27-Sep-2021 13:34 GMT.
UNDER CONSTRUCTION
Anonymous, there are 54 items in your selection [1 - 50] [51 - 54]
[News] Transcript of Alan Redhouses speech at AmiGBG 2003ANN.lu
Posted on 31-Mar-2003 19:40 GMT by Peter Gordon54 comments
View flat
View list
I have transcribed Alans AmiGBG speech. It can be viewed here.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 1 of 54ANN.lu
Posted by Ole-Egil on 31-Mar-2003 18:04 GMT
Slight correction:

" Can I just say, Ole-egil and Justin have done sterling work right from
the beginning.. is there anything you'd like to add to all this stuff?

People who have done sterling work: [muffled] you're on your own, hehe

Alan: Right. heh.

People who have done sterling work: [muffled]

Alan: What are they?

People who have done sterling work (and then possibly the audience again):
[muffled]"

Actually, The first line was me (that he was on his own). The second and third line was actually the audience, from what I recall. I'll try to think up the correct questions that was asked during all of this, gimme some time (I'm assembling the new A1200 I bought from Gunne at the show at the moment :-)
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 2 of 54ANN.lu
Posted by Ole-Egil on 31-Mar-2003 18:07 GMT
"Audience Member: [muffled]

Alan: I don't know if they want a copy of that or what else."

The question was something along the lines of: "How are IBM viewing MorphOS?" or something. I think so, at least.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 3 of 54ANN.lu
Posted by Ole-Egil on 31-Mar-2003 18:10 GMT
"Audience member: [muffled question]

Alan: It has to be linux on the server farm, yeah, but the point is...

Audience member: [muffled] but you were also talking about the.. switch on and
it works, switch off..."

Q1: "Is it only Linux you're thinking of for the server farm concept?"
Q2 is actually pretty much complete, just add "mumble mumble OS4 mumble mumble" and you get the picture. Shouldn't be too hard from the context...
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 4 of 54ANN.lu
Posted by Ole-Egil on 31-Mar-2003 18:12 GMT
"Audience Member: [muffled]

Alan: How long will it take before what?

Audience Member: [muffled]

Alan: Thats on the AmigaONE board? "

This was basically a follow-up on the PPC970 question...
How long will it take before that is on the AmigaOne motherboard...
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 5 of 54ANN.lu
Posted by Peter Gordon on 31-Mar-2003 18:25 GMT
In reply to Comment 4 (Ole-Egil):
OK thanks. I've updated it a bit.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 6 of 54ANN.lu
Posted by SlimJim on 31-Mar-2003 18:40 GMT
I also have a gap to fill ;-)

---------------------------------------------------------------
Audience member: [muffled]

Alan: Hmm.. well, some of that stuff is sort of in flux.
[...]
---------------------------------------------------------------

The Audience member is me, and I asked "To what extent is Amiga Incoorporated
participating in the AOne and AOS4 endeavour"?

...That is what I meant to ask anyway, but I think Alan
misunderstood/misinterpreted me as asking about AmigaInc's plans in general.

Sterling work on the transcript by the way! Thanks!
.
SlimJim
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 7 of 54ANN.lu
Posted by SlimJim on 31-Mar-2003 18:45 GMT
... And the "Gunnar" it says Alan is referring to as getting the boards - it
should be "Gunne" in all cases.
.
SlimJim
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 8 of 54ANN.lu
Posted by SlimJim on 31-Mar-2003 18:54 GMT
[...]We do have a very good relationship with MAI logic. I talk to the chairman
two or three times a week, and its a two way thing, I mean he asked me for
some advice on some stuff, not *technical* stuff I hasten to add, professional
stuff[...]

The word "technical" was actually very muffled for us in the audience too. But I
got the impression he was actually saying "not *personal* stuff". But I'm not
sure.
.
SlimJim
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 9 of 54ANN.lu
Posted by Ole-Egil on 31-Mar-2003 19:02 GMT
In reply to Comment 8 (SlimJim):
It was "not technical stuff". Alan isn't an engineer, really.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 10 of 54ANN.lu
Posted by Peter Gordon on 31-Mar-2003 19:13 GMT
In reply to Comment 9 (Ole-Egil):
Thanks guys. I've updated it again.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 11 of 54ANN.lu
Posted by James Carroll on 31-Mar-2003 22:48 GMT
Thanks for typing this up. It was a good read.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 12 of 54ANN.lu
Posted by Atheist2 on 31-Mar-2003 23:50 GMT
Works for me!! :))))

Thanks for the transcript!


