Intel's Dual Core Plan Overview

[raretech]

Anandtech has put together the information they have about Intel's dual core strategy into one article. If the information is accurate, it's worse than I thought for Intel. On the desktop and in servers at least.

<A HREF="http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2252" target="_new">http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2252</A>

<i>Nemo me impune lacesset</i><P ID="edit"><FONT SIZE=-1><EM>Edited by raretech on 10/22/04 08:36 PM.</EM></FONT></P>

 

[wolverinero79]

Um, 2 Prescott cores running at 3.2 Ghz? Ok, so we didn't get the Banias/Dothan speed stepping, but 3.2 Ghz for a dual-core chip? Dang!

So a minimum of 2 MB of cache (I think that's safe to assume), dual core running b/w 2.8 and 3.2 with 64-bit enabled that will be compatible with Microsoft's 64 bit OS's. What's so bad? Oh, and they use the same socket already available (so no mobo upgrades for all the early adopters who invested in Prescott already). Wow...

Granted, I don't like the initial chips being limited by the 800 FSB, but that's the way it goes. This could turn out to be really really nice or a very bad flop. And considering that AMD is attacking server dual-core rather than desktop initially, Intel has a chance to reclaim face in the desktop performance arena. A chance, looks like Intel's making a serious wager...

I'm just your average habitual smiler =D

 

[Spitfire_x86]

HOT!!!

------------
<A HREF="http://www.foood.net" target="_new">FOOOD's Icons</A>

<A HREF="http://geocities.com/spitfire_x86/myrig.html" target="_new">My Rig</A> & <A HREF="http://geocities.com/spitfire_x86/benchmark.html" target="_new">3DMark score</A>

 

[Mephistopheles]

Erm... a dual-core 3.2Ghz prescott versus a dual-core Athlon 64 at 2.0Ghz? That competition is exactly what's going on now, except it's with single-cored processors.

Plus, if those are little more than two prescotts slapped on a die instead of the much, much more interesting dual-core dothan possibility (even if clocked, say, at a "mere" 2.4-2.5Ghz), then this is not the news I was hoping for. It's not good.

Two prescotts on a die is too much of a bad thing. :frown:

Or at the very least too much of too little in the first place.

The 90nm A64 will kick a dual-core prescott's *ss anytime...<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 10/23/04 08:55 PM.</EM></FONT></P>

 

[Xeon]

Two Prescott’s on a die is too much of a bad thing.

Why with all those extra transistors in there it could very be they are for dual core operations, exceptionally efficient thread manipulation, resource balancing, threaded branch prediction, thread vectorizeing, shared microcode swapping, enhanced power management, enhanced system tick distribution, enhanced memory prefectch system to alleviate bandwidth limitations, shared SSE,SSE2 engines.

Any of those technologies could be implemented in there with regards that HT is disabled makes the logical assumption that there is a great deal of thread manipulation logic on those cores. Otherwise they are dual cores and with regards to the current performance war with AMD that is nearly improbable.

The 90nm A64 will kick a dual-core prescott's *ss anytime...

You sound like P4Man speculations and theories don’t make facts.

Xeon

<font color=red>Post created with being a dickhead in mind.</font color=red>
<font color=white>For all emotional and slanderous statements contact THG for all law suits.</font color=white>

 

[Spitfire_x86]

And remember, they're going to miss the performance boost from HT in some encoding/rendering apps, which saves current P4s from being a total loser.


------------
<A HREF="http://www.foood.net" target="_new">FOOOD's Icons</A>

<A HREF="http://geocities.com/spitfire_x86/myrig.html" target="_new">My Rig</A> & <A HREF="http://geocities.com/spitfire_x86/benchmark.html" target="_new">3DMark score</A>

 

[Mephistopheles]

You do have a point, I was highly speculative.

<speculation> (yes, again, beware!)
Actually, this whole discussion reminds me of the fact that prescott contains a gazillion dark transistors... Useless ones, as far as we know.

Now this is even more speculative, but maybe all that junk has a function... which has not yet been activated?...

And as speculative as it gets: <i><b>two northwood cores</b> would fit just fine on a single prescott die!!!</i> The transistor count is compatible with that theory. Prescott has 2.5x the transistor count that NW had... <b>(!)</b> Maybe that beast is already dual-core and that's why it dissipates 115W?...

Highly unlikely. But speculation is interesting nonetheless.
</speculation>

 

[Xeon]

HT helps more on multitasking and can help in threaded applications. I have seen apps that were threaded and it did nothing in fact sometimes hinder it.

