Llano - when will we start seeing reviews?

[SickBeast]

By several accounts, AMD is currently shipping Llano APUs to manufacturers. When do you guys think we'll start seeing reviews?

Also, how do you think it will perform relative to a K10? Is a 10% boost clock-for-clock too high of an expectation?

I think we'll see them hit 4.2ghz or so, and I think they'll turn out to be awesome budget CPUs for the majority of people who only need basic graphics. As a matter of fact, they'll probably make good laptop CPUs as well.

 

[Vesku]

I think it will be similar to Zacate and we won't see reviews until the first early laptops ship. Must be frustrating for the AMD engineers to see OEMs so cautious about launching products that have so far turned out to be very solid competition to Intel in their market space.

 

[dma0991]

I doubt we'll be seeing review anytime soon because I see the Llano would be at it's best in the notebook market. Till an OEM announces they have a new Fusion laptop, we're not going to get a glimpse of Llano's performance.

 

[Tuna-Fish]

Also, how do you think it will perform relative to a K10? Is a 10% boost clock-for-clock too high of an expectation?

10% more IPC? Probably yes. IPC isn't what I'm waiting for from Llano. By all accounts, the process AMD is using for phenoms is quite pathetic when compared to the Intel ones (it's more comparable to Intel 65nm than Intel 45nm, and Intel has had 32nm out for how long now?), while the upcoming 32nm process will be more or less comparable with Intel 32nm.

So, expect a huge increase in performance/watt, mostly from running at higher clock speed at lower power consumption.

 

[Vesku]

Thuban 45nm doesn't seem like Intel 65nm to me, process tech is mostly separate from IPC. Seems like cost/process might be part of the reason behind AMD's cache disadvantages to Intel's but there are design differences that make it hard to figure out.

 

[SickBeast]

10% more IPC? Probably yes. IPC isn't what I'm waiting for from Llano. By all accounts, the process AMD is using for phenoms is quite pathetic when compared to the Intel ones (it's more comparable to Intel 65nm than Intel 45nm, and Intel has had 32nm out for how long now?), while the upcoming 32nm process will be more or less comparable with Intel 32nm.

So, expect a huge increase in performance/watt, mostly from running at higher clock speed at lower power consumption.

Isn't Llano still made on the 40nm node?

 

[AtenRa]

Isn't Llano still made on the 40nm node?

32nm SOI HKMG

Bobcat is on TSMC 40nm

 

[SickBeast]

32nm SOI HKMG

Bobcat is on TSMC 40nm

If that's the case, then Llano is a much more interesting CPU than I thought it was.

 

[Tuna-Fish]

Thuban 45nm doesn't seem like Intel 65nm to me,

Look at the data I linked -- the GF 45nm is faster than Intel 65nm, but not by very much. GF 32nm should be almost as much better than GF 45nm as Intel 32nm was better than Intel 65nm.

process tech is mostly separate from IPC.

Well, yes. Your point?

 

[Vesku]

Look at the data I linked -- the GF 45nm is faster than Intel 65nm, but not by very much. GF 32nm should be almost as much better than GF 45nm as Intel 32nm was better than Intel 65nm.


Well, yes. Your point?

"IBM’s 45nm SOI process with high-k dielectrics and metal gates is very impressive, nearly matching the performance for the high performance 32nm processes. However from the description was more research oriented and will probably never go into production. "

^ from your link, isn't Thuban on that high-k process? My point is that the actual performance difference between 45nm Intel and AMD cpus is partly from an IPC lead that Intel has and that saying it's barely better than Intel 65nm is a bit hyperbolic.

 

[Idontcare]

isn't Thuban on that high-k process?

Nope. Thuban is on the same doped-poly nitrided-gate 45nm SOI CMOS that the rest of AMD's 45nm CPU's use at GloFo.

All the high-k metal gate stuff at 45nm that AMD talked about never made it into production. Slideware at its finest.

 

[jimbo75]

Nope. Thuban is on the same doped-poly nitrided-gate 45nm SOI CMOS that the rest of AMD's 45nm CPU's use at GloFo.

All the high-k metal gate stuff at 45nm that AMD talked about never made it into production. Slideware at its finest.

Thuban included low-k dialectrics which was why it had such an improvement in power draw. I don't remember high-k ever being talked about except in some theories by Dresdenboy, which he later admitted was wrong.

 

[Idontcare]

Thuban included low-k dialectrics which was why it had such an improvement in power draw. I don't remember high-k ever being talked about except in some theories by Dresdenboy, which he later admitted was wrong.

Thuban includes the same low-k dielectrics that all the other 45nm GloFo CPU's include. Low-k is for the BEOL (metalization levels).

Athlon II, Phenom II, Thuban...all the same process tech. Same xtors, same metalization, same low-k dielectric.

http://www.xbitlabs.com/news/cpu/di..._45nm_Microprocessors_with_HKMG_Incoming.html

The high-k stuff was explicitly advertised by AMD as an "optional feature for 45nm". People got the impression that AMD was probably going to roll out a "45nm Phase 2" with HKMG because that was what AMD talked about.

There was a pdf hosted by AMD that talked about it in some detail in Q/A format, it is now conveniently missing, so I will link to a web article that talks about it, although it was talked about quite a lot so I'd think you'd know of it or easily come up to speed on it with a bit of googling.

We have high-k+metal gate as an option at 45 nm; we haven't announced any products that will use high-k+metal gate," spokesperson Gary Silcott told us. "It will come later on in our 45 nm process, so our initial 45 nm process does not include that."

http://www.betanews.com/article/AMD-moves-cautiously-into-the-45-nm-era/1204662746

 

[Vesku]

They definitely gave me the impression that they were introducing HKMG into their matured 45nm products, they need to transfer that marketeer to direct sales.