AmigaOne! "...don't give up now!"
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 13 of 54ANN.lu
Posted by Anonymous on 01-Apr-2003 01:39 GMT
Great transcript, thank you very much for the effort!

What did Alan mean with just 3% speed gain through L3 cache?
On Macs it makes quite a difference and games (Hyperion!) like Quake3 benefit
greatly from it. Some examples of iMac 1GHz w/o L3 cache vs. PowerMac 1GHz with
(just) 1MB L3 cache [ct 6/2003 page 108]:

Cinebench 2001
Shading (C4D): 7.44 vs. 8.08
Shading (OpenGL): 9 vs. 9.79

Photoshop 7
Action 1: 23s vs. 20s
Action 2: 48s vs. 42s

iTunes
MP3 encoding: 5min48s vs. 4min40s

Quake3
Demo Four 68: 67.7fps vs. 82.9fps

Quicktime
Sorenson 2: 3min vs. 2min47s
Sorenson 3: 1min16s vs. 1min09

Both had the same gfxcard (GF4MX 64MB DDR) and memory subsystem (PC2100).

http://www.xlr8yourmac.com/G4CARDS/giga_designs_800_1ghz/index.html shows
that L3 cache can make an average difference of well over 10% (please also
notice the extreme Q3 and UT results...).
http://macspeedzone.com/ ist another source for comparative numbers.
http://www.powerlogix.com/downloads/SDRDDR.pdf comes to the following conclusion: "Clearly it is important to have some sort of L3 cache, but the type and
speed of the L3 are almost irrelevant. In fact, the only factor that matters to any degree, is
L3 cache size, and even that is not a huge difference"

So _if_ Alan only gets 3% performance gain from L3 cache there must be something
wrong with AmigaOne - also the G4 working only up to 40 degrees sounds like a
joke since the CPUs are specified for at least 65 degrees operation.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 14 of 54ANN.lu
Posted by reflect on 01-Apr-2003 06:00 GMT
In reply to Comment 13 (Anonymous):
well, I haven't checked your links yet(no time) but is it the exact same CPU?
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 15 of 54ANN.lu
Posted by Ole-Egil on 01-Apr-2003 06:44 GMT
In reply to Comment 13 (Anonymous):
Look, if Motorola admits to only getting 3% speedup (min, typical, max? I dunno), then those numbers are MOST probably NOT from the same CPU. If Motorola had said "that's not correct" I would MOST probably have agreed with you. But they didn't. They admitted to seeing the same results.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 16 of 54ANN.lu
Posted by Ole-Egil on 01-Apr-2003 06:46 GMT
In reply to Comment 13 (Anonymous):
Actually, you HAVE to be comparing with a different CPU, as the one used in the AmigaOne is brand new and doesn't have a 1G version yet.

I suggest you take it up with Motorola, ok?
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 17 of 54ANN.lu
Posted by Jim Forbes-Ritte (AGAfaster) on 01-Apr-2003 07:16 GMT
Mr Gordon,

A superb read !
Well in son, definitely top drawer ! ;-)
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 18 of 54ANN.lu
Posted by alan buxey on 01-Apr-2003 08:33 GMT
In reply to Comment 13 (Anonymous):
if IBM and Motorola engineers say that theres only 3% speedup
on that particular config (due to memory BUS for example - this isnt a PC2100 system if you note!) . then i believe them. if some unknown anonymous troll
on slashdot or ANN says that it should be XX%, then i dont believe them.

if someone at motorola says that those particular CPUs (due to formfactor,
mounting type or production style) should only run at 40 degrees, then i believe
them. i decent passive heatsink or fan should keep the temperature below 35 degress anyway. after all, these things dont kick out that much heat!

