Llano A8 and A6 for sale

[podspi]

Exactly. Draw a line in the sand for at least the FX-4110 (maybe the A8-3850 beats the eventually released FX-4100, but that is more about saving dies).

D: That would be incredibly disappointing, unless the FX-4100 is very, very low power (and low clocks).

 

[Topweasel]

D: That would be incredibly disappointing, unless the FX-4100 is very, very low power (and low clocks).

Low clocks. From what I can gather Low 2k to mid/high 2k Turbo. If the the IPC for Husky is 6% better then Stars, and BD is ~10% better then Stars, and at Turbo its probably running around 200MHz or more slower, then yeah they would probably be pretty close together. Remember the 4110 is already priced at about $150, a 4100 would be priced around $125 or less depending on how BD pricing moves after launch.

 

[Bman123]

Im waiting for a mc combo deal before I buy

 

[drizek]

Let's think about this...why would AMD seemingly cap the upper-end of the performance an end-user can extract from AMD's lower-budget platform parts?

The answer is self-evident, is it not? To prevent sales cannibalism of the Bulldozer SKU's that were supposed to be in the channel right now...

We all talk about the top-end Zambezi part in the forum threads but there are numerous lesser-performing Zambezi parts planned.

So what does that say about Bulldozer performance?

The only potentially good interpretation of this is that their 32nm is so good that an uncapped Llano can hit really really high clocks.

 

[drizek]

D: That would be incredibly disappointing, unless the FX-4100 is very, very low power (and low clocks).

It is not low power. I think it is a 95W TDP, which means the cpu portion will use more power than the cpu portion of Llano.

 

[podspi]

Low clocks. From what I can gather Low 2k to mid/high 2k Turbo. If the the IPC for Husky is 6% better then Stars, and BD is ~10% better then Stars, and at Turbo its probably running around 200MHz or more slower, then yeah they would probably be pretty close together. Remember the 4110 is already priced at about $150, a 4100 would be priced around $125 or less depending on how BD pricing moves after launch.

Honestly, I'm expecting average IPC to increase by more than that (not necessarily sustainable peak, though).

If quad-core K10.x can compete with quad-core BD at anywhere near equal speeds, that bodes poorly for BD...

 

[LOL_Wut_Axel]

Honestly, I'm expecting average IPC to increase by more than that (not necessarily sustainable peak, though).

If quad-core K10.x can compete with quad-core BD at anywhere near equal speeds, that bodes poorly for BD...

I have to agree with this. I'm expecting a 20% jump in IPC to be satisfied, which would make it comparable to Nehalem. If that happens, then the FX-6110 will be able to match the Core i7 2600K (same situation as Core i7 970 vs. 2600K) overall, being a bit faster in multi-threaded programs and a bit slower in single-threaded ones--but as of now those are pretty much limited to audio encoding.

 

[AtenRa]

http://developer.amd.com/assets/apu101.pdf?elq=9a578abfcd244faca0f02e65418670b8&elqCampaignId=44

 

[Idontcare]

So what does that say about Bulldozer performance?

The only potentially good interpretation of this is that their 32nm is so good that an uncapped Llano can hit really really high clocks.

Let's hope this is the case.

Although I have to say I was rather taken aback by the power-consumption numbers in the AT articles.

I was thinking the 100W TDP for the 2.9GHz part was more just to add a ton of padding to the TDP spec and that the chip would actually come in more around 65-70W. (meaning the 2.9GHz would have had lots of thermal headroom in my magical world of fairy's and unicorns...but alas that does not appear to be the case, if any conclusions can be reliably based on a sample size of one)

But definitely they could have enabled a 3.5GHz turbo-core on the 2.9GHz model and still fit within the 100W footprint. Why they did not is the basis of speculation here, only rational reason I can conclude is because they need to keep Llano from stealing Bulldozer's thunder in the low-end SKU's.

AMD is filled full of bright savvy people, I'm betting this is all strategy and not a reaction to something that needed fixing by castrating the turbo-core feature.

 

[drizek]

Maybe they were afraid of the SEC stepping in if they were performing too much better than Intel.

The 120W number I believe was for the entire system, so I think the real TDP for Llano is 100W or maybe a little lower.

Disable the graphics part and you are down to 65W tops, that gives you around 80W of thermal headroom with a good cooler.

 

[MACMAC]

That's the reason why they showed that they were able to over-clock but didn't show over-clocking performance when both the CPU and GPU are over-clocked.

Just to clarify, the reason why we didn't show any overclocking performance numbers, and in fact tried to make it clear that our overclocking results weren't final or even stable, is that we literally only spent 30 minutes overclocking this new platform.

We were writing the review with one hand and overclocking the Llano system with another, that's literally how time constrained we were.

We would never try to hide issues from our readers, so if we find a TDP cap or poor GPU performance when overclocking, we will let you know

 

[mnewsham]

Just to clarify, the reason why we didn't show any overclocking performance numbers, and in fact tried to make it clear that our overclocking results weren't final or even stable, is that we literally only spent 30 minutes overclocking this new platform.

We were writing the review with one hand and overclocking the Llano system with another, that's literally how time constrained we were.

We would never try to hide issues from our readers, so if we find a TDP cap or poor GPU performance when overclocking, we will let you know

Thanks for the clarification

 

[Idontcare]

Welcome to the forums MAC_HWC :thumbsup:

 

[LOL_Wut_Axel]

Just to clarify, the reason why we didn't show any overclocking performance numbers, and in fact tried to make it clear that our overclocking results weren't final or even stable, is that we literally only spent 30 minutes overclocking this new platform.

We were writing the review with one hand and overclocking the Llano system with another, that's literally how time constrained we were.

We would never try to hide issues from our readers, so if we find a TDP cap or poor GPU performance when overclocking, we will let you know

Alright. Thanks for clarifying.

Almost all the reviews I've seen mention that Llano has a TDP cap that gives higher priority to the GPU and that if you try to surpass it, performance is lowered.

Think of it like this: if you want to get raise CPU performance, you need to downclock the GPU so that it has more thermal headroom, and vice versa. The GPU is still given priority, though, so you'll have more thermal headroom for over-clocking that.

That's what I've understood from what I've read.

 

[busydude]

Alright.

Almost all the reviews I've seen mention that Llano has a TDP cap that gives higher priority to the GPU and that if you try to surpass it, performance is lowered.

Think of it like this: if you want to get raise CPU performance, you need to downclock the GPU so that it has more thermal headroom, and vice versa. The GPU is still given priority, though, so you'll have more thermal headroom for over-clocking that.

That's what I've understood from what I've read.

That is exactly how I pictured it as well.

 

[pantsaregood]

Is the TDP cap really hardwired?

I would expect it to be circumvented in some way. Just getting up to 130W would really give a lot to work with.

 

[jvroig]

Count me in as one of the people interested in knowing if this TDP cap cannot be circumvented (through a bios setting or anywhere else). As I understood it from the AT review, there was no way to do this.

Nice to hear from someone at HWC, it's one of the sites I do like aside from AT.

 

[MACMAC]

Count me in as one of the people interested in knowing if this TDP cap cannot be circumvented (through a bios setting or anywhere else). As I understood it from the AT review, there was no way to do this.

Nice to hear from someone at HWC, it's one of the sites I do like aside from AT.

Let me put it to you this way,

I can very confidently state that no one, aside from the small inner circle that designed Llano, knows everything there is to know about these chips.

Shockingly, even the motherboard manufacturers are as clueless as us reviewers right now with regard to certain aspects of this new processor family, and I am pretty confident that additional capabilities will be unlocked (or at least improved) within the next few weeks.