Intel : Prescott successor canned? Pentium-M is the future?

[imgod2u]

So what could we expect to see from a Desktop Banias/Dothan CPU?

400/533/800 FSB? What about implementing HyperTransport?
SSE3?

If they used Hypertransport, they'd need another channel for memory access (such as a dedicated memory controller and channel). I doubt they'd do that and Hypertransport is not very suitable for memory access.

Improved efficiency Dynamic Branch Predictors (similar to ones installed on Prescott)?

Already in there. The advanced branch prediction algorithms in Prescott were in Banias first.

Increased size of Branch Target Buffer (similar to what AMD with with Opteron)?

BTB's will only get you so far. I think Banias has branches handled pretty well.

Increased size of Global History Counter?
Optimized scheduler improvements (A LA Prescott)?

Currently, Banias's execution backend isn't that strong (it's very similar to that used in the P6 core), so I doubt the limitation is scheduling. The bottleneck probably lies in the execution units available.

Increased number of execution units (bring it back up to par with AMD's Athlon)?

Not if you still want to keep it low-powered. The 9-way 9-issue back-end on the K7 is complete overkill in most cases. Considering that Banias, with a 6-way, 5-issue backend achieves about the same performance. Too much and you'll just leave idle units there to waste power. Netburst probably had the right execution backend. Put that on Banias and you'll be set.

High speed 512 KB or 1024 KB L2 (shorter pipeline, faster cache needed)?

Has that already. Shorter pipeline would pretty much put it where the P3 was. Stuck at ~1.4 GHz. Banias wasn't designed solely for clockspeed like Prescott was, but that doesn't mean having more isn't a good thing.

Could be really interesting to see what the desktop version of Banias/Dothan brings to the desktop CPU market.

Currently, it seems like it's just the Banias core we have now, at 90 nm, on a dual-core. Intel doesn't have enough time to redesign everything if they're only starting now (which it seems like they have). Which brings up the question of how to deal with Banias's sub-par FPU performance.

 

[stmok]

I'm sure Intel can whip out something to remedy Banias's FPU performance...Or they've completely lacked the innovation nowadays?

Haven't they done it with the Dothan?

(Banias's Successor that is using 90nm process...45W of heat though, which caused delays in its release...Compared to the current Banias which are less than 28W).

If Intel fails...Well, there's always the AMD alternative...I have yet to invest in their products yet. (holding out for that Dual core Opteron.)

 

[SLee]

Haven't they done it with the Dothan?

(Banias's Successor that is using 90nm process...45W of heat though, which caused delays in its release...Compared to the current Banias which are less than 28W).

21W is the expected TDP for 400MHz FSB Dothans, which isn't too bad given the additional 50+ Million transistors and 300MHz increased maximum clock speed.

 

[stmok]

Hmmm, but for a desktop role, it won't be too much of an issue would it? Since heat requirements are not as strict as mobile platforms. So I'm sort of contradicting myself here...

This is what I'm sort of hoping...If Intel improves FPU performance while still making the Pentium-M (or future derivative) acceptable in thermal levels for both single- and dual-cores, then everything should go well.

(acceptable = not requiring 1 kg HSF just to cool the bloody thing!...Something of reasonable size and weight that it doesn't fall off when you take it to a LANparty. Low noise at max load.)

If Intel's engineers can accomplish those, hell, I'll be pleased to invest further in an Intel product...(in addition to an AMD product).

But if they do a sloppy job...I'm gone. I've waited long enough for them to shake off that ridiculous "P4 fever" they have. (Non-P4 buyer here).

 

[Audioaficionado]

The rumor mill reports that AMD’s .09 micron die shrink will meet more traditional expectations for a die shrink, which is to say that AMD’s .09 micron chip will run cooler, draw less power, and scale well.

I cannot imagine how Intel R&D did not know the P-4 Core would have this scalibility issue when it was deployed and for them to have missed it is a big sign that Intel is in trouble OR just cares nothing about it's customers.

Imagine all the P-4 Boards that are going to be left in the junk drawer becaue the entire chip platform has to be scrapped ? LOL..

I would be so mad at Intel right now if i owned one,that The filters would be having trouble keeping up with me. What is really sad is how DUMB Intel must be to not have been ready to phase out the P-4 smoothy with an introduction of a replacement line. DURRING THIS ANNOUNCMENT..............

Wow and to think people have been brainwashed so well as to not see this as a kick in the butt from a giant manufacture simply amazes me.......................if I was an INTEL OWNER I WOULD JUMP TO AMD SO FAST THE PRESHOTT WOULD HAVE NO TIME TO COOL DOWN...Come on here people, look at what Intel just did ?
Is no one really mad at Intel for this surprise bombshell ?

I'll spring for a dual Opteron system once they come down in price and PCI-Express gets fully deployed. Future DDR-II isn't that importaint to me as DDR gets the job done nicely now.