alan
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 19 of 54ANN.lu
Posted by steffo / cryonics on 01-Apr-2003 09:31 GMT
In reply to Comment 13 (Anonymous):
About the speed tests above in the forum.
Most of your tests actually show around 3% speed gain, no? :)
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 20 of 54ANN.lu
Posted by Anonymous on 01-Apr-2003 09:40 GMT
In reply to Comment 18 (alan buxey):
But we have no quality evidence that Motorola said anything at all.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 21 of 54ANN.lu
Posted by Ole-Egil on 01-Apr-2003 09:45 GMT
In reply to Comment 20 (Anonymous):
No, of course. Alan is a bad troll who actually on several occasions has been known to lie to my face.

Nah.

Have you SEEN my face up close? I'm a pretty ugly dude, and I have been known to lift heavier things than Alan to throw it in the ocean (which was about 50 meters away when he said this). Come on, not everyone lies. Some people actually tell the truth from time to time...
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 22 of 54ANN.lu
Posted by Anonymous on 01-Apr-2003 11:05 GMT
In reply to Comment 19 (steffo / cryonics):
****bzzzzt*

Wrong answer, thanks for playing.

(Hint - try doing the maths on those results, rather than simply guessing the answers you WANT to hear. Then explain to us how 15% to 20%+ somehow equals 3%)
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 23 of 54ANN.lu
Posted by Don Cox on 01-Apr-2003 11:20 GMT
In reply to Comment 21 (Ole-Egil):
"Have you SEEN my face up close? I'm a pretty ugly dude, "

Oh I dunno. Judging from the photos, more like Santa Claus in his younger days, when his family were worrying about whether he would ever find a real job.

;-)
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 24 of 54ANN.lu
Posted by Jim Forbes-Ritte (AGAfaster) on 01-Apr-2003 11:53 GMT
In reply to Comment 23 (Don Cox):
Kinda Dilbert-like ... :-]
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 25 of 54ANN.lu
Posted by Ole-Egil on 01-Apr-2003 12:48 GMT
In reply to Comment 24 (Jim Forbes-Ritte (AGAfaster)):
OI, I HEARD THAT ONE! :-)
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 26 of 54ANN.lu
Posted by Ole-Egil on 01-Apr-2003 12:49 GMT
In reply to Comment 23 (Don Cox):
I am thinking of spending at least one christmas with a white beard. Depends if my family force-shaves me or not, I guess :-)
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 27 of 54ANN.lu
Posted by SlimJim on 01-Apr-2003 16:12 GMT
In reply to Comment 26 (Ole-Egil):
I won the beard-contest...

...on style and quality! :-
.
SlimJim
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 28 of 54ANN.lu
Posted by Anonymous on 01-Apr-2003 19:09 GMT
In reply to Comment 15 (Ole-Egil):
>They admitted to seeing the same results.

The question is which results they were looking at?
L3 cache is only good for _cached_ data obviously.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 29 of 54ANN.lu
Posted by Anonymous on 01-Apr-2003 19:20 GMT
In reply to Comment 16 (Ole-Egil):
>Actually, you HAVE to be comparing with a different CPU, as the one used
>in the AmigaOne is brand new and doesn't have a 1G version yet.

First test were done with MPC7455-1000, the 800MHz test without L3 cache is MPC7451
similar to AmigaOne-XE while the 800MHz with L3 cache is a clocked down MPC7455.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 30 of 54ANN.lu
Posted by Anonymous on 01-Apr-2003 19:28 GMT
In reply to Comment 18 (alan buxey):
>IBM and Motorola engineers say that theres only 3% speedup

Who says IBM?

>on that particular config (due to memory BUS for example - this isnt a PC2100
>system if you note!) . then i believe them.

So the benefit should be even higher because AOne only supports PC133? :-)

>if some unknown anonymous troll on slashdot or ANN says that it should be XX%,
>then i dont believe them.

