[U n i o n 0015]

I'd love for an announcement at NAB, but realistically, Apple will disappoint me and not announce anything until June. Or later. ARGH!

[Eug Wanker]

It'd be nice to see the Power Mac at NAB, but the eMac is in even worse shape. On Wednesday the current eMac is a year old. However, it has been over 10 months already for those Power Macs, too.

Now, even if the Power Mac does get updated, one wonders if Apple would wait until later before selling quads.

Also, while I think there's no question IBM could release a 3.0 GHz G5 today, I'm growing more worried they won't be able to do the same for a dual-core chip. A dual-core 3.0 GHz G5 970MP seems like a tall order. Even a 2.8 GHz dual-core is pushing it. I think at least a 2.5 GHz dual-core is pretty much guaranteed though.

[goMac]

I don't expect a dual dual core G5 would cost that much to build. Probably just as much as current ones. It only costs IBM a few cents to make each chip, but a lot to build the factory. That said, there is no reason that they couldn't charge the same for a multi core chip.

[Eug Wanker]

Originally posted by goMac:
It only costs IBM a few cents to make each chip

Uh, no.

Addendum:

BTW everyone... WWDC is only 2 months away. It's amazing how quickly a year can go by...

[goMac]

Originally posted by Eug Wanker:
Uh, no.

Addendum:

BTW everyone... WWDC is only 2 months away. It's amazing how quickly a year can go by...

Actually I've looked at this in economics class. The actual production cost of the chip is very little. It doesn't cost much to make the wafer. Where the costs come in are the few billion for the plant, and the research. Thats why chips cost so much.

A dual core chip won't cost that much more to produce than a single core. Chipmakers usually charge a markup for said chips though. In addition all mhz chips are actually the same. They're all made in the same batch together. A chipmaker will assign speeds to chips based on which price points they want to sell at (and any chips that are somewhat defective can be clocked down). For example, my 3.2 ghz P4 was made in the same batch as all the 3.6 ghz chips. Thats why I can run it at 3.6 ghz just fine. There is no difference between the two, only that my chip has been marked as a 3.2 ghz one. Of course not all 3.2 ghz chips will run at 3.6 because you might get one that has a defect.

Anyway, thats my mini lesson on the economics of chip making.

Edit: Anyway, why we were talking about this in economics was that chips are very easy and cheap to make, that in theory it would be a great market for any firm to move into. But the entry costs are so high (the plant and the research) that firms can't really enter. This is why there are so few chip makers.

[Catfish_Man]

Originally posted by goMac:
It only costs IBM a few cents to make each chip, but a lot to build the factory..

About halfway down the page here is info on the wafer costs for a 50mm^2 chip on a non-cutting edge 250nm manufacturing process with 4 layers and aluminum interconnects. http://www.informit.com/articles/art...&seqNum=4&rl=1

Definitely quite a bit more than "a few cents", and a 90nm SOI process is going to cost a good bit more.

[goMac]

Originally posted by Catfish_Man:
About halfway down the page here is info on the wafer costs for a 50mm^2 chip on a non-cutting edge 250nm manufacturing process with 4 layers and aluminum interconnects. http://www.informit.com/articles/art...&seqNum=4&rl=1

Definitely quite a bit more than "a few cents", and a 90nm SOI process is going to cost a good bit more.

Well, my professor could have been a bit carried away. That said, you can fit quite a few chips on a wafer. It could probably get down to the $50-$100 range per chip. And a dual core vs. a single would still be the same price to produce (not counting r & d).

[Catfish_Man]

Originally posted by goMac:
Well, my professor could have been a bit carried away. That said, you can fit quite a few chips on a wafer. It could probably get down to the $50-$100 range per chip. And a dual core vs. a single would still be the same price to produce (not counting r & d).

I'd guess <$50 personally. I don't see where you're getting the "dual core is the same price" though. It is twice as big...

