Bulldozer delayed til Q3?

What kind of strategy is it to delay your products? May be they are just waiting for the next recession so they could go bust...

They could also be trying to get enough Llano chips ready for shipping, since it would be ridiculous to launch LLano, yet not meet the demand.

Just a possibility.
 
They could also be trying to get enough Llano chips ready for shipping, since it would be ridiculous to launch LLano, yet not meet the demand.

Just a possibility.


So its better to sell $70-90 Llanos than $800 a piece BDs ?

Interesting.
 
The top end bulldozer chip was supposed to cost only a little more than the i7 2600K not $800.


Why ship BD to desktops when you could ship it where it is most needed, servers ? There the price range is $300-1500. AMD's market share in servers is 6% , much less than in desktops.
heflys20 said:
Llano ranges to a $150. And yes to your question, particularly if BD costs $800.

You're comparing end user price to what AMD charges. FYI, the Llanos shipped in Q2 had an ASP of $70 as per Mercury Research info.

Secondly, BD die size is in the same ballpark as Llano. Why would it cost 15x more to produce ? The difference is probably less than 20% ( more expensive packaging and more testing ).

high end desktop chips are laughable sales to the big 2.

High end is 5% of the volume and 90% of the profit. Look at AMD how much money is making without a presence in the high end. You can alternavely compare their ASPs. High end does matter.
 
You're comparing end user price to what AMD charges. FYI, the Llanos shipped in Q2 had an ASP of $70 as per Mercury Research info.

You're talking about the ultra-portable model, right? What about the upper mobile and desktop parts? What's the pricing on those, since they're only just coming out now.

Secondly, BD die size is in the same ballpark as Llano. Why would it cost 15x more to produce ? The difference is probably less than 20% ( more expensive packaging and more testing ).

Who said anything about it costing 15x more to produce? I thought the argument pertained to what was more profitable? Regardless, I imagine the difference being more than 20%. Thus, I doubt they're manufacturing Bulldozer for dirt cheap prices.

High end is 5% of the volume and 90% of the profit. Look at AMD how much money is making without a presence in the high end. You can alternavely compare their ASPs. High end does matter.

Well, that makes sense if AMD doesn't have much of a competitor and offers a superior/competitive product; not to mention overcoming Intel's strong branding. With Llano, AMD doesn't have any real opposition; hence, higher demand.

Scratch that, I'm skeptical that high-end (stuff over $500) makes up 90% percent of a company's profit.
 
Last edited:
Scratch that, I'm skeptical that high-end (stuff over $500) makes up 90% percent of a company's profit.

It might if you consider server sales. In the desktop market I bet it's the lower mid-range stuff that get packed in "beige boxes" that make most of the profit. The high-end stuff is marketed so that the Average Joe will go to the store and say, "I want one of them there Pentiums. No, the $300 one!"
 
Where does the myth that Llano doesn't have oppossition come from ?

Anand benchmark clearly shows than Intel+nvidia optimus can offer similar performance at nearly identical power consumption levels.

http://www.anandtech.com/show/4444/amd-llano-notebook-review-a-series-fusion-apu-a8-3500m/7

And that's without having their cpu performance cripled by ancient phenom II architecture running at up to 2 Ghz.

Optimus sucks. It's OS-dependent (not supported in linux), really flaky, generally overpriced, and a pretty half-assed implementation.

It's not a real competitor. Do you own an optimus device? They're pretty much junk and I'd expect them to be unsupported pretty fast as the idea gets abandoned going forward. The implementation as it is now needs to be changed for it to be well adopted.
 
You're comparing end user price to what AMD charges. FYI, the Llanos shipped in Q2 had an ASP of $70 as per Mercury Research info.
.

APU's != Llano, but Llano=APU.

The APU's they shipped in Q2 they are referring to the Zacate chips. The $70 is based off of more then likely the E-350.

It has a 5th of the Transistors at 1/4th the die size.

At that size and price its as good as selling a Llano CPU at $250.
 
Anyone has any idea when exactly Q3 would it be released? Debating whether to just go ahead with a 2600k MIVE build, though the bulldozer Crosshair IV build sounds more intriguing...
 
Anyone has any idea when exactly Q3 would it be released? Debating whether to just go ahead with a 2600k MIVE build, though the bulldozer Crosshair IV build sounds more intriguing...

Could be August or September. You could just wait a bit, since it appears press is getting their hands on the chips in late July (if you consider the GameExperience to be a press event). As, such, they might release in August.

You need it now? Prices will likely fluctuate when the chips appear (better deals).
 