Beleive what you want, I only gave comparative Mac numbers and their sources.
If AOne performs worse we should ask why this is the case (but feel free to
buy it just for the sake of the name without looking at stupid performance
figures).
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 31 of 54ANN.lu
Posted by David Scheibler on 01-Apr-2003 22:03 GMT
In reply to Comment 18 (alan buxey):
>if some unknown anonymous troll on slashdot or ANN says that it should be XX%,
>then i dont believe them

These test were done by the German magazine c't (ct 6/2003 page 108). That was
said in the post. I guess the guys there know what they're doing...
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 32 of 54ANN.lu
Posted by StormLord on 01-Apr-2003 22:05 GMT
In reply to Comment 30 (Anonymous):
I 'm not going to cover anyone ass here but if the results are correct (I have no reason to belive the opposite) I must remind you a CRUSIAL point: The test are from different motherboards, and please don't tell me that a mac is a mac and they are all the same cause I have proof that Macs with the same cpu and speed (and OS ofcourse) but different models give different results EVEN on RC5 that is an only a cpu matter!!! Just try to compare an 4400 with an 53xx or 54xx and you will get the point! ;-)
Maybe none of them are lying! ;-)
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 33 of 54ANN.lu
Posted by Anonymous on 02-Apr-2003 03:50 GMT
In reply to Comment 32 (StormLord):
It's funny you are providing RC5 as example for different results sincehttp://personal.inet.fi/cool/pekosbil/a1benchmarks.htm shows exactly thesame numbers for the different AmigaOne and Pegasos G3-600 boards!The iMac and PowerMac seem to have the same memory controller, and the800MHz tests were done on the same machine (3rd party CPU upgrade).
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 34 of 54ANN.lu
Posted by Ole-Egil on 02-Apr-2003 06:18 GMT
In reply to Comment 33 (Anonymous):
Well, he's actually right when it comes to a Mac not being "just" a Mac. There can be some REALLY big differences (let's face it. you don't think they have ONE guy designing everything, right? So naturally it's going to be different from model to model). If I am going to believe that Alan is lying, I need to see tests done _with and without L3 cache on the same motherboard with the same cpu_. OS version and applications must also be IDENTICAL. So a test done on two machines is NOT useful. It must be the same machine with and without L3 cache with NO other changes done in between. Alan claims Motorola told him this, and I believe him. Alan is one of the few sales people who have never lied to me. Note that I wouldn't call a missed deadline a lie if anyone else did it either, so it might just be my attitude...
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 35 of 54ANN.lu
Posted by samface on 02-Apr-2003 07:35 GMT
In reply to Comment 34 (Ole-Egil):
Yes, there seems to be alot of confusion in the Amiga community about estimated timeframes and "promises". Not many seem to be able to tell the difference around here...
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 36 of 54ANN.lu
Posted by IanG on 02-Apr-2003 10:10 GMT
So far everyone has considered:

1) Alan is stupid/lying.
2) Motoroal/IBM don't know what their own chip does.
3) The AOne must be crap because the L3 cache doesn't help.

Why does nobody consider:

4) The Mac (or whatever architecture the tests quoted in this thread were done on) is inefficient and requires an L3 cache to get full performance, whereas an AOne utilises the PPC full performance as intended and the L3 cache therefore only gives 3% like the manufacturers say it will.


BTW: Well done to the person who transcibed this speech. Very informative and well presented.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 37 of 54ANN.lu
Posted by David Scheibler on 02-Apr-2003 10:17 GMT
In reply to Comment 34 (Ole-Egil):
Well the best way would just be that Eyetech releases the benchmarks they have
done to compare the G4 with and without L3 cache. It seems that they did many
benchmarks and even did benchmarks of what they thought that the l3 impact must
be bigger because otherwise they hadn't asked Motorola about it later. So just
publish the benchmarks and everyone can compare himself.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 38 of 54ANN.lu
Posted by Anonymous on 02-Apr-2003 10:41 GMT
In reply to Comment 36 (IanG):
>The Mac (or whatever architecture the tests quoted in this thread were done on)>is inefficient and requires an L3 cache to get full performance, whereas an AOne>utilises the PPC full performance as intended and the L3 cache therefore only>gives 3% like the manufacturers say it will. Sorry, but this sounds like a fairy tale.It's rather the other way round with Macs having the more professional and efficientarchitecture while AOne uses a STB chipset with inferior throughput...
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 39 of 54ANN.lu
Posted by Anonymous on 02-Apr-2003 11:21 GMT
In reply to Comment 36 (IanG):
Are you SERIOUSLY trying to claim that in a well-designed system, an external L3 cache will only give 3% performance improvements?