[MORT A POTTY]

actually, 90nm is cheaper on a chips per wafer basis than was 130nm, and 180nm... smaller they get, more chips per wafer they can get. R&D we can figure remains constant, or slightly more expensive, so it's all gravy baby

[power142]

Originally posted by goMac:
A chipmaker will assign speeds to chips based on which price points they want to sell at (and any chips that are somewhat defective can be clocked down). For example, my 3.2 ghz P4 was made in the same batch as all the 3.6 ghz chips. Thats why I can run it at 3.6 ghz just fine. There is no difference between the two, only that my chip has been marked as a 3.2 ghz one. Of course not all 3.2 ghz chips will run at 3.6 because you might get one that has a defect.

This process you allude to is called 'bin splitting' - the chip manufacturers test each chip and determine what the most reliable speed is for a given chip and throw it in the respective speed bin. If a given P4 chip runs perfectly at 3.2GHz but not to _their_ specifications (since they give it the warranty) at 3.4 or 3.6, then that's where it ends up. Typically, the yield of higher speed chips will be less than those at lower speeds, thus making them better candidates to be sold at higher prices.

The number of transistors for a given chip also has an effect on the yield of a wafer. For example, a wafer might give better yields of G5 chips than of the P4 expensive edition (with 2MB cache) since the latter consists of many more transistors, while the former has fewer... assuming that the manufacturing process is reasonably well optimized. As we've seen in G5 history, this may not always be the case, as 2.5GHz PowerMac availability was severely constrained by IBM's ability to produce them in quantity. In an ideal process, chip fabrication would be perfect every time and all chips would be rated at the speed Intel/AMD/IBM intended - that isn't the way it is today.

To summarize, my point is that all chips in a given wafer might be made equal in theory, but that doesn't carry through in practice. The relevant point for the dual-core G5 chips is that because of the greater transistor count, fewer chips are possible per wafer, and unless IBM has (close to) perfected the manufacturing process, there is also a greater chance of defects for a given chip (but they could easily add extra 'bins' for cases where parts of the increased cache are defective, for example). This will inevitably drive up costs to the system builders, in our case, Apple. Whether the unit cost of the chips are significantly higher than the earlier G5 chips is up to IBM and Apple to determine.

[Catfish_Man]

Random semi-offtopic comment about the effects of cache area on yields: according to some posts at realworldtech.com by Paul DeMone (who seems to quite know what he's talking about), cache area on a chip tends to not change the yield % very much due to two factors: it's apparently fairly easy to include a few extra blocks of cache in case one or two are defective, and, as mentioned before, if enough is defective you can just turn some off and sell it as a low-cache version.

[power142]

That's very interesting... glad to see things have moved on somewhat since the days I picked up the info in my previous post - back in the era of Athlon/Duron

[Eug Wanker]

Originally posted by Catfish_Man:
Random semi-offtopic comment about the effects of cache area on yields: according to some posts at realworldtech.com by Paul DeMone (who seems to quite know what he's talking about), cache area on a chip tends to not change the yield % very much due to two factors: it's apparently fairly easy to include a few extra blocks of cache in case one or two are defective, and, as mentioned before, if enough is defective you can just turn some off and sell it as a low-cache version.

Except Apple doesn't sell lower cache G5s. They didn't sell lower cache G4s either.

[Catfish_Man]

Originally posted by Eug Wanker:
Except Apple doesn't sell lower cache G5s. They didn't sell lower cache G4s either.

hence the "semi-offtopic". The post(s) I referenced were talking about Itanium 3.

[Simon]

ThinkSecret says new PowerMacs with the 970MP or 970GX at NAB. They expect new iMacs with the 970FX this month.

[BZ]

Originally posted by Simon:
ThinkSecret says new PowerMacs with the 970MP or 970GX at NAB. They expect new iMacs with the 970FX this month.

Vague as always.

Nobody knows nuthin!!!

BZ

[U n i o n 0015]

It's the same rumor report every time. It just gets recycled around the same sites.

Oh well, NAB is next week, we'll know then won't we? For know, I'm going to bask in the glory that is the Tiger ship date announcement.

[Eug Wanker]

AI claims that they have info that quads are being tested although they can't verify that new Power Macs are going to be released at NAB:

However, sources recently reported that Apple is performing quality assurance tests on new Power Macs based on IBM's unannounced dual-core PowerPC 970MP processor, code-named "Antares." In order to properly test the systems, which reportedly sport two of the dual-core chips, engineers have been seen upgrading office and testing lab power outlets to enable proper testing of the desktop's power-hungry power supply.

