Intel spills beans on Yonah, the next notebook chip

Quote:

---

Originally Posted by Mikesta

OK so I am speculating, but I am going off some reasonable assumptions. The current Dothan's O/C extremely well as they are even with weak heatsinks/fans. With a proper socket (and also a proper heatspreeder) we'd be able to get better heatsinks or watercooling for these baby's. I am also assuming that 65nm will help in this regard. Lookng at the TDP these things aren't putting out much heat to begin with.

OH and BTW when will AMD have quad-core availalble on desktop? We will be extremely lucky to get that any earlier then very late 2006. Yonah will be out far before then.

I do assume that 65nm process will benefit A64, but at the moment 90nm Dothan > 90nm A64. Yonah will be a highly optimised revamped dual-core version of Dothan, so I'm willing to bet that 65nm Dothan > 65nm A64. It requires so much less power then a64 as well.

What I am waiting for is to see what/when K9 is going to be/out. K8 will not be able to compete with an optimised version of Dothan if the Intel Processor is not hamstrung my platform architecture.

As I have stated I am speculating but I believe I'm going off reasonable assumptions. Make of it what you will.

---

According to the AMD, quad core a64's will be quarter 1 2k6, not late 2k6.

You also have to remember, they're going to have to up the transistor count alot on the P-M to add in the features it needs. This will, inturn, up the power requirements of it, which will make for a hotter running chip... Now, they don't really need to add any more transistors to the A64 for quadcore as the chips were made with multiple cores in mind, thus the die-shrink will do nothing but lower power and heat. If they use low power transistors like they did in newark, they could probably match the power/heat of the yohan with ease.

This, too, is going off very reasonable assumptions, as well as more than a few cold hard facts.

It was noted before that Yonah will have HT. If so, the theoretical edge of QC AMD cpu's will shrink.

And don't forget that if Yonah's architecture will be more advanced than AMD's QC (as some will say that Dothan is more advanced than K8) - Yonah will definitly be with the upperhand.

Actually the first quad-core AMD will be Opteron. And it will come along with Socket F. Don`t think that AMD will release quad-cores for desktop so soon after that. Even Intel`s going to release their quad-core in early 2007. So I don`t think there is a big hurry to increase core quantity. Even now dual-core processors have performance boost only in professional apps like rendering or coding.

Dothan`s and Yonah`s architecture is based on P6 architecture that took place in PIII. So it`s not that new. It`s just had few improovements time-to-time like new tech processes and some new functions.

Quote:

---

Originally Posted by shklak

It was noted before that Yonah will have HT. If so, the theoretical edge of QC AMD cpu's will shrink.

And don't forget that if Yonah's architecture will be more advanced than AMD's QC (as some will say that Dothan is more advanced than K8) - Yonah will definitly be with the upperhand.

---

More advanced? It's a modified P3... It's definitely not more advanced, it's just that the P3 was intel's BEST chip hands down, and the P4 was their biggest mistake, thus they're bringing back what they need to.

As for HT, If you look at the P-D 840EE, you'll see that HT with dual core doesn't have much benefit, and in most cases NEGATIVELY impacts overall performance.

Cooper, of course they'll do quad core op's first, they also did dual core first op's first... They almost always start with the opterons before going to the desktop with everything. They haven't, however, made any claims that quad core will not reach the desktop, and thus we have reason to believe they will, and NONE to believe they won't.... Also, didn't amd say they have a "very special" FX chip for when multithreading becomes mainstream(think for a minute here)? Of course, this final sentence is purely a guess and nothing more, but it does make perfect sense as making the FX quad and regulars dual would be the ultimate market divider.

Finally, you must remember, next year releases Socket M... Neither you, nor I , nor anyone without an AMD name tag have any idea what to expect out of that. To assume that yohan will be more technologically advanced is pure heresay, and blatant fanboyism at it's finest. We don't know what will happen, but we do know both companies are going to come next round with a completely fresh twist, and that we'll all have to buy new EVERYTHING to run em..... :banana: :banana: :banana: :banana: 'ing technology!