The technology isn’t needed once there are 2 physical cores on the same chip. As well SSE and SSE2 gave the P4 the edge it needed in encoding and rendering more than HT ever did.

Also the statement the P4 being a total loser is almost malicious in nature. P4C's seemed to have throttled the living [-peep-] out of the AXP's. The A64 owes its dominance to the on die memory controller not the failure from Intel to make a competitive product.

I see dual Prescott’s actually performing quite well, if the load balancing is as good as a HT enabled processor, there could very well be some interesting performance wars approaching especially keeping in mind the clock speeds will be 2.8-3.2 is quite remarkable IMO.

Useless ones, as far as we know.

Trust me they aren’t useless transistors, lithography is a time consuming process and expensive as bloody hell. Anything on there has a purpose, if not they literally have some crack head engineers over there enjoying the six digit salaries.

And as speculative as it gets: two northwood cores

Highly unlikely since the core was revamped by specialized IC layout software with load balancing, electrical balancing, clock signal balancing, and machine execution balancing in mind. Prescott’s are as similar to a Northwood core as I am to a black person.


Xeon

<font color=red>Post created with being a dickhead in mind.</font color=red>
<font color=white>For all emotional and slanderous statements contact THG for all law suits.</font color=white>

 

[raretech]

The bus is what makes it bad for me. You can say "that's the way it is" and I say "yeah, that's why it's worse than I'd thought it would be."

AMD is being smart about targeting workstations and servers first. There is already demand there as well as a plethora of apps ready to benefit from the technology. Targeting the desktop with this first doesn't make a whole lot of sense.

Except maybe for gaining bragging rights in a couple of benchmarks for desktop apps.

<i>Nemo me impune lacesset</i>

 

[Xeon]

Well how long has HT been activated in P4's? Long enough to guarantee OS and most apps will be able to take advantage of dual processors. The desktop move works well on quantity for them as well. Lowering cost of ownership for workstation and HPC users in the end, I fail to see a disadvantage to that.

Xeon

<font color=red>Post created with being a dickhead in mind.</font color=red>
<font color=white>For all emotional and slanderous statements contact THG for all law suits.</font color=white>

 

[raretech]

Well how long has HT been activated in P4's? Long enough to guarantee OS and most apps will be able to take advantage of dual processors.

Did you just say MOST apps? Are you a comedian?

The desktop move works well on quantity for them as well. Lowering cost of ownership for workstation and HPC users in the end, I fail to see a disadvantage to that.

You think they're going to move a high quantity of those at first? Define quantity and I might be willing to offer you a friendly wager.

This chip is going to have little impact on the workstation market and ZERO impact on the HPC market. The bandwidth limitations alone, compared to other alternatives, would make this a piss poor choice for HPC applications compared to existing alternatives. I guess we'll see what the pricing is on this, that could help. But uh, I'm not one to hold my breath for Intel to come out as the value leader with a new chip.

<i>Nemo me impune lacesset</i>

 

[Atolsammeek]

Problem we had in the past in the 1980s with 16bit and 32bit taking 10 years to have 32bit take over.

I think we are going to have the same problem with Dual cores and AI 64 bit.

I for one am glad of the HT. For I can more then one program at once. Where 10 Websites would lag my desktop system. The hypertreading shares tasks.

I think Intel looking at it this way. We went from Dual core Motherboards (two cpu) To hypertreading acts as two chips and help in speed. Back to dual cores but on one chip.

It like this

1980s 1990s Dual Core Motherboards
2002 Hypertreading
2005 Dual core cpu
???? Dual core cpu with hypertreading. two differnt ideas
1. The dual core cpu with ht. Will run a 3 HT cpu of shared info.
2 The dual core cpu with ht. Will run 2 hypertreading cpus.

I could be wrong I could be right. But at least is something to think about.

 

[Spitfire_x86]

Hyperthreading does nothing to improve light to average multitasking. My non-HT system doesn't slow down when I have 10 websites open. I can use everything without problem while encoding video. The secret: I give encoding "idle" priority. HT may help somewhat when you're encoding video, rendering things at 3DSMax and playing Doom3 at once, but since it's not real SMP system, performance loss will be very obvious. In that case, suppose a non-HT system gets 10 fps in Doom3, and HT system gets 15 fps. 50% improvement, but both are practically unusable.

The main HT benefit was it made P4 faster in encoding/rendering apps. For multitasking, it's mainly a marketing gimmick. HT may provide some real multitasking beneifit if you're trying to run a second OS through VMWare/Virtual PC, but I haven't seen how much improvemnt these apps get from HT. But it's safe to say, it's definately worse than SMP/Dual Core systems.


