RENAMED - Intel Yonah review thread

[Mats]

Anandtech:
Preview Part I.
Preview Part II.
Notebook performace review.

A preview by Matbe, translated to English here.

[ckolivas]

Look at that glorious power consumption. /me can't wait for Merom dual core for all this and more

[Mats]

Well, I'd be impressed if they can go lower than AMD's 35 W TDP. The only problem is to actually find one at 35 W though, thanks to AMD's variable thermal data... I don't like it.

[afrost]

that looks pretty sweet......finally some competition for AMD next year in the performance per watt area.

Hopefully this competition will get us some kick ass 65nm Athlon 64s some time next year.

[~El~Jefe~]

a64 wont be better than this. I wish it were so, but I am doubting that. a64 is better for several reasons in heat and such, but this is a short pipeline dothan with 2 megs cache total.

dothans presently when overclocked go just about 50% over what their base speed is while still being cooled by a zalman or whatever. this 2.0 oc'd is what im interested in personally.

that could be the fastest chip out.

[Aris]

i just hope they dont make boards with proprietary heatsink mounting holes on these new ones. what good is a kickass cpu with low power consumption if the heatsink/fan is either 1. not adequit, or 2. too loud, and you cant replace it with anything better.

[Ackelind]

i just hope they dont make boards with proprietary heatsink mounting holes on these new ones. what good is a kickass cpu with low power consumption if the heatsink/fan is either 1. not adequit, or 2. too loud, and you cant replace it with anything better.

Of course they will make it like that. It seems like they are struggeling to make one component as good as posible, so they can make the other cheap and bad, and still end up with something mediocre in the end. Like switching to square wheels when getting a really powerful car.

[/irony off]

I sure would be good to see. Seems like they are still using 478-pin but changed the pin layout AGAIN! Was it not enough with the 478/479 mess?
Intel seems to be releasing more sockets and chipsets than AMD is releasing processors. Every time there is a new processor, there is a new socket and ten new chipsets.

[stromgald]

Well, I'd be impressed if they can go lower than AMD's 35 W TDP. The only problem is to actually find one at 35 W though, thanks to AMD's variable thermal data... I don't like it.

Lower than 35W TDP? I think Intel already achieves that with its notebook processors, and AMD only gets down to that level with its Turion line. The newer Intel Pentium-Ms can get down to 25W or so. IMO, This Yonah is for the most part, a dual-core desktop processor advertised as a notebook processor. The processor in the article should only be compared to the A64 X2 or the current Pentium 4 dual cores. The Yonah 'L' and 'U' versions all promise to be less than 25W according to Intel's roadmap. There should also be single core Yonahs that will probably be more common in notebooks.

MikeC: Shouldn't it say 'first half of 2006' not 2007? Also, you spelled 'for' with two 'r's. A little eager to get the good news out?

[Aris]

I sure would be good to see. Seems like they are still using 478-pin but changed the pin layout AGAIN! Was it not enough with the 478/479 mess?
Intel seems to be releasing more sockets and chipsets than AMD is releasing processors. Every time there is a new processor, there is a new socket and ten new chipsets.

mabey they do it because there losing ground on their CPU sales. if they change sockets on a new cpu, it requires people to not only upgrade a chip, but now a whole new board (which are prodominantly using intel chipsets btw). also when you think about major distributors of notebooks like dell and hp, it will require them to buy all new hardware from intel, and not just new cpu's that they can plug into their pre-existing socket 479 notebooks that can be purchased at a lower cost than their newer chipset motherboards.

[asitic]

Correction: The Yonah's will be available beginning of 2006 not 2007.

I want one of these for my living room HTPC.

[afrost]

a64 wont be better than this. I wish it were so, but I am doubting that. a64 is better for several reasons in heat and such, but this is a short pipeline dothan with 2 megs cache total.

dothans presently when overclocked go just about 50% over what their base speed is while still being cooled by a zalman or whatever. this 2.0 oc'd is what im interested in personally.

that could be the fastest chip out.

The A64 has been out for a long time now. If you think that AMD doesn't have any new stuff they have been working on all this time, I would like some of what you are smoking.