Yes Socket M is surely a mystery for now. But about FX. This is purely gamers proc.(as AMD claims). Do you know any game that has an performance boost, playing it on dual-core platform ? I don`t. Heh and they`ll come only next year. Of course quad-core processors will go desktop. But IMO not next year. It`s just too costy and won`t have any performance improovement for avarage (or extreem) gamer. Even thought AGEIA is coming to market. Multycores are good for professionals needs-that`s their role. For now.
P.S. heh I think soon will be facing great problem: what to choose? Intel or AMD ?
i think it`s nice :)
P.S. what is QC :confused:

Cooper, AMD has stated it's saving a VERY special chip for when games go multi-threaded.. ALL games based on the UE3.0 engine will be multi-threaded. That alone would be enough to warrant releasing a multi-core fx... As for it being too expensive, they charge close to $1,000 for FX's anyway, so I don't see how they couldn't do it.

p.s. QC is Quad Core.

Quote:

---

Originally Posted by DilTech

According to the AMD, quad core a64's will be quarter 1 2k6, not late 2k6.

You also have to remember, they're going to have to up the transistor count alot on the P-M to add in the features it needs. This will, inturn, up the power requirements of it, which will make for a hotter running chip... Now, they don't really need to add any more transistors to the A64 for quadcore as the chips were made with multiple cores in mind, thus the die-shrink will do nothing but lower power and heat. If they use low power transistors like they did in newark, they could probably match the power/heat of the yohan with ease.

This, too, is going off very reasonable assumptions, as well as more than a few cold hard facts.

---

Mate I have to disagree on a number of things you have said in this and other posts.

I can't see the P-M transistor count needing to rise so much. Remember that the Yonah will have 2mb lvl2 Cache total (1mb per core).

Also are you sure that AMD will not need to add to their transistor count, or are you telling me that they will run quad-core off a 128 bit Memory controller? If so....No thanks AMD.

Also your assumption that A64 could match the power/heat of Yonah is crazy. No K8 processor currently in the ball park with Dothan regarding power/heat (also add size/cooling requirements). Not even the Turion's. Not the mention that a Dothan outperforms a Turion in general and would widen the gap if you make a performace / ( Power/Heat) ratio.

Before you launch into me or others about fanboyism, take a look at yourself. I am actually slightly biased toward AMD but I recognise that AMD dominance is under severe threat by P-M. Lucky for AMD that Intel is lazy and greedy (wants to cash in on Prescott based processors and maintain price premium of Dothan's) otherwise Intel could trash AMD in no time. Meanwhile AMD is being far too complacent, resting on their laurals. This is so frustrating to witness because I want good competetion.

Also I can't see what AMD is going to do to make a 'special' FX dual/multi-core chip. They have the K8 to work with. They vary it by having single/dual channel memory controllers, 512k to 1024K lvl2 cache, and frequency.

AMD will have to come up with the K9 to compete with the successor's of Dothan, and it seems like that is going to be in 2007 and no earlier. AMD has to pray that Intel doesn't get too aggressive.

Imagine if AMD does realise a Quad only for Intel to respond with a Quad-core Yonah based tech. WTF are AMD going to do? they will be up :banana::banana::banana::banana: creek. Feel free to debate my points.

I'm no fanboy either, my secondary pc is a p4 2.6C with an ATi 9800pro... The exact opposite of my main one ironically enough. My PC before that, a AXP 2100+, before that a p3 1 ghz.... I go with whatevers better at that time! Right now, it's ALL amd!

Now, without further ado...