Because if so, then you are talking out of your backside, as anyone with even a vague clue about system designs will realise (unless, of course, your L1 and L2 cache are so large - say 8MB - that L3 would not be needed - this is blatantly not the case here).
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 40 of 54ANN.lu
Posted by IanG on 02-Apr-2003 14:30 GMT
In reply to Comment 39 (Anonymous):
> Are you SERIOUSLY trying to claim that in a well-designed system, an external
> L3 cache will only give 3% performance improvements?
> Because if so, then you are talking out of your backside,

Whoa! I'm not claiming anything. I'm simply pointing out that you are *assuming* that the AOne is the crap item in the loop here. You have no more evidence of that than I do of my *hypothesis* that the Mac systems may be. You may not like Amigas, Eyetech or AOne's, but that does not automatically mean they are crap just because you say so.

> as anyone with even a vague clue about system designs will realise (unless,
> of course, your L1 and L2 cache are so large - say 8MB - that L3 would not be
> needed - this is blatantly not the case here).

You say "anyone with a vague clue". No offence, but your hand-waving suggests you only have at best a layman's understanding of the issues involved. Motorola and IBM *who designed the chip we're talking about* probably do have a fairly good clue about cache system designs. And they say you should see 3%.

The only real solution is to see if the Mac outperforms the AOne by as much as the L3 cache tests suggest. We can't do that because too many other variables come into play when the systems differ by so much.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 41 of 54ANN.lu
Posted by IanG on 02-Apr-2003 14:41 GMT
In reply to Comment 38 (Anonymous):
> Sorry, but this sounds like a fairy tale.
> It's rather the other way round with Macs having the more professional and
> efficient architecture while AOne uses a STB chipset with inferior throughput...

And you are probably right. I do not know, nor particularly care, whether the L3 cache gives 3% or 15%. Neither is particularly significant. I am merely pointing out that several comments here have been written with the unspoken assumption that the AOne must be inferior because it is an AOne. Using the pyschology of speaking as if something is true beyond the need to even mention it is quite an effective way of getting people to subconciously believe the things you want them to.

And before anyone starts on the subject, I'm not an "Amiga zealot". I'd be making similar comments about the Pegasos were it under discussion. Which I note that nobody has yet tried to claim is better in this thread. Yet.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 42 of 54ANN.lu
Posted by Anonymous on 02-Apr-2003 16:03 GMT
In reply to Comment 40 (IanG):
>Motorola and IBM *who designed the chip we're talking about* probably do have a
>fairly good clue about cache system designs. And they say you should see 3%.

Again - where does IBM say that?
Apple is (was?) part of AIM alliance and has the most experience when it comes
to desktop computer systems (MOT concentrates on embedded appliances).

Apple say on http://www.apple.com/powermac/architecture.html :
"L3 cache keeps the PowerPC G4s engine stoked
L3 cache is high-speed Double Data Rate memory. It provides fast access to data
and application code through a dedicated bus to the processor. This dedicated
bus provides throughput of up to 4 gigabytes per second (GBps), and is completely
unhindered by any other data transfers. The high speed made possible by L3 cache,
with its dedicated bus, allows the PowerPC G4 processor to receive data more
than five times faster than it could from main memory (where a shared bus
lowers speed). Because of this low latency, the G4 processor is constantly fed
with data. So it doesnt just sit idle, waiting for the next set of data to arrive."
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 43 of 54ANN.lu
Posted by Peter Gordon on 02-Apr-2003 16:23 GMT
All the people moaning about L3 cache performance on the AOne: He didn't say AmigaONE, he said Motorola board, which most likely means their developer board.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 44 of 54ANN.lu
Posted by Anonymous on 02-Apr-2003 16:33 GMT
In reply to Comment 43 (Peter Gordon):
So he tests his MegArray G4 cards on some Motorola board? Why?
My first interpretation was that he called the G4 CPU card "Motorola board"
because it uses a CPU by Motorola (unlike IBM/G3 card). Corrections welcome!
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 45 of 54ANN.lu
Posted by Peter Gordon on 02-Apr-2003 19:11 GMT
In reply to Comment 44 (Anonymous):
Why would Motorola know what sort of speed increase L3 cache would give on an AmigaONE board?

