

The new machines WILL be significantly faster than the old ones, but not exactly for the reasons you cite, and I'm not sure about the clock for clock claim either, although you may be right.

[Simon]

Originally posted by Eug:
The new machines WILL be significantly faster than the old ones, but not exactly for the reasons you cite, and I'm not sure about the clock for clock claim either, although you may be right.

They will be, I'm rather confident. We'll see.

[Shaddim]

Originally posted by pbjudge:
I think you are right, the Japanese sight agrees with the American site - there is no L3 listed as a part of any of the new PB's.

Does this adversely affect the performance or have they made up for it somehow else?

So, does the ability to use DDR to it's full potential and doubling the L2 make up for a lack of L3? Personally, I believe so.

[BrunoBruin]

Originally posted by Eug:
The new machines WILL be significantly faster than the old ones, but not exactly for the reasons you cite, and I'm not sure about the clock for clock claim either, although you may be right.

Oh hell, as long as it's faster than my 667, I don't care exactly how MUCH! The 1.25GHz is already CHEAPER than mine was a year ago, and when I bought it, it was the low-end model!

[Eug]

Originally posted by MacNStein:
So, does the ability to use DDR to it's full potential and doubling the L2 make up for a lack of L3? Personally, I believe so.

The DDR is a hack. The 7457 CPU cannot make use of DDR. Other things can use DDR, but in reality, it's the CPU's ability to use DDR that makes the difference.

In most benches (and in real-world testing), DDR on a G4 Mac makes little difference compared to SDR.

[PeterKG]

Can I take my 512MB DDR ram from my 12" Powerbook and use it in my new 15" Powerbook?

[Alpha-sphere]

The dutch apple site stated something about L3 also this morning but i can't seem too find it it now.

[ibulldog]

This is a serious issue for me. The lack of a L3 cache on my current 12 inch powerbook is what caused me to consider the updrade to the new machines. Frankly, the lack of the increased cache on my current machine causes issues with speed and responsiveness. I'm running the latest version of OSX and my ram is at 640.

I did order the 1 ghz version, today, but will cancel my order is the speed boost is just "marginal."

Any thoughts out there?

[kerl]

Have a look at the 15" benchmarks on macbidouille.com, they don't look to shabby. Apparently the missing L3 cache is not slowing down the system in any way: http://www.macbidouille.com/niouzcon...003-09-16#6636

e.g. the value of 40+ in the interface test: the "old" 12" has 22 something there. my 17" 1ghz has ~36

[RBM]

1Mb 3rd level cache was on the Dutch website for sure this afternoon. Now it shows 512Kb 2nd level instead - maybe a last minute decision to leave out the 3rd level cache?

[klinux]

Originally posted by RBM:
1Mb 3rd level cache was on the Dutch website for sure this afternoon. Now it shows 512Kb 2nd level instead - maybe a last minute decision to leave out the 3rd level cache?

Ditto. On UK site last night there was a mention of L3 and now it is gone too.

[Eug]

Have a look at the 15" benchmarks on macbidouille.com, they don't look to shabby. Apparently the missing L3 cache is not slowing down the system in any way: http://www.macbidouille.com/niouzco...2003-09-16#6636

e.g. the value of 40+ in the interface test: the "old" 12" has 22 something there. my 17" 1ghz has ~36

Xbench? Xbench is pretty useless for testing these kind of differences. Furthermore, the numbers have changed, with the 1.1.1 update.

This is a serious issue for me. The lack of a L3 cache on my current 12 inch powerbook is what caused me to consider the updrade to the new machines. Frankly, the lack of the increased cache on my current machine causes issues with speed and responsiveness. I'm running the latest version of OSX and my ram is at 640.

I did order the 1 ghz version, today, but will cancel my order is the speed boost is just "marginal."

Any thoughts out there?

Personally, I'd think it's rather pointless upgrading from an 867 to a 1 GHz, unless you absolutely needed DVI or something.

An upgrade to a 1.25 15" would be OK for a speed boost though (although a rather expensive upgrade for an incremental speedup).

[thunderous_funker]

The Register reports it is the PowerPC 7457, which accounts for the presence of 512K L2 cache, but doesn't explain the notable absence of L3 cache. The jury is out on how this will affect real world performance in certain applications that have typically relied on L3 cache.

<edited for corrections on cache data>

[alex_kac]

I can say for certain that the L2 cache more than makes up for the lack of a L3 cache. The CPU can use a L2 cache directly for all of its code. Making a 512k cache is a tremendous performance increase.

Remember, the G5 has no L3 cache either. An L3 cache is really a hack for CPUs that just don't want to add the bulk of transistors on the die.

The only good thing abot an L3 cache is that it can be turned off for power savings.

So a 256k L2 cache + 1MB L3 cache is good; but a 512k L2 cache is far better.

[beefstu01]

I'm saying that it's the 7447 actually. It's pretty much the exact same processor as the 7457, but it doesn't support L3 cache. If the processor in the 'books could handle L3 cache, I'd bet my left shoe that Apple would use that. I don't think it's much of a drag on the battery, and it could only bump the speed up, so it would be used if it could, but it can't so it won't.

[Simon]

Originally posted by beefstu01:
I'm saying that it's the 7447 actually.

It would really be grand if somebody could confirm this. Can't anybody just buy a new PowerBook, open it, rip it apart and tell me the numbers printed on top of the chip.

[nobitacu]

Originally posted by acadian:
Cant find anything on this yet.