Isn't K9 suppose to make it's appearance with Socket M? 2k6 marks the release of socket M with the DDR2 memory controller, 2k7 it jumps up to DDR3! There's an article about it at extremetech.com...

http://www.extremetech.com/article2/...1826595,00.asp

One part to look at is where they mention a special form of multithreading, where multiple cores will be able to share specific parts of the die yielding in an 80 PERCENT INCREASE in performance, but will cost an extra 50 percent in resources....

There are 2 new versions of Hypertransport coming up as well, we have no clue what these will do for the A64.

As for a "special" FX...Who knows, we obviously don't, all that's been said is that amd has a very special multi-core FX for when multithreading hits gaming. I don't work for them, but I'm sure it's more than just another 200mhz on the x2 4800+.... A good guess would be quad core, as it won't have to show up til games go multi-threaded, which will be next year.

Yonah is made for dual core, intel has stated to handle amd's quadcore they'll slap 2 yonahs together EXACTLY like they did their P-D... We can all see clear as day how that turned out.... Of course, that was with prescott cores, the worst core EVER made, so this might be a bit different...

Intel is actually going to have to add quite a bit to the P-M to make it desktop ready... Adding HT, 64-bit instructions, sse3, and a few other neccessities alone is going to up the transitor count by itself quite a bit, which inturn will raise the power requirements up quite a bit, as well as the heat output... This is a fact.

Yohan will have one other very large obstacle, and that would be desktop support. There's currently what, 2 mobo's and an adapter for only ONE mobo(which only supports ONE heatsink) for the P-M(which has been around for what...2 years?)? That'll HAVE to change, or the yohan still won't have a place in the desktop world, and if it does it's place will still be small except amongst those who change out their parts as often as they wash their laundary until they add 64 bit to them, which won't happen initially. They'll need it though as Longhorn will be out next year.

Personally, I'd LOVE to see intel do something big, they HAVE to, as netburst MUST go. Netburst has ALWAYS been a mistake, and intel should've stayed with the p3 instead of going for the p4. CPU's have been in a slump(besides multicore, which is currently pointless for anything non-professional) for quite awhile, since the release of the A64 there hasn't been a single leap forward besides a few speedbumps. This is because the competition hasn't been there, as you stated....

Now, with the upcoming K9 and intel dropping the netburst tech(which the first desktop based on the p-m should show up around 2k7) things are about to really pick up again. Hopefully intel comes correct, or amd will be taking ALOT more chunks out of their market share.

Diltech. I think your latest post here is quite reasonable.

However I was under the impression that K9 was going to be when a DDR3 platform is released in 2007? It appears that K9 will consist of a large shared Lvl3 cache. Other then that we have no info!

I sincerely hope they do a fair bit more then that. At least 1 single channel memory controller (64bit) per core will be a must. Personally I would like to see AMD reduce the lvl2 cache latencies of K8 and maybe lvl1 cache as well. This will benefit their processors greately! HT will also help a64 if implemented correctly.With its 128 total lvl1 cache HT will be a killer on a64's (just as it will be in 32+32 lvl1 cache of Dothan's/Yonah.

You see A64's and Dothan's are very similar beasts.

Both have Large lvl1 caches a64=64+64, Dothan 32+32 as opposed to p4's and most P3's 16+12 etc

Both have short pipelines relative to Prescott.

After this is where they diverge but they are basically trying to do the same thing. Increase processing power by reducing latencies.

a64 does this by including an ondie memory controller.

Dothan does this by including lvl1 and especially lvl2 cache with very low latencies. Dothan includes a large cache to ensure pipeline's are always filled.

If either processor took on the main aspect of the other, then we would be looking at a super processor. Yonah with on die mem controller or a64 with signiificantly reduced cache latencies would be monsters!

However in current form it seems that Intel's solution is actually superior to AMD's. Intel howver is happy to cripple it (so it can't effect sales of p4's which it is still trying to get a good return on investment from) and sell
P-M's for crazy prices.

