PDA

View Full Version : Intel Haveing trouble with a consumer 64bit processor?


zulgand04
Feb 23, 2003, 05:14 PM
the other day while bored, i saw an artical about intel not comeing out with a consumer type 64bit processor till at least late 2010 i think, due to they belive the consumer market does not need the advantages of it.
Now is this Intel actully beliveing it won't sell or is this covering a problem they have producing these chips?

heres the link to the artical
http://news.com.com/2100-1001-985432.html


-Neal

howard
Feb 23, 2003, 05:28 PM
very interesting article, i'm sure they've got something though. maybe it is a little later than the competitions and there just say oh theres no need for a 64 bit chip just cause they can't get one out there fast enough. only time will tell....

zulgand04
Feb 23, 2003, 06:08 PM
thats what i figured, or even because of the ghz rateing on their 32bit chips, their 64bit chips arnt even close to being close to being 3ghz+.

if thats true, thats kinda funny how their screwing them selves


-Neal

Jaykay
Feb 23, 2003, 06:08 PM
Im sure Intel have enough money to push out a 64bit processor sometime soon, it probably just wont be up to scratch or may be missing a load of features but thats intel for ya.....

DavPeanut
Feb 23, 2003, 06:28 PM
The Itaniums are only at 1 Ghz, and I would expect that consumer processors would be even slower, so they would have trouble getting the majority of users to buy them.

ddtlm
Feb 23, 2003, 06:54 PM
Intel is saying that because they don't want to make one, not because they can't. If Intel made a low-end 64-bit machine (presumably x86-64) they would shoot down their own Itanium.

DavPeanut:

The Itaniums are only at 1 Ghz, and I would expect that consumer processors would be even slower, so they would have trouble getting the majority of users to buy them.
Clockspeed and 64-bitness are not related. The Itanium is low-clocked because of other features, including its short pipeline (about the same length as a 7455) and because it is still fabbed on old-school 180nm tech.

Jaykay:

it probably just wont be up to scratch or may be missing a load of features but thats intel for ya.....
Why don't you name a single feature that Intel is missing? Why don't you name me an Intel processor line that is not up to scratch?

Catfish_Man
Feb 23, 2003, 07:00 PM
Late 2004? I had heard 2010. 2004 would be an EXCELLENT schedule for Intel, since I have yet to hear of a 64 bit consumer chip from them, despite hearing about their roadmap up until at least 2005.

zulgand04
Feb 23, 2003, 07:08 PM
sorry bout that i'll fix that i wasnt sure if i was right bout that till i reread the artical. u are right bout 2010.


-Neal

Shrek
Feb 23, 2003, 07:50 PM
IMHO, I think that Intel should just drop out of the chip market and do what they do best, and that is to make networking products. ;)

MisterMe
Feb 23, 2003, 09:17 PM
Originally posted by Jaykay
Im sure Intel have enough money to push out a 64bit processor sometime soon, it probably just wont be up to scratch or may be missing a load of features but thats intel for ya..... Let us not forget that for years now Intel has been trying to develop a mass-market 64-bit processor to replace its ancient x86-based product line. A lot of people on this forum forget that IA-64 was intended to replace--not supplement--IA-32 and HP PA. The Itanium is the best that it has been able to do after the better part of a decade. Intel has retargeted Itanium toward engineering workstations, but it was not always thus.

Catfish_Man
Feb 23, 2003, 09:57 PM
Originally posted by MisterMe
Let us not forget that for years now Intel has been trying to develop a mass-market 64-bit processor to replace its ancient x86-based product line. A lot of people on this forum forget that ISA-64 was intended to replace--not supplement--ISA-32 and HP PA. The Itanium is the best that it has been able to do after the better part of a decade. Intel has retargeted Itanium toward engineering workstations, but it was not always thus.

Deerfield (one of the next Itanium revisions) is supposed to use 65 watts of power (although it runs at 1GHz and "only" has 1.5MB of on chip L3 cache). That's a good deal lower than the Pentium 4. I think in another few years, IA-64 will have the price/performance/power usage to go up against x86 and win, the nasty part will be getting everyone to give up backwards compatibility. Perhaps they're working on an x86 emulator for IA-64? (The Itanium series has an x86 core on die, but it's about equal to a Pentium MMX, so it's not much use)

MisterMe
Feb 23, 2003, 10:25 PM
Originally posted by Catfish_Man
Deerfield (one of the next Itanium revisions) is supposed to use 65 watts of power (although it runs at 1GHz and "only" has 1.5MB of on chip L3 cache). That's a good deal lower than the Pentium 4. I think in another few years, IA-64 will have the price/performance/power usage to go up against x86 and win, the nasty part will be getting everyone to give up backwards compatibility. Perhaps they're working on an x86 emulator for IA-64? (The Itanium series has an x86 core on die, but it's about equal to a Pentium MMX, so it's not much use) The x86 emulator is the problem from where I sit. When Apple switched from the 680x0 to the PowerPC, it did not insist on on-chip 680x0 instructions. Apple was content to add an awesome 680x0 emulator to the Mac's ToolBox ROM. It worked with Motorola and IBM to reduce the POWER chip set to a microprocessor. Intel chose HP's Precision Architecture as the basis of its RISC effort. Had Intel been satisfied with reducing the HP PA to a single chip, leaving the x86 emulator to software or firmware, it is likely that the chip would have been ready years ago and into its fourth or fifth generation by now.