I am not an AMD fanboy.....I would buy a Yonah IF the price is right. All that I'm saying is that AMD finally has some performance competition that will force them to release some new stuff. A straight die shrink to 65nm will give them a boost in clock speed and lower power consumption. The graphs show that the Yonah and current 3800 X2 are similar performers......so when "X2" gets its 65nm process it comes down to which one can clock higher and which platform doesn't cost an arm and a leg.

[MikeC]

Several points:

Intel has had an operational dualcore P-M in their labs since the spring. My guess is that Anand got one of these. We have no idea how close it is to a production product. Maybe Intel doesn't either. What we do know is that Intel has once again deflected attention from its current lineup and put the focus on a product that won't be on the market for at least another quarter. Very clever for Intel, pitting a current x2 product from AMD against a future x2 product from Intel.

AMD engineers are surely not sitting back with martinis in their labs. They have to be working in anticipation of future competition from Intel. I would be floored if AMD did not already have something in their labs that offers lower power consumption & at least maintains performance.

Finally, Yonah is not even slated for the desktop. So how far off are Menon and other desktop-specific "high efficiency" Intel dualcores?

In other words, take the Anand article for what it is -- a preview of a future potential product against an existing product. in 3~6 months, through the transition from lab to mass production, much can change, including the competition.

[smilingcrow]

My initial reaction was of the Wow nature. But, there are so many unknown variables in this article that it’s really hard to see what’s going down. I wonder how much the power hungry AMD chipsets are distorting the picture! Were CnQ and Speedstep enabled? Either way, it seems obvious that dual core on 65 nm is going to be excellent for silent types, be it Intel or AMD. At the platform level, I think Intel may have an advantage though.

[Mats]

Well, I'd be impressed if they can go lower than AMD's 35 W TDP. The only problem is to actually find one at 35 W though, thanks to AMD's variable thermal data... I don't like it.

Lower than 35W TDP? I think Intel already achieves that with its notebook processors, and AMD only gets down to that level with its Turion line. The newer Intel Pentium-Ms can get down to 25W or so. IMO, This Yonah is for the most part, a dual-core desktop processor advertised as a notebook processor. The processor in the article should only be compared to the A64 X2 or the current Pentium 4 dual cores. The Yonah 'L' and 'U' versions all promise to be less than 25W according to Intel's roadmap. There should also be single core Yonahs that will probably be more common in notebooks.

1 - T64 MT-40 (2.2 GHz) seems to be the coolest running CPU today with it's 25 W (I think they will make a 2.4 GHz/25 W too). PM 780 runs at 2.26 GHz and 34 W (taken from cpuheat.wz.cz).

2 - I was talking about dual core of course, the coolest running dual core AMD I've seen (don't know all models) runs at 2.2 GHz and 35 W (Opteron 175), the problem is just to find them since you first have to start it up to know the TDP. Intel on the other hand have a 24 W max for their "L" models, and a 1.66 MHz for max speed. Above that they go 49 W with "T", god knows what speed it will end up with, maybe 2.66 GHz.

So AMD is desktop/workstation/server, 90 nm, 2x1 MB L2 cache, 2.2 GHz, integrated memory controller (needs power) and still only 35 W.
Intel Yonah is mobile, 65 nm, 2 x 1 MB cache, either 1.66 GHz and 24 W or something higher on both, no integrated memory controller, and remember - that's Intel's TDP which always is lower than AMD's.

I'm not impressed of Yonah, AMD will take the lead when they go 65 nm since they still are already competitive when comparing their 90 nm desktop to Intels upcoming 65 nm mobile. I don't know how long Anandtech actually had that CPU, but they said a month, which means two months before release. And why did Intel raise the L2 cache latency?

Too bad Anandtech didn't tell us the TDP of that X2 3800+, it could be anywhere between 30 and 100 W (dont know the exact numbers since AMD have no new data for A64 available AFAIK).

[dragmor]

And why did Intel raise the L2 cache latency?

I think Anand had the right idea on why the cache latency was raised. Basically the dynamic nature of the cache (i.e. complete turning off of the cache blocks when not in use) required a slower cache. So intel went for power over speed.

I cant remember wether it was Yohan or Mermon, but one of the die shots I've seen had the L2 cache place all over the CPU instead of in a large block.