------------
<A HREF="http://www.foood.net" target="_new">FOOOD's Icons</A>

<A HREF="http://geocities.com/spitfire_x86/myrig.html" target="_new">My Rig</A> & <A HREF="http://geocities.com/spitfire_x86/benchmark.html" target="_new">3DMark score</A>

 

[Xeon]

Did you just say MOST apps? Are you a comedian?

Most was supposed to come before OS's I made a grammar error, thank you again for catching that.

You think they're going to move a high quantity of those at first? Define quantity and I might be willing to offer you a friendly wager.

Well they are selling two at a time I really can’t see them not selling an immense amount of CPU's like they already do. Also they will be the first to have them for desktop machines which will drive their sales up as well.

This chip is going to have little impact on the workstation market and ZERO impact on the HPC market.

Ya go figure a desktop processor sure will impress workstation users and HPC users.

The bandwidth limitations alone, compared to other alternatives, would make this a piss poor choice for HPC applications compared to existing alternatives.

It has yet to be finalized what the final bandwidth of the Xeon based cores and the Itanium based cores will be like.

With regards to the both of them I don’t see Intel castrating them further. Bandwidth maybe less important if the cores work together and not independent of each other.

Xeon

<font color=red>Post created with being a dickhead in mind.</font color=red>
<font color=white>For all emotional and slanderous statements contact THG for all law suits.</font color=white>

 

[trooper11]

that is not reflected in any tests ive seen. today, basiclaly hyperthreading does nothing to make the p4 a better choice except in some encoding/3d apps and of course in multitasking. and while the multitasking part is nice, in reality, most users wont take advatnage of it becuase most ppl dont run doom3 while encoding video and running 3dmax in the background. thats why HT didnt solicit much app development, and thats why even when dual cores come, app development wont be overnight unfortunately, but at least faster then it has been so far.

now i dont expect the first p4 dual core to be a big hit, there are just too many questions right nwo for any of us to be able to say it will be good or not, but most first generation new products arent the best the product can offer later down the road. alos, i think amd will have an edge in the first generation becuase of thier day one architecture for dual core, but itnel cna easily get bakc into the game. im excited to see what happens

 

[trooper11]

"Also they will be the first to have them for desktop machines which will drive their sales up as well."

wait, your saying intel will be first to market with dual core for the desktop before amd? last i heard, intel stated 2006 for its dual core desktop and amd has stated 3Q-4Q 2005. did you see something different?

im sorry, that 2006 date is for the server/workstaion dual core parts, intel seems to still be saying the end of 2005 of dual core desktop. but that still means amd and intel will be at the same time, no lead time for intel.
<P ID="edit"><FONT SIZE=-1><EM>Edited by trooper11 on 10/24/04 10:19 PM.</EM></FONT></P>

 

[raretech]

Ya go figure a desktop processor sure will impress workstation users and HPC users.

I disagree. Not when dual core opterons will be on the market , targeting those very segments.

It has yet to be finalized what the final bandwidth of the Xeon based cores and the Itanium based cores will be like.

We were talking about the dual core scotties I thought, that is the chip that Intel will be targeting the desktop with first. That is, assumning the information available is accurate.

<i>Nemo me impune lacesset</i>

 

[Mephistopheles]

Correct me if I'm wrong, but while Intel and AMD will have dual core parts out by the same time, those will not be targeted at the same market niche:

Intel: mainstream desktop initially
AMD: Opteron initially, then Athlon FX, then mainstream desktop

These are the latest plans I saw. So Intel might actually be the first one to release a dual-core chip. That still doesn't mean it won't be patchwork. And again, patchwork is still not to say it won't be interesting. It just makes it more unlikely.

 

[Mephistopheles]

Based on up-to-date reports and information:

I suspect Xeon's bandwidth will continue to be stupendously low. It won't match HT-like integration anytime soon.

On the other hand, it has already been stated that Montecito, the next big thing in IA64, will feature a point-to-point bus running at 667Mhz (128=bit wide) for a total of 10.6GB/s. Because it's point-to-point, though, system bandwidth will more than triple by the time Montecito gets introduced.

A lot of people are paying close attention to Montecito. A successful montecito might just manage to get IA64 to a respectable position... but if it's just another speed and small feature-bump in the Itanium line, IA64 will continue its downwards spiral.

OS support is also crucial, but that still depends on a lot of things... If Montecito turns out to be anything close to the paper tiger it currently is, the scenario will probably change a bit.