If they say "Oh, we knew that. Didn't you know?", that implies to me that the Motorola dev board only gains 3% when you put L3 cache on it. Thats what I thought he meant anyway.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 46 of 54ANN.lu
Posted by Bill "tekmage" Borsari on 02-Apr-2003 19:26 GMT
I find it interesting people are debating the performance of L3 cache on a product which is many months late to ship :)

3% or 20% I don't care, I want an Amiga1XE with OS4! :)

I think that Hyperion and Alan's team have been working hard with no or little money to get this done in the best way possible. I'm really looking forward to giving Alan my money so I can play with the fruits of their labor.

I was very happy to see Alan was thinking of the "what next" for the mobo, it sounded like a laptop version would be very do-able if the market existed. I was also impressed by the rack server idea, gig-E and onboard raid would be very cool.

I hope Alan is able to make AmiWest and so is Hyperion along with shipping versions of the A1 and OS4. I'll personally buy a round at the bar for those guys!

Bill "tekmage" Borsari
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 47 of 54ANN.lu
Posted by Zorro on 03-Apr-2003 09:31 GMT
In reply to Comment 46 (Bill "tekmage" Borsari):
I fully agree with you
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 48 of 54ANN.lu
Posted by IanG on 03-Apr-2003 10:10 GMT
In reply to Comment 42 (Anonymous):
>>Motorola and IBM *who designed the chip we're talking about* probably do have a
>>fairly good clue about cache system designs. And they say you should see 3%.

>Again - where does IBM say that?
>Apple is (was?) part of AIM alliance and has the most experience when it comes
>to desktop computer systems (MOT concentrates on embedded appliances).

OK. I am going by the comment that was made in Alan's speech which said their inability to achieve much speed gain from the L3 resulted in Mot/IBM(?) saying that was what they expected. It is entirely possible that Alan misinterpreted what was said or has misrepresented the situation, or maybe I've misunderstood what he said.

> Apple say on http://www.apple.com/powermac/architecture.html :
> "L3 cache keeps the PowerPC G4s engine stoked

Well, yes, I'm sure Apple, just as any other company, will squeeze every marketing opportunity out of "extra" hardware or features, whether they exist to truly beneift or just to overcome shortcomings.

> L3 cache is high-speed Double Data Rate memory. It provides fast access to
> data and application code through a dedicated bus to the processor [....]

The description is little more than a description of any generic cache system. I had assumed (feel free to correct me) that an L3 cache system exists in addition to an L2 cache, or that the tests that Alan refer to are relative between an L3 and L2 (or other) cache system. Not the lack of any cache whatsoever.
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 49 of 54ANN.lu
Posted by Ole-Egil on 03-Apr-2003 10:41 GMT
In reply to Comment 48 (IanG):
Excactly.

The L3 gave the G4 in this setup a 3% increase in throughput versus the board with only L1/L2.

I think that sounds about right, really. Not much application-wise should end up in the L3 cache, since it's either just running inside L1/L2, or running in memory (either small or large, not too much happens to be about 2MB in memory footprint).

Remember that if you have L3 enabled you will often get an extra penalty because of an L3 cache miss. It takes longer to miss cache 3 times than 2, so system without L3 has faster memory access...
This might be a reason...
Transcript of Alan Redhouses speech at AmiGBG 2003 : Comment 50 of 54ANN.lu
Posted by Anonymous on 03-Apr-2003 12:47 GMT
In reply to Comment 49 (Ole-Egil):
>I think that sounds about right, really. Not much application-wise should end up
>in the L3 cache, since it's either just running inside L1/L2, or running in memory

What about a petition for Alan to get us rid of the expensive useless L3 cache then?
Only Macs need it for their marketing machine to have an "advantage" over PC.
Anonymous, there are 54 items in your selection [1 - 50] [51 - 54]
Back to Top