They did manage to get some better OCing performance from the C3 stepping regardless.

 

[podspi]

By several accounts, AMD is currently shipping Llano APUs to manufacturers. When do you guys think we'll start seeing reviews?

Also, how do you think it will perform relative to a K10? Is a 10% boost clock-for-clock too high of an expectation?

I think we'll see them hit 4.2ghz or so, and I think they'll turn out to be awesome budget CPUs for the majority of people who only need basic graphics. As a matter of fact, they'll probably make good laptop CPUs as well.

I bet LLano's IPC is going to come in around Phenom II's. It has some tweaks that all else equal will increase IPC (so I've heard?) but will be hurt by the lack of L3 cache. Its turbo modes should be interesting.

 

[StrangerGuy]

First thing I wanna know about Llano is...Is there hybrid crossfire?

 

[jimbo75]

The high-k stuff was explicitly advertised by AMD as an "optional feature for 45nm". People got the impression that AMD was probably going to roll out a "45nm Phase 2" with HKMG because that was what AMD talked about.

There was a pdf hosted by AMD that talked about it in some detail in Q/A format, it is now conveniently missing, so I will link to a web article that talks about it, although it was talked about quite a lot so I'd think you'd know of it or easily come up to speed on it with a bit of googling.


http://www.betanews.com/article/AMD-moves-cautiously-into-the-45-nm-era/1204662746

Well tbh that looks a lot more like it's saying low-k will be used instead of high-k, which is what happened. Leaving something open as an option isn't quite the same thing as making a definitive statement. Certainly nothing was said about it since details of Thuban were announced in mid-late 2009 and nobody was under the impression that it was definitely coming with high-k, just that there was a very small chance of it happening.

As for Thuban, it didn't use the same low-k, it used a lower-k according to Hans de Vries http://www.xtremesystems.org/forums/showpost.php?p=4330334&postcount=48 and it was also mentioned in a lot of the reviews on release day.

I'm guessing this is the pdf you were talking about - http://forums.overclockers.com.au/showpost.php?p=9099422&postcount=16 (you're right it's been removed)

Again it should be pretty clear that in no way did AMD state that HKMG was definitely coming at 45nm, just that it was an option, and that ultra low-k would be used later on the process.

 

[Soleron]

Thuban includes the same low-k dielectrics that all the other 45nm GloFo CPU's include. Low-k is for the BEOL (metalization levels).

Athlon II, Phenom II, Thuban...all the same process tech. Same xtors, same metalization, same low-k dielectric.

"AMD had another ace up its sleeve when designing Thuban. The folks at GlobalFoundries have made tweaks to their 45-nm silicon-on-insulator process, adding a low-k dielectric layer to reduce leakage power. The result? Within a given thermal envelope, AMD can achieve nearly the same clock speeds with six cores as a Phenom II X4 did with four cores."

http://techreport.com/articles.x/18799/1

 

[Fox5]

Low-k = reduced power consumption, high-k = better performance, right?

 

[Gigantopithecus]

I don't think we'll start seeing Llano reviews until products are shipping. I hope Llano returns AMD to competitiveness in the laptop segment. If it's priced right, and its integrated GPU is sufficiently powerful, it could very well make Intel's i3 and below CPUs irrelevant, much like Zacate made Atom irrelevant.

 

[nonameo]

I don't think we'll start seeing Llano reviews until products are shipping. I hope Llano returns AMD to competitiveness in the laptop segment. If it's priced right, and its integrated GPU is sufficiently powerful, it could very well make Intel's i3 and below CPUs irrelevant, much like Zacate made Atom irrelevant.

I don't know if I would go that fast so quickly. Intel's marketing and name is enough to make it relevant, even if it's selling saltines with traces on them.

 

[Gigantopithecus]

I don't know if I would go that fast so quickly. Intel's marketing and name is enough to make it relevant, even if it's selling saltines with traces on them.

Ha, true, true. Plus Atom's market penetration is already very high. However, for those who know (like us), the E-350 is so superior to D525 that there's no reason to consider the latter anymore aside from cost (i.e. D525 would be more compelling if substantially cheaper). Similarly, it would be nice for AMD (and those of us who appreciate competition) to make the less capable Pentiums and Celerons obsolete in the laptop segment.

 

[Tuna-Fish]

If it's priced right, and its integrated GPU is sufficiently powerful, it could very well make Intel's i3 and below CPUs irrelevant, much like Zacate made Atom irrelevant.

It won't. SNB's integrated GPU is simply good enough for a huge chunk of the market. (While almost everyone who bought an Atom would be willing to pay more for more performance on the same power envelope.) Llano does, however, seem to be ready to return AMD to competitiveness on laptops and low-end desktops.

 

[podspi]

First thing I wanna know about Llano is...Is there hybrid crossfire?

It should have hybrid crossfire, if this is to be believed:

http://www.xbitlabs.com/news/other/...d_Outlook_Along_with_Launch_of_New_Chips.html

and

http://news.softpedia.com/news/AMD-...logy-Similar-to-Hybrid-CrossFire-179740.shtml

Though I know these aren't the best sources, it makes a lot of sense. Would give AMD a huge leg up in laptops with discrete GPUs, as well as help add-ins on lower-end desktops.

 

[wahdangun]

to be honest i really expecting AMD would use sideport memory like thier current integrated GPU but use GDDR5 instead, so it won't be bottlenecked by dual channel ram config