[Ironic]

2 - I was talking about dual core of course, the coolest running dual core AMD I've seen (don't know all models) runs at 2.2 GHz and 35 W (Opteron 175), the problem is just to find them since you first have to start it up to know the TDP..

35w is a number I find hard to believe for a double core amd.
The pentium m and yonah have been thought from the start with power consumption in mind, it's not the case with the opteron.
Why would a x2 3800+ (1.8hz, 512ko cache) need more energy that a 2.2 ghz (1mo cache) opteron?
It would have to run on a very very low voltage.

[Mats]

35w is a number I find hard to believe for a double core amd.

Well you don't have to. Some people just can't believe that there's no more low power than PM. Why? I don't know. Maybe the Centrino marketing guys made a good job.

The pentium m and yonah have been thought from the start with power consumption in mind, it's not the case with the opteron.

How do you know that about Opteron for a fact? Opteron was the first server/workstation CPU ever made that lowers clock speed and Vcore when idle, I'd call that "thought from the start with [low] power consumption in mind". Ok, it wasn't enabled until revision D showed up, revision B and C didn't have it.

Why would a x2 3800+ (1.8hz, 512ko cache) need more energy that a 2.2 ghz (1mo cache) opteron?

I don't know, ask AMD. X2 3800 runs at 2 GHz. It's quite difficult to understand the TDP of A64E, it's not one simple number in a list, and not in the OPN on the lid anymore. You have to use the CPU to be able to find out the TDP. Max TDP is just misleading.

It would have to run on a very very low voltage.

It runs at 1.3 V AFAIK.

[Ironic]

35w is a number I find hard to believe for a double core amd.

Well you don't have to. Some people just can't believe that there's no more low power than PM. Why?

Maybe becasue in every single test they have seen comparing the A64 (or turion) with the pentium m, the intel have been ahead in terms of power consumption?

I still believe the 35w is a typo.
Maybe amd will reach that with a revised core and the 65nm process, but at the present time, I don't buy it.

[stromgald]

35w is a number I find hard to believe for a double core amd.

Well you don't have to. Some people just can't believe that there's no more low power than PM. Why? I don't know. Maybe the Centrino marketing guys made a good job.

Why would a x2 3800+ (1.8hz, 512ko cache) need more energy that a 2.2 ghz (1mo cache) opteron?

I don't know, ask AMD. X2 3800 runs at 2 GHz. It's quite difficult to understand the TDP of A64E, it's not one simple number in a list, and not in the OPN on the lid anymore. You have to use the CPU to be able to find out the TDP. Max TDP is just misleading.

It would have to run on a very very low voltage.

It runs at 1.3 V AFAIK.

The 35W number is for the max TDP of some opterons at its lowest P-state, which I think means when it is stepped down to its lowest level, hence the lower frequencies at lower P-states (i.e. Cool'n'Quiet). You have to look at the max TDP and the highest P-state to make an accurate comparison. However, there is a series of Opterons (140, 240, 840) that do have a max TDP of 30W, but they're rated at 1.4Ghz and runs at 1.15V. Its most likely single core also. Most of the Opterons have a TDP of 55-115W.

The Opteron 175 you mentioned is listed as a single core running at 1.3V and 110W in that same document.

Too bad Anandtech didn't tell us the TDP of that X2 3800+, it could be anywhere between 30 and 100 W (dont know the exact numbers since AMD have no new data for A64 available AFAIK).

The 3800+ X2s are Manchester core, which have a TDP of 95W. The older X2 Toledo cores, which range from 4200+ to 4800+, have a TDP of 110W.

[Mats]

Well as I said, you don't have to believe me if you don't want to. I'm not talking about that a 95 or 110 W Max TDP figure, not about EE or HE models running at 30 or 55 W, and not about lowest P-state. You obviously missed the part about Thermal Profile in the PDF (page 10) and the XS thread, and assuming that I'm wrong.

For instance, my CPU have a Max TDP of 67 W in Thermal Profile A (as in worst case scenario, the highest TDP a CPU like this can have, but it can be lower.), but when I checked the TDP for this individual CPU it was 61.8 W (not big difference though)

[Mats]

Maybe becasue in every single test they have seen comparing the A64 (or turion) with the pentium m, the intel have been ahead in terms of power consumption?