Quote:

---

However in current form it seems that Intel's solution is actually superior to AMD's

---

Heh. Dothan is based on P6 architecture. Don`t think that this old style is better than K8 :) But it relly is better than NetBurst

Quote:

---

Originally Posted by Cooper

Heh. Dothan is based on P6 architecture. Don`t think that this old style is better than K8 :) But it relly is better than NetBurst

---

Man I'm alittle tired of people saying Dothan is based on PIII's.

Tell me how many PIII's had 32+32 Lvl1 cache? Or 1-2 MB of Level2 cache.
It is its own architecture. It's not like the son, or brother of the PIII, maybe a cousin.

If Netburst was not created and Intel just kept on using shrinkage to improve performance of PIII's it would not turn out like Dothan at all.

heh...if Intel woudn`t be at that way of big numbers and just kept improoving P6 it wouldn`t be like Dothan at all - it would be much more greater thing.
And architecture is not just caches capacity. I don`t think that you will deny the fact that Dothan is based on P6.

Quote:

---

Originally Posted by Cooper

heh...if Intel woudn`t be at that way of big numbers and just kept improoving P6 it wouldn`t be like Dothan at all - it would be much more greater thing.
And architecture is not just caches capacity. I don`t think that you will deny the fact that Dothan is based on P6.

---

Your not serious are you? If intel kept improving P6, say 130nm then 90nm process it would not be better then a Dothan (Even though it would likely be operating between 2-3ghz). P4 wasn't great granted, but it still isn't that bad. Northwoods are quite nice processors. Personally I wonder how a 90nm Northwood would have faired....

Now regarding architecture - caches are VERY important. the size and latencies make a big difference.

I won`t argue with you Mikesta - you are right. Just wanna speak my thoughts about K9. As you all know AMD is going to release it`s quad-core processor next year already - and it`s going to be server one. Though that innovations oftly comes in this sector there`s just might be that K9 will comes with this high-tech piece of silicon. Also it would be nice to see next to it DDR3 or XDR RAM, which is a lot better than DDRII. More than that these types of memory are already being produce. This could possible be quad-channel 128 bit compared to dual-channel 64 bit that DDR & DDRII has. Personnaly I don`t wanna see DDRII next to new models of AMD with it`s on-die mem controller.

Quote:

---

Originally Posted by Cooper

I won`t argue with you Mikesta - you are right. Just wanna speak my thoughts about K9. As you all know AMD is going to release it`s quad-core processor next year already - and it`s going to be server one. Though that innovations oftly comes in this sector there`s just might be that K9 will comes with this high-tech piece of silicon. Also it would be nice to see next to it DDR3 or XDR RAM, which is a lot better than DDRII. More than that these types of memory are already being produce. This could possible be quad-channel 128 bit compared to dual-channel 64 bit that DDR & DDRII has. Personnaly I don`t wanna see DDRII next to new models of AMD with it`s on-die mem controller.

---

I too am hoping that AMD gets off its high horse and starts scrambling to get a high quality innovative new product in a K9.

Slapping together a quad-core 65nm on K8 architecture is NOT innovative and I hope to God that it is NOT AMD's answer to Dothan's successor's.

I hope they will continue to push the envelope and continue to give Intel a run for their money.

Sorry to dig this back up...

According to the various sources I've seen, Yonah and Sossaman won't have:

* Hyperthreading (HT)
* EMT64

As many of us have seen with dual-core Pentium-D and XE models, HT doesn't offer much gain compared to dual-core itself. And there's little incentive in adding more complexity if it doesn't result as much gain as it does on the Pentium-M.

Remember what HT was originally for. To improve the number of instructions per cycle on the P4. Pentium-M is already there without HT.

Clock for clock, both the A64 and P-M solutions outgun P4 in most tasks. (Except for certain digital media applications).


As for EMT64 capability, well, that'll come in 2007. ;)


There is also no mention of a new feature with Yonah.

"Digital Media Boost"...That's the marketing name.