Sources also report that the company is testing lower-end systems based on IBM's single-core PowerPC 970GX.

Both the 970GX and the 970MP offer better performance while running much cooler than the PowerPC 970FX chips found in the current Power Mac offerings. This will allow Apple to do away with the costly, and seemingly unreliable, liquid-cooling systems featured in Apple's high-end dual-processor Power Mac G5.

These liquid-cooling systems, which were designed by Delphi, manufactured in Mexico, and shipped to Apple's manufacturing facilities in Asia, have been the topic of concern at several Apple exec meetings.

According to reliable sources, Delphi only guarantees the cooling systems to Apple for a period of 2.5 years, though the expected life-span of each unit is rumored to be closer to 2 years. The lifespan is reportedly limited by potenial leakage of the thermal conductive fluid inside the systems. Apple fears the fluid could cause damage to consumer's valuables or expensive carpets, which would leave the company exposted to potential lawsuits.

Seems like a bit of a strange article though.

[osxisfun]

>According to reliable sources, Delphi only guarantees the cooling systems to Apple for a period of 2.5 years, though the expected life-span of each unit is rumored to be closer to 2 years. The lifespan is reportedly limited by potenial leakage of the thermal conductive fluid inside the systems. Apple fears the fluid could cause damage to consumer's valuables or expensive carpets, which would leave the company exposted to potential lawsuits.



that's scary.

[jamil5454]

HAHA @ all the people who bought the DP 2.5 PowerMac.

[ReggieX]

Originally posted by Eug Wanker:
According to reliable sources, Delphi only guarantees the cooling systems to Apple for a period of 2.5 years, though the expected life-span of each unit is rumored to be closer to 2 years. The lifespan is reportedly limited by potenial leakage of the thermal conductive fluid inside the systems. Apple fears the fluid could cause damage to consumer's valuables or expensive carpets, which would leave the company exposted to potential lawsuits.

Wow. I've had my over Sawtooth 5 years, if my G5 started leaking everywhere after 2 years I'd be pretty friggin' angry.

[Simon]

Originally posted by Eug Wanker:
Seems like a bit of a strange article though.

Indeed it does.

Especially, I call BS on the part where they mention that developers were 'testing lab power outlets' because of the power supply requirements of the dual duals. What a load of crap. In our EE labs we have 220V/20A on all outlets -> 4.4 kW power. No way in the world would any Power Mac ever require that kind of electrical power. The Power Mac's issue might be a new type of PS, but certainly not the electrical outlets. Or do they really think customers would want to upgrade their outlets for a computer? BS. Either Apple's labs are entirely ancient (with mice running generator wheels or something) or the 'sources' just made it up. I bet a six pack of Anchor Steam it's the latter.

[Eug Wanker]

Originally posted by Simon:
Especially, I call BS on the part where they mention that developers were 'testing lab power outlets' because of the power supply requirements of the dual duals. What a load of crap. In our EE labs we have 220V/20A on all outlets -> 4.4 kW power. No way in the world would any Power Mac ever require that kind of electrical power. The Power Mac's issue might be a new type of PS, but certainly not the electrical outlets. Or do they really think customers would want to upgrade their outlets for a computer? BS. Either Apple's labs are entirely ancient (with mice running generator wheels or something) or the 'sources' just made it up. I bet a six pack of Anchor Steam it's the latter.

I agree, unless they're just mixed up about the outlets.

[Simon]

Come to think of it, what's the deal with their mentioning of this mexican Delphi business? I mean 2.5 years contracts with Apple - WTF? Apple knew they were going to sell these 2.5 GHz Power Macs for at least 6-8 months and they knew they were selling AppleCare for three years of warranty. Why in the world should Apple make a contract for only 2.5 years? Certainly Apple needs spare parts, technical assistance, etc. for more then just 2.5 years. The article is really weird.