First of all, I've never seen a review showing the actual TDP for the CPU they're testing, so it could be as high as 95 for a X2 3800+ for instance. Most reviewers look at total power draw, and intels mobile chipsets uses less power.

I still believe the 35w is a typo.

The application for checking the TDP is quite new, but there are both people getting those low numbers and it's documented by AMD. It is most unlikely a typo (or five typos ).
Dual core Opterons seem to use less power than X2, the opposite to when comparing single core Opterons and A64. I still haven't seen a dual core Opteron review.

[Mats]

stromgald:I see you have two A64. If any of them is revision E, please download the application from here and tell us what TDP you have. If you use it on an older CPU you just get the max TDP value, which I believe is 67 W for yours.

This is a new type of program, which checks AMD64 CPUs for their maximum rated case temp or TCaseMax and reports the TDP which corresponds to the processor in question.
On most AMD processors TCaseMax and TDP values are constant, and depend on CPU type and model, but on the E Revision chips, both this values and the default voltage are variable.
These are the chips that have variable values:
Athlon64 (Venice, San Diego)
Athlon64 FX (San Diego)
Athlon64 X2 (Manchester, Toledo)
All Rev E Opterons and Dual Core Opterons
What this all means is this: On 90nm manufacturing process, the power leakage of transistors on individual processors differs greatly, this is why AMD implemented TCaseMax. Each CPU has a different TDP rating...

Results can be found here, starting from about page 9, and here.

[perplex]

...What we do know is that Intel has once again deflected attention from its current lineup and put the focus on a product that won't be on the market for at least another quarter...

i think 5 weeks from now

[Mats]

Yeah, January 1 2006 is the day AFAIK. So strange that they're replacing it only like 6 - 8 months later. Maybe because they need a new desktop CPU and see no reason not to update the mobile one too (even though it's not old by then).

5 weeks, but how long do we have to wait for motherboards? I think it will take less than last time (was it like 18 months or more?), they've got a bigger market now, and Merom will work with them too I think.

Even though I don't think anything will be changed on the CPU they release compared to the one Anandtech got (2 months older) I expect motherboards to become better, maybe showing some better results.

[vertigo]

Well new DDR2 AMD chips will need a new motherboard with socket M2. It could be that Intel wants to get people buying new boards before those cpu's come out.

I don't think people upgrade motherboards all that often, so the early bird will surely catch the worm.

[perplex]

Yeah, January 1 2006 is the day AFAIK. So strange that they're replacing it only like 6 - 8 months later. Maybe because they need a new desktop CPU and see no reason not to update the mobile one too (even though it's not old by then).

i dont hear anyone complaining

[~El~Jefe~]

I dont think that intel will ever lose the ground that it did so long as it stays with Dothan/Tualatin processors. My piii 1.2ghz is an insane processor that has very little heat and blew away the p4's until 2.0ghz was exceeded in the p4 line.

A dothan can oc into the 2.8 ghz range and show faster fps compared to fx57 (anandtech.com) in games like doom3, and that on air cooling. I dont like intel at all, however, I dont see amd winning this. the turion dual core will be the test for this I guess.

[stromgald]

AMD won't win because Intel is bigger (higher volume), has contracts with most major computer manufacturers (i.e. Dell), and has the edge on advertising. Right now, the performance/price gap is enough to make people and companies think of switching, but with the new Yonah and future processors, Intel will probably stay close to AMD in price and performance. With similar products from two different companies, customers will usually choose the larger, more reputable (from advertising) company. To win AMD has to consistently beat Intel's products by a large margin like they're doing now.

[accord1999]

1 - T64 MT-40 (2.2 GHz) seems to be the coolest running CPU today with it's 25 W (I think they will make a 2.4 GHz/25 W too). PM 780 runs at 2.26 GHz and 34 W (taken from cpuheat.wz.cz).

No P-M uses 34w in any situation.

http://babelfish.altavista.com/babelfis ... /page4.php

Between idle 800MHz/0.988v and full load 2.13GHz/1.34v with cpuburn, the P-M has difference of 17w.

[Mats]

No P-M uses 34w in any situation.

Do you even know what TDP is?

Oh and Welcome to SPCR!