So what is it? According to http://www.theinquirer.net/?article=21549

It's a fusion of the SSE and SSE2 micro operations with some new streaming extensions and improvements to floating point performance and to to the SSE decoder throughput.

It addresses performance areas that are lacking in the current "Dothan" core. What can we likely to expect? Newer Pentium-Ms becoming more capable as P4 in nearly everything. (If not, you'll see single-cored Yonahs are much improved over their Dothan brothers).

There is also rumour that Yonah might get virtualisation or Vanderpool Technology. It would be great to run multiple OSs simultaneously like I do with VMware. ;) (at least there is no longer a need to spend $350 Aussie on VMware any more).

As many of you know, the Yonah is made on the 65nm process.
The process is manufactured at Oregon, USA, at Intel's D1D Fab facility.


With regards to the argument or comment about Pentium-M being a PIII.

Technically, you should view it as a hybrid design.

The Israeli-based Intel team that developed the Banias (1st Generation P-M) was given the task of developing a mobile platform. They chose the Pentium III as the basis, as at the time, it offered the best power to performance ratio compared to the P4 (Netburst).

Using the PIII as basis, many changes were made. The FPU was much improved, the cache was larger but it had far less latency than both P4 and A64, the P4 bus is used, and power saving technologies were implemented.

Essentially, what Intel did was take what was already good from the PIII and the P4 designs, and made changes to produce the Pentium-M.

It does have characteristics of the Pentium III, but it doesn't mean its a complete PIII design...What PIII uses the P4 bus? ;)

BTW, I've tested PIII-S 1.4Ghz "Tualatin" and a Pentium-M 1.4Ghz "Banias" against each other. The differences between them is very noticable. I compressed a 4GB file using WinRAR, and the difference is quite dramatic. (we're talking about an almost 50% difference).

Even when I overclocked my old Celeron "Tualatin" 1.2Ghz to 1.6Ghz, it still took an hour longer to compiler API tools I was using. (2 hours for my Banias based notebook vs 3 hours for Celeron at 1.6Ghz).

I was simply impressed with the Pentium-M that I didn't bother with the P4. None of the PCs I own are P4-based. They're either Pentium III, Celeron (PIII-based) and Pentium-M. (Ok, I also have a VIA EPIA PD10000, which uses a C3 "Nehemiah" CPU).


As for the AMD K9....I don't see much info on it.
All I know its gonna be made on 65nm, is likely to be a quad-core solution, it was last heard to run at 3Ghz, and development of the K9 started back in March 2003.

Given the nature of how things are progressing with AMD, I expect K9 to come in late 2007. (It typically takes 4 to 5 years from starting the project to until the first retail product comes).

Quote:

---

BTW, I've tested PIII-S 1.4Ghz "Tualatin" and a Pentium-M 1.4Ghz "Banias" against each other. The differences between them is very noticable. I compressed a 4GB file using WinRAR, and the difference is quite dramatic. (we're talking about an almost 50% difference).

---

Well WinRAR performance is very depended on cache size, which is way bigger in Dothan than in PIII.

Quote:

---

at least there is no longer a need to spend $350 Aussie on VMware any more

---

Do you use it on commercial purposes ? If not why you spend yr money on it ? ;)

Quote:

---

As for EMT64 capability, well, that'll come in 2007

---

As Intel represantatives had informed this feature will come as market will require it. I think it will be sooner than 2007.

Quote:

---

As many of us have seen with dual-core Pentium-D and XE models, HT doesn't offer much gain compared to dual-core itself.

---

That is cos of Windows XP problems - it cannot work fine with 4 threads at once. Soon there will be some patch to fix this problem.

Quote:

---

Except for certain digital media applications

---

Which are very depended on actual processor speed.

Quote:

---

There is also rumour that Yonah might get virtualisation or Vanderpool Technology

---

It is not a rumour - it is a fact. At least some more powerfull models will support it.