Why not use C32 for the Zambezi desktop?

[Sn0wm@n]

but from your understanding of everything you read so far do you see amd having a high end enthusiast platform with a 1000$ absurdly fast cpu on the horizon?
do you think amd would care about socket costs if a c32 bulldozer was potentially competitive with anything intel could make?
im just looking at this from a product positioning marketing view.

unless of course its the inverse and am3+ bulldozer turns out so fast that 8 core will compete well with even the top 8core sb 2011 that there is no need for c32. i hope so but im skeptical. believe me nothing would please me more than seeing amd and intel offering equally fast high end products at a competitive price =)

who knows about the pricing that amd will use if bulldozer do give intel a good run for its money ...


and i dont know what else to say besides the point i allready made earlier ... to wich is amd would be losing money since they barely have market shares overall... but if you think your plan could yield them some proffit instead .. make a good presentation on paper .. send it to amd and hope for the best ???

[Dimitriman]

who knows about the pricing that amd will use if bulldozer do give intel a good run for its money ...


and i dont know what else to say besides the point i allready made earlier ... to wich is amd would be losing money since they barely have market shares overall... but if you think your plan could yield them some proffit instead .. make a good presentation on paper .. send it to amd and hope for the best ???

hehe i would write an awesome paper if it got me a job or traineeship with them on their marketing