[BZ]

TS has "confirmed" specs... (although not which proc)

http://www.thinksecret.com/news/0504macs.html

"Dual-2GHz
512K L2 cache per processor
Dual 1GHz frontside buses
512MB PC3200 DDR SDRAM (4GB max.)
160GB Serial ATA Hard Drive
128MB DDR SDRAM ATI Radeon 9600 video card
Dual-2.3GHz
512K L2 cache per processor
Dual 1.15GHz frontside buses
512MB PC3200 DDR SDRAM (8GB max.)
250GB Serial ATA Hard Drive
128MB DDR SDRAM ATI Radeon 9600 video card
Dual-2.7GHz
512K L2 cache per processor
Dual 1.35GHz frontside buses
512MB PC3200 DDR SDRAM (8GB max.)
250GB Serial ATA Hard Drive
256MB DDR SDRAM ATI Radeon 9650 video card"

BZ

[Zoro]

isn't the 970MP supposed to have 1Mb l2 cache per core ? or is that total L2 cache (ie 512Kb per core)

[blackwind]

Originally Posted by Zoro

isn't the 970MP supposed to have 1Mb l2 cache per core ? or is that total L2 cache (ie 512Kb per core)

You are correct. The 970MP does have 1 MB of L2 cache per core. The above specifications indicate another 970FX-based release (assuming that they are accurate).

[Eug Wanker]

Yeah, the article talks about 970GX and 970MP, but those L2 cache specs are for the 970FX. And where is the quad? WWDC?

P.S. This new board seems like it may be a bit faster on my Cube.

[Pierre B.]

Originally Posted by Eug Wanker

Yeah, the article talks about 970GX and 970MP, but those L2 cache specs are for the 970FX. And where is the quad? WWDC?

Next year perhaps? Although we should wait and see. It is not that much time anyway.

[Simon]

Originally Posted by Eug Wanker

Yeah, the article talks about 970GX and 970MP, but those L2 cache specs are for the 970FX. And where is the quad? WWDC?

Yeah yeah, it's the FX again. Ha, you guys are just not patient enough.

I told you (here, here, and here) there would be no quads this time. Everybody will be disappointed, as expected.

Wake me up when there's news.

[Eug Wanker]

Originally Posted by Simon

I told you (here, here, and here) there would be no quads this time.

[Simon]

Mmmmh, thanks. I love cookies.

[osxisfun]

Originally Posted by Simon

Mmmmh, thanks. I love cookies.

so no dual duals?



Its a powerbook 1.67 for me.....


Probably going to be easier on my back.

[Simon]

Originally Posted by osxisfun

so no dual duals?

Nope.

Originally Posted by osxisfun

Its a powerbook 1.67 for me.....

Good choice, very nice machine. However, I doubt that's an alternative to a dual 970MP...

[osxisfun]

yah. I'll just have to buy the PB G5 in Jan (or whenever)

[Pierre B.]

Originally Posted by osxisfun

yah. I'll just have to buy the PB G5 in Jan (or whenever)

I feel sad to bring you the bad news, but you missed perhaps this one:

- Q: Last conference call, mentioned that putting a G5 in a PowerBook would be a significant technical challenge. A: Still think it's the "mother of all" thermal challenges.

Not gonna happen anytime soon. Maybe never.

[Leonard]

http://www.macworld.com/news/2005/04...wave/index.php

Lightwave has just announced support for Multi-core Processors. Is this another hint that Multi-core Power Mac G5's are coming soon?

[osxisfun]

Originally Posted by Pierre B.

I feel sad to bring you the bad news, but you missed perhaps this one:



Not gonna happen anytime soon. Maybe never.

I posted a thead about it.

the same guy said in a previous conf call that apple would never get into the sub $500 pc market and they did 3 months later.

There will be a pb g5, will it be the same g5 chip that we know today. no it will be some GX-super lite version that has been tweaked beyond belief.

I am stil betting Jan. anounce. ship a couple of months later?

[Eug Wanker]

Originally Posted by Leonard

http://www.macworld.com/news/2005/04...wave/index.php

Lightwave has just announced support for Multi-core Processors. Is this another hint that Multi-core Power Mac G5's are coming soon?