High end is 5% of the volume and 90% of the profit. Look at AMD how much money is making without a presence in the high end. You can alternavely compare their ASPs. High end does matter.

I really don't understand how you came up with this number. You do know that to both AMD and Intel, server parts are not included in the 'high end'. High end is a subset of a market, i.e. High-end Mobile, High-end Desktop/Workstation, High-end Server/HPC.

Mid-range and Value segments of those markets make up ~95% of volume, but have lower ASPs, actually resulting in higher revenues, while the High-end segment has higher ASPs, but around ~5% of volume, which results in lower revenues than the lower segments.

How many >$500 processors do you really think Intel and AMD (during the Athlon FX era) sell in one quarter, versus <$300 processors?

EDIT: To put it another way, if Intel sold processors for no less than $350, who would make more money in the current PC market, AMD or Intel?
 
Last edited:
I really don't understand how you came up with this number. You do know that to both AMD and Intel, server parts are not included in the 'high end'. High end is a subset of a market, i.e. High-end Mobile, High-end Desktop/Workstation, High-end Server/HPC.

High end from an ASP POV. I don't care if it's server or extreme desktop.

You have low end CPUs <$100; middle end $100-$200 and what's over $200 is IMO part of the high end.
Intel and AMD do not make money on the low end parts which are the bulk of the volume.
Mid-range and Value segments of those markets make up ~95% of volume, but have lower ASPs, actually resulting in higher revenues, while the High-end segment has higher ASPs, but around ~5% of volume, which results in lower revenues than the lower segments.

How many >$500 processors do you really think Intel and AMD (during the Athlon FX era) sell in one quarter, versus <$300 processors?

EDIT: To put it another way, if Intel sold processors for no less than $350, who would make more money in the current PC market, AMD or Intel?

I'm actually refering to profit, not revenue. It is fairly obvious where the lion share in revenue is. That doesn't mean however that the profit is also coming from there. For example Intel probably sells 200k Pentiums a day and only 10k Sandy Bridge 2600k. Which one brings more profit in your opinion ?

APU's != Llano, but Llano=APU.

The APU's they shipped in Q2 they are referring to the Zacate chips. The $70 is based off of more then likely the E-350.

It has a 5th of the Transistors at 1/4th the die size.

At that size and price its as good as selling a Llano CPU at $250.

No. They also shipped 7k Llanos in Q2 at an ASP of $70. The E350 ASP was around $31 if memory serves.

FYI, the high end desktop model, A3850 is rumored to be launched at an list price of $129.
 
Could be August or September. You could just wait a bit, since it appears press is getting their hands on the chips in late July (if you consider the GameExperience to be a press event). As, such, they might release in August.

You need it now? Prices will likely fluctuate when the chips appear (better deals).

They said 60-90 days in early June, so that would imply early to late August.
 
High end from an ASP POV. I don't care if it's server or extreme desktop.

You have low end CPUs <$100; middle end $100-$200 and what's over $200 is IMO part of the high end.
Intel and AMD do not make money on the low end parts which are the bulk of the volume.


I'm actually refering to profit, not revenue. It is fairly obvious where the lion share in revenue is. That doesn't mean however that the profit is also coming from there. For example Intel probably sells 200k Pentiums a day and only 10k Sandy Bridge 2600k. Which one brings more profit in your opinion ?



No. They also shipped 7k Llanos in Q2 at an ASP of $70. The E350 ASP was around $31 if memory serves.

FYI, the high end desktop model, A3850 is rumored to be launched at an list price of $129.

Even better then if they can sell a Llano at a profit at $70, surely a CPU 1/4 its size should be enormously profitable at $31. If you have any links to these reports on ASP that would be great though, didn't seem like to many non-A8's were being sold.

Also remember AMD no longer is responsible for establishing the manufacturing base for the new CPU's. This means they just pay a flat fee on wafer's. That gets rid of a lot of overhead. So really they only need to cover the production of the chip (with packaging) and R&D. So they can in theory with enough sales make large profits on $1 margins. The shift is at a below $300 CPU in the computer, $50-~$200, with people wanting a a performance BD $200- $300. Seems like they are targeting sales not ASP. Nobody tries to sell stuff at a loss, if they really didn't think they could make a profit off of Llano then they would all be 2C 160SP(or 220) instead of 4C CPU's. Specially since 220SP would still be loads better then HD2000 and CPU power isn't the focus.
 
For example Intel probably sells 200k Pentiums a day and only 10k Sandy Bridge 2600k. Which one brings more profit in your opinion ?