FelixDerKater
Feb 24, 2003, 09:04 AM
Originally posted by MisterMe
The x86 emulator is the problem from where I sit. When Apple switched from the 680x0 to the PowerPC, it did not insist on on-chip 680x0 instructions. Apple was content to add an awesome 680x0 emulator to the Mac's ToolBox ROM. It worked with Motorola and IBM to reduce the POWER chip set to a microprocessor. Intel chose HP's Precision Architecture as the basis of its RISC effort. Had Intel been satisfied with reducing the HP PA to a single chip, leaving the x86 emulator to software or firmware, it is likely that the chip would have been ready years ago and into its fourth or fifth generation by now.

That is why the Power4 is so great. It already accepts 32-bit instructions along with 64-bit instructions.

barkmonster
Feb 24, 2003, 09:22 AM
Why don't you name a single feature that Intel is missing? Why don't you name me an Intel processor line that is not up to scratch?

The Pentium 4!!!

Pentium 4 Denormalised Bug (http://phonophunk.phreakin.com/p4denormal.html) - This makes the CPU spike up to 100% usage whenever the level of audio drops below a certain threshold.

This (http://www.emulators.com/docs/pentium_1.htm) very detailed article has tons of info about why the Pentium 4 isn't the wonder CPU Intel would have people believe. It details a few things that DEFINATELY make it 'not up to scratch' and 'missing features'

1. The Pentium 4 has no barrel shifter, it's a circuit which can shift or rotate any 32-bit number in one clock cycle. Every cpu from the 386 and 68030 onwards has had one of these.

2. The L1 cache is tiny, smaller even than the cache on a 486!

3. Only one machine language instruction can be decoded per clock cycle. Again, this is a pre 486 thing and not something an efficient CPU would do.

I could list more but reading the article explains it all. The article isn't biased towards any platform, in fact it entirely focuses on CPUs that are compatable with x86 code. The only mention of PowerPC or 68K chips is for background details on chip design.

Can you believe it was motorola who started using pipeline stages first ? that comes as quite a shocker. Maybe they could be the ones with a 3Ghz Chip these days if they'd carried on in that direction with the PowerPC instead of focusing on low power and altivec.

Catfish_Man
Feb 24, 2003, 10:41 AM
Originally posted by barkmonster
The Pentium 4!!!

Pentium 4 Denormalised Bug (http://phonophunk.phreakin.com/p4denormal.html) - This makes the CPU spike up to 100% usage whenever the level of audio drops below a certain threshold.

This (http://www.emulators.com/docs/pentium_1.htm) very detailed article has tons of info about why the Pentium 4 isn't the wonder CPU Intel would have people believe. It details a few things that DEFINATELY make it 'not up to scratch' and 'missing features'

1. The Pentium 4 has no barrel shifter, it's a circuit which can shift or rotate any 32-bit number in one clock cycle. Every cpu from the 386 and 68030 onwards has had one of these.

2. The L1 cache is tiny, smaller even than the cache on a 486!

3. Only one machine language instruction can be decoded per clock cycle. Again, this is a pre 486 thing and not something an efficient CPU would do.


Actually, some of those aren't so bad.
1) bad
2) The reason why it's so small is so that it can be accessed in 2 cycles. No other L1 cache has been able to do that to my knowledge. The next P4 (Prescott) will have a 16k L1 cache instead of an 8k L1 cache.
3) This is because decoded instructions are stored in the L1 cache, removing the need to decode them again. It's one of the P4's major (only?) innovations.

4) There's another bad one. There's an instruction called FXCH that has been free (takes 0 cycles to execute) on previous x86 processors. It's very very useful to have it free, and the Pentium 4 doesn't.

Basically, the original Pentium 4 (Willamette) sucked big time. It was released prematurely due to the competetion from AMD. The current Northwood Bs are decent, and Prescott looks positively dangerous (800MHz bus, double L1 and L2 cache, even more optimized for high clock frequency, special threading instructions, etc...). The 970 is going to have a lot of work to do keeping up with Prescott, but I think it stands a chance. If the 970 and succesors can keep up until 2005/2006, the 32 bit Intel chips won't be able to address enough memory to keep up with them (assuming included memory doubles every year, high end computers will be shipping with upwards of 4-8GB by then).

dongmin
Feb 24, 2003, 11:33 AM
I can't say I'm an expert on CPU design but Intel seems to have a good point. If being able to address more than 4GB of memory is the chief advantage of a 64-bit chip (as implied in the article), then yeah, I don't see much point of it in consumer machines.

Being the first to the market with a consumer 64-bit machine seems to be purely for bragging rights. I always thought the attraction of the IBM 970 was its fast bus, scaling potential, etc. and not its 64-bitness.

Catfish_Man
Feb 24, 2003, 12:19 PM
Originally posted by dongmin
I can't say I'm an expert on CPU design but Intel seems to have a good point. If being able to address more than 4GB of memory is the chief advantage of a 64-bit chip (as implied in the article), then yeah, I don't see much point of it in consumer machines.

Being the first to the market with a consumer 64-bit machine seems to be purely for bragging rights. I always thought the attraction of the IBM 970 was its fast bus, scaling potential, etc. and not its 64-bitness.

For the next few years, that's certainly true. However, if current trends continue, 4GBs is going to be pitiful by 2010.