No one know if they are in fact using the 7457 chip, everyone is just guessing at this point. As for the L3, no, no L3, that's a fact. People need to understand though that the speed of the L2 is in fact faster than the L3, so by increasing the size of the L2 and not having L3 might in fact be a better choice, faster choice.

Ming

[Simon]

Originally posted by nobitacu:
People need to understand though that the speed of the L2 is in fact faster than the L3, so by increasing the size of the L2 and not having L3 might in fact be a better choice, faster choice.

Definitely it's a better choice.

Proof can be found at this Barefeats page where you will see how a 15" 1GHz Ti compares to a 15" 1.25GHz Al. The Ti needs 300.3sec and the Al needs 227.2sec. The additonal 250MHz would only account for a time of about 240sec, so the Al actually gained more speed than what the clock would tell you.

Of course the reasons for this (bus, RAM, GPU, etc.) can be debated, but it is already a good sign that the new Al PowerBooks (clocked higher, use less power, better specs, stay cooler) will hold up fine against the older Ti (more power, hotter, worse specs, etc.).

[Eug]

Originally posted by Simon:
Definitely it's a better choice.

Proof can be found at this Barefeats page where you will see how a 15" 1GHz Ti compares to a 15" 1.25GHz Al. The Ti needs 300.3sec and the Al needs 227.2sec. The additonal 250MHz would only account for a time of about 240sec, so the Al actually gained more speed than what the clock would tell you.

Of course the reasons for this (bus, RAM, GPU, etc.) can be debated, but it is already a good sign that the new Al PowerBooks (clocked higher, use less power, better specs, stay cooler) will hold up fine against the older Ti (more power, hotter, worse specs, etc.).

Looking good. So I guess you're right with this bench at least - the 1.25 is doing better than clock for clock performance.

Note: It's not the GPU, since this test does not utilize it.

[mainemanx]

Originally posted by Simon:
It would really be grand if somebody could confirm this. Can't anybody just buy a new PowerBook, open it, rip it apart and tell me the numbers printed on top of the chip.

I give it one one week before we see a post, probably from Tokyo, where someone will have done just that ... shredded their new PB

[michaelb]

Originally posted by Simon:
The Ti needs 300.3sec and the Al needs 227.2sec. The additonal 250MHz would only account for a time of about 240sec, so the Al actually gained more speed than what the clock would tell you.

How are you calculating this?

Relative to the 1 Ghz Ti, the time saving is 300 - 227 = 73 seconds.

So the speed increase is:

73 secs / 300 secs = ~25%.

Moving from a Ti to an Al has increased the speed by 25% in this test.

Or am I missing some special way of calculating performance increases?

[cambro]

Originally posted by Simon:
Definitely it's a better choice.

Proof can be found at this Barefeats page where you will see how a 15" 1GHz Ti compares to a 15" 1.25GHz Al. The Ti needs 300.3sec and the Al needs 227.2sec. The additonal 250MHz would only account for a time of about 240sec, so the Al actually gained more speed than what the clock would tell you...

Yeah, but look at that Centrino!!! Clock for clock, it is well ahead of the G4. Ouch.

[Simon]

Originally posted by Eug:
Note: It's not the GPU, since this test does not utilize it.

Ah, good to know. Thanks for that info. I don't use Cinebench so I have no idea what it really stresses. I trusted Rob on that one.

This makes it only better.

Prelimenary conclusion: The new PowerBook, although it has a low-power G4 and no L3 cache, runs at least as fast as the old PowerBook clock per clock. And since it's also clocked faster it really should be faster in overall everyday-use. That's great!

[Simon]

Originally posted by mainemanx:
I give it one one week before we see a post, probably from Tokyo, where someone will have done just that ... shredded their new PB

I love Japan. Very cool people.

Arigato-Gosai-Mas! (or something like that)

[Simon]

Originally posted by michaelb:
How are you calculating this?

I did it in my head this morning before coffee so let's check.

1GHz : 1.25 GHz is 0.8.

0.8 = 240 : 300 <- that's where I got the 240 from.

But, the Al got 227 instead of 240, i.e. it was actually 6% faster. So I concluded it's doing better than equal clock per clock.

I still think my calc is OK, but since I haven't had much coffee in the meantime, I guess I could still be wrong.

I hope not though.

[acadian]

Someone had posted a link to an Apple developer note which confirms it is indeed the 7457, they have simply omitted the L3 cache which leads me to believe it was a cost issue. Moto's tech sheet shows that the 7457 can support up to 2Mb of L3 cache. While it is obvious that the 7457 may indeed outperform the 7447 clock per clock, why would they incorporate L3 support if was redundant towards the chips performance? Again, it must have been a cost issue on Apple's part in that they chose an overall reduction in the products cost over an even greater speed boost.

[Simon]

Originally posted by cambro:
Yeah, but look at that Centrino!!! Clock for clock, it is well ahead of the G4. Ouch.

Well, I guess we will have to see if this is a general rule or if the Centrino is just doing especially good on this test (Altivec? Bandwidth?).

Then again, the Centrino is no option for Apple I suppose...

I'm wondering how a 90nm 970-successor would compare in terms of power and heat with the Centrino.

[schk]

The new PB may be using 7447. The Apple docs don't specifically state which chipset is being used. All they do is link to the 7457 PDF file, but that doesn't mean it's not 7447. The 7447 specs are contained in the same 7457 PDF file, there is no separate PDF for the 7447 rather they are combined in one.