That depends purely on manufacturing costs. I'd need to be provided that information first before determining which product is more "profitable." As it stands, the revenue from those 200k pentiums is 16 mill to 3 mill (with an average price of $80 for the Pentiums and $300 for the 2600k). Nearly 5.5 times that amount.
 
That depends purely on manufacturing costs. I'd need to be provided that information first before determining which product is more "profitable." As it stands, the revenue from those 200k pentiums is 16 mill to 3 mill (with an average price of $80 for the Pentiums and $300 for the 2600k). Nearly 5.5 times that amount.
But remember, those 2600K's bring in more 'profit' than those Pentiums do.
 
I still think Zambezi must be really good, not beat SB by 50% 'good', but close enough 'good'. Why else would Nvidia allow SLi on the upcoming AMD 9-series chipsets? Can't be out of the goodness of their hearts, because SLi would have been 'enabled' on 7xx and 8xx chipsets after Nvidia exited the chipset business.

I don't think so. NVIDIA finally relented yes, but not because AMD's Zambezi parts are so badass but rather because they are out of options. NVIDIA is a stubborn company and they'll only relent once they've seen enough of an impact to their bottom line to swallow their collective pride. It also probably took them awhile to come to an agreement with AMD. Something I'm sure NVIDIA procrastinated on even approaching AMD with.

So far every indication is that Bulldozer / Zambezi are disappointing thus far. It remains to be seen if a silicon revision can fix that but my guess is that it won't totally be resolved. I'm fairly certain based on pricing information and performance information known at present that a silicon revision will only partly address the problem. We may not see Bulldozer get where it needs to be until it's been refreshed at least once. By then Intel may be even further ahead.
 
thread has been derailed : ( On Topic: What do you guys think is the cause of the delay? In my eyes it falls to being two possibilities; bulldozer having some bugs that they know how to fix in that short a time or Llano having such a high demand that they wanted to ensure that there were no chip shortages.
 
I don't think so. NVIDIA finally relented yes, but not because AMD's Zambezi parts are so badass but rather because they are out of options. NVIDIA is a stubborn company and they'll only relent once they've seen enough of an impact to their bottom line to swallow their collective pride. It also probably took them awhile to come to an agreement with AMD. Something I'm sure NVIDIA procrastinated on even approaching AMD with.

So far every indication is that Bulldozer / Zambezi are disappointing thus far. It remains to be seen if a silicon revision can fix that but my guess is that it won't totally be resolved. I'm fairly certain based on pricing information and performance information known at present that a silicon revision will only partly address the problem. We may not see Bulldozer get where it needs to be until it's been refreshed at least once. By then Intel may be even further ahead.

Imho nvidia situation is quite simple they had license for amd chipsets. Something in that license probably stops them from releasing AM3+ chipsets. So they allowed SLI on 990 mobos.
 
2600K are $115 a piece on Intel Retail Edge so I guess they are making crap load of profit.

But that is assuming that the the actual sales of the 2600k is really that great. Its easy to measure what a chip is profitable at when you look at the price of the lowest priced version of it. So it is nice to have a chip that sells for that much higher a price. But the point is speccially on a Fabless company that market penetration matters more then margin, because as long as you are making a profit then its all good.

AMD seems to have a really good lineup in terms of market penetration and pricing. They have an embedded netbook chip selling in the 10s of dollars that is faster and priced right around its competitor. They have a thin/general use CPU selling in the $60-$150 dollar range that is great for general computing where CPU speed doesn't matter as much as graphics and features, creating a whole subset of highly usable $500-$700 laptops. They have a performance CPU (not sure performance) Selling in the $150-$350, and are using that same CPU for servers where the sale price can go from $250-$2000, places where high priced CPU's actually sell.

Margin is great an all, but it doesn't take a rocket scientist to see that they going into this year has drawn up a really advanced strategy to target their components at market prices that are targeted at the prices where the market is at its best based on purchase rates.

In retrospect sure a 2600k is probably worth more in profits, but a lot of that has to do with the lack of purchases in the Pentium end. This is an avenue that something like the Llano is going to rectify. For intel and OEM's the current Pentium doesn't fill it's spot very well when you consider target audiences. Sure its cheaper then an i3 but how much does it bring to the table over a Atom or Zacate? Is the price/performance decrease worth it to not use an i3?

My dad got a really nice Asus Windows 7 tablet for $1k considering the price of the phone based tablets its pretty well priced. Instead of going with an i3 or Pentium they went with an ULV i5. I don't think the Pentium is well positioned in performance and features compared to where it is priced.
 