Not really, since Lightwave runs on Windows too, and dual-core desktop x86 chips are already here.

[OreoCookie]

Originally Posted by Eug Wanker

Not really, since Lightwave runs on Windows too, and dual-core desktop x86 chips are already here.

Well, they arrived today

[osxisfun]

TS is still sticking to their leaked specs.

Last bump before multicore?

[Pierre B.]

Originally Posted by osxisfun

TS is still sticking to their leaked specs.

Last bump before multicore?

Quite possible. I hope not in a year from now.

[Eug Wanker]

Originally Posted by osxisfun

TS is still sticking to their leaked specs.

Last bump before multicore?

Think Secret's followup article is here. Yeah, it does sound like a last bump before the 970MP.

It would make sense that a new 970MP-capable (and likely dual 970MP-capable) Power Mac motherboard would also have PCIe, and maybe DDR2 support.

If it's sometime in 2005, that's reasonable, but if it's Macworld January 2006, that's a bit too far behind the curve for my tastes.

[Zap2]

apple needs to make a call on the emac, update it or kill it off and make a cheaper 17'' or 15'' screen and hope the mini and the cheap screen make up for it. I hope they updat it to at least 1.42 or 1.5ghz

so the imac would be 1.8 or 2ghz,the powermac 2,2.3,2.7ghz,mini,1.25 or 1.42,emac would be 1.25ghz for schools and 1.5ghz

[Simon]

Originally Posted by Simon

Yeah yeah, it's the FX again. Ha, you guys are just not patient enough.
...
Everybody will be disappointed, as expected.

TS is driving the point home just for those who still had hopes.

No PCIe, no DDR2, no BlueRay, no Antares, not even AntaresSP. Looks like another round of the FX. I'm glad I'm not buying a Power Mac now.

[Eug Wanker]

Originally Posted by Simon

TS is driving the point home just for those who still had hopes.

No PCIe, no DDR2, no BlueRay, no Antares, not even AntaresSP. Looks like another round of the FX. I'm glad I'm not buying a Power Mac now.

Yup. It does make sense though as I said for Apple to release 970MP on a PCIe board (with PCI-X as well). May as well kill two birds with one stone. I'm not actually counting on the DDR2 though. DDR2 has latency issues, making it less desirable unless they jumped to 667 MHz right off the bat. There are even rumours that AMD might skip DDR2 altogether.

I don't see Blu-Ray any time in 2005 in a Mac. Hell, the next-gen format for HD isn't even finalized yet. Toshiba, Sony, and friends are still in discussion, trying to avert a format war. Apple, while a supporter of Blu-Ray, has already announced HD-DVD on DVD media support in DVD Studio Pro and DVD Player 4.6 (in Tiger).

As for AntaresSP/970GX, I'm wondering if they may simply be 970MP rejects, where one core has failed or runs at too low a clock speed. I'd really like to see the next gen iMacs go to L2 cache-doubled higher clocked G5s though, after this imminent revision. Maybe MW2006 for 970GX iMacs? I'll buy one then (unless I finally give in and buy one in the next few months ).

P.S. The US Apple Store just went from same day shipping to 3-5 business days today. AI's prediction of new Macs coming April 26 may just be right. And it's of note that it's only the Power Macs that are seeing the delay. The eMacs, iMacs, and iBooks still are at same day shipping.

[Luisc]

How much is the new Dual-2ghz PowerMac is going to cost?

[zed57]

anyone's guess.

[Todd Madson]

Grumble. I think we'll need to wait a week to figure on what the heck will really be shipping.

Makes it difficult - my Sawtooth G4 just gave up the ghost this morning (typing on a work
laptop they provide me).

Grumble.

[P]

Originally Posted by goMac

Well, my professor could have been a bit carried away. That said, you can fit quite a few chips on a wafer. It could probably get down to the $50-$100 range per chip. And a dual core vs. a single would still be the same price to produce (not counting r & d).

No, it would be almost twice as expensive. Not quite, if you do it right, because you can share some stuff, but dual cores have one L2 cache for each core, and that's the majority of the chip area - especially if you add in the actual execution units.