I don't think they did this purposely, but honestly thought they'd solve the problems like they always did in the past. This time the laws of physics and quantum mechanics stopped them dead in their tracks.

None of my future P4 purchases will be in a junk drawer. Now that P4s are going to get cheap, I'm buying a lot of them and making a folding farm of my dreams. P4Cs spank any AMD XP processors when it comes to folding and multimedia apps. My 3GHz P4c Northwood spanked my XP3200 Barton all day long every day.

I have a very nice dual Prestonia Xeon system that is P4 based. Eventhough my Iwill DH800 is built for those PressHott like Noconas, I'll just stick with overclocking my Prestonias past 800fsb and be very happy. When I feel the need to go dual 64bit CPUs, I'll just get an Opteron setup.

These are just business decision by both intel and AMD and you shouldn't take it so personally.

 

[CPL.Luke]

didn't the P3 replace the p2 like less than two years after its release.

Anyway, The new G5's are 90 nanometer from IBm now aren't they so Does anyone know how those are doing as they should be a good indicater of amds 90 nm because they use the same process.

 

[Gregory_WE]

They aren't doing well. At first, they projected that they would be able to pump out 3GHz G5's, but now they're hoping that they can even get 2.5GHz this year.

 

[funnyperson1]

didn't the P3 replace the p2 like less than two years after its release.

Anyway, The new G5's are 90 nanometer from IBm now aren't they so Does anyone know how those are doing as they should be a good indicater of amds 90 nm because they use the same process.

Yes but the PIII was also based on the P6 core, in fact the difference between the PII and PIII wan't that much different architecturally.

 

[MetalStorm]

Yeah, but by going Dual core don't you need software / OS to utilize them?

Longhorn.

 

[mamisano]

Longhorn.

I am aware that LongHorn will be multi-processor compliant, but we need applications to be written for them too.

 

[stmok]

As long as they're multi-threaded, you'll see some noticeable improvements...Otherwise, you notice your system being able to handle higher loads. (Heavy Multi-tasking)...Games are gonna be the sticking point as dualie CPU setups today don't perform well under this scenario. (Unless the game is re-written and optimised for SMP use).

I doubt the "feel" of a dual core would be any different from today's Dualie CPU setups. (I own a few dual setups, so having a dual-core in the future is not gonna feel too much different from what I have today...Sure it'll be 1Ghz faster, but the overall "feel" of responsiveness under high loads is gonna be about the same.

 

[MetalStorm]

I am aware that LongHorn will be multi-processor compliant, but we need applications to be written for them too.

Sorry, that was more aimed at the OS part of your question, I'm guessing though as the hardware comes out, software will be written to support the new features like SSE and SSE2, though I guess it will be more of a novelty like 64bit computing until it has a large user base and software is seriously written for it.

 

[diehrd]

QUOTE:From stmok
"Games are gonna be the sticking point as dualie CPU setups today don't perform well under this scenario. (Unless the game is re-written and optimised for SMP use").

That is totally untrue. For many reasons but most of all is All modern video games are Vide card dependant and any dual from 1 gig up with an agp slot and a good video card will game perfectly fine.

I started the below thread because there is a misconception about a dual processor system and gaming. What is funny is this use to be the oposite misconception where people thought 2 CPU would double there performance in a game lol........Is there a performance difference ? yes in FPS but in real life play the user will not experence anything less then a smooth gaming enviroment


http://www.ocforums.com/showthread.php?s=&threadid=293259

 

[GotThaJuice]

stmok, I loved your chip wars

So, I guess that the fact that Intel decided to make insane pipelines just so that they could put a 3GHz sticker on their retail boxes has come back to bite they right in their asses.
OK, so does this mean that my P4P800 will soon become mature for a museum or will any of these desktop mobile derivatives fit in mPGA-478 (I wish )?

It's kind of ironic that chips with the most potential and efficiency are rotting inside laptops. Dangerously undervolted, without fast memory for a companion and worst of all: non-overclocked.
And meanwhile, I am using my PC as a radiator during the winter. Splendid .

 

[PsychiatristNET]

i've been told that the inquierer has some issues with false news statements?

 

[JaY_III]

Pentium Pro
Pentium II
Pentium III
Pentium M

All based on the same core.
P6 lives on....
now only if i can get a Pentium M to work in my BX board

 

[Jimbob7]

I don't know if this link has already been posted or not, but its pretty cool

http://support.intel.com/support/processors/sb/CS-010854.htm

 

[Captain Newbie]

Yes but the PIII was also based on the P6 core, in fact the difference between the PII and PIII wan't that much different architecturally.

Main differences between Pentium II and Pentium III:

1) Different manufacture process (.25 mics to .25/.18mics)
2) 70 new instructions, aka SSE or KNI
3) The stupid, stupid, stupid unique ID number flaw.

I'd like to see what innovations Intel will have for us. It should prove very interesting.

 

[Cowboy X]

Not to mention more cache ( at least initially ) and the ATC or advanced transfer cache that came with coppermine .