How much a "latest" tech product really worth depends on when the next "latest" tech comes out. When both the latest and the next latest product are determined by the same company, they have all the flexibility to control their profit. So what we are really buying from them is the time until they release their next product.
 
Imho nvidia situation is quite simple they had license for amd chipsets. Something in that license probably stops them from releasing AM3+ chipsets. So they allowed SLI on 990 mobos.
Since Nvidia was a HyperTransport licensee, they were able to make chipsets for anything that used HT for I/O. Nvidia didn't have to license anything from AMD to make AM3/AM3+ chipsets. Nvidia's problem was that:
1. They could not get a QPI/DMI license to make chipsets for Intel CPUs, making them cave on licensing SLi for Non-Nvidia chipsets.
2. They were being marginalized on the AMD side by AMDs chipsets
3. OEMs were only buying GeForce 6100/6150 chipsets, not newer GeForce 8-series (for proof of this, look at all of the Nvidia-based AM3 OEM systems)

I still feel that the whole SLi on AMD 9-series chips has more to do with something that Nvidia knows about BD.

I wonder, if Komodo will use the FM2 socket, will Nvidia allow SLi on it?
 
Imho nvidia situation is quite simple they had license for amd chipsets. Something in that license probably stops them from releasing AM3+ chipsets. So they allowed SLI on 990 mobos.

No, because Nvidia pulled out of the AMD chipset market as well. Their existing chipsets couldn't be reused for Bulldozer either, unlike the AM2+ Phenom 1 chips rebranded for AM3.
They stand to lose on SLI sales if they ignore the AMD business.
 
I can wait no longer, I was hoping for bulldozer to be out by this time. If it was on par with SB I would've gotten it. If not it would of have at least lowered SB prices a bit. Oh well. Sandy Bridge here I come...somewhat reluctantly!
 
... Their existing chipsets couldn't be reused for Bulldozer either, unlike the AM2+ Phenom 1 chips rebranded for AM3.

How so?
Technically, the chipset has nothing to do with whether or not a board is AM2, AM2+, AM3 or AM3+. As long as the CPU uses HyperTransport, any chipset will work, because ALL CPUs support HT multipliers from 4x to 20x (800MHz to 4GHz).

It is just like people that say if a board does not have a 900-series chipset, it isn't an AM3+ board...
 
How so?
Technically, the chipset has nothing to do with whether or not a board is AM2, AM2+, AM3 or AM3+. As long as the CPU uses HyperTransport, any chipset will work, because ALL CPUs support HT multipliers from 4x to 20x (800MHz to 4GHz).

It is just like people that say if a board does not have a 900-series chipset, it isn't an AM3+ board...
Argh! lack of sleep is bad, brain and typing hands don't say the same thing

What I meant was that there was no attempt by mobo makers to make AM3+ motherboards using Nvidia's 980a chipset, and even only Asus used the 980a chipset for about 1 board at most only. Also, Nvidia is pulling out from the chipset business on all fronts. It just makes sense to make a quick buck like what they did with SLI certification for Intel chipsets.
 
Doesn't really matter if they didn't want or couldn't produce chipsets or couldn't find anyone who would want to use NF980 for am3+.

There won't be nvidia based am3+ mobos - so it's natural decision to allow SLI on 990FX and get cash for licensing and selling more cards.
 
Doesn't really matter if they didn't want or couldn't produce chipsets or couldn't find anyone who would want to use NF980 for am3+.

There won't be nvidia based am3+ mobos - so it's natural decision to allow SLI on 990FX and get cash for licensing and selling more cards.

I'll laugh my butt off if AMD ever tells them they don't want their stinking SLI! :D

I personally like Nvidia cards better at the moment, but AMD is catching up big time and in 1-2 years I see them being the better graphics card makers eventually.

Who knows really, just my current opinion.
 
I personally like Nvidia cards better at the moment, but AMD is catching up big time and in 1-2 years I see them being the better graphics card makers eventually.

.

Catching up? They've been caught up. But that's a discussion for another topic.
 
I'm hoping that that article is incorrect; I mean it is BSoN, and they haven't exactly been known to be the greatest source.
 
Yeah, AMD's going to charge premium price (over $300 for the top unit) for CPU' that are barely faster than the ones they are EOL'ing. If so, economic suicide must be their new MO.

The bulldozer has potential for both an IPC increase and IPC decrease. As for charging more money, 8 core and much higher base + turbo frequencies can make up somewhat for a less than great improvement in IPC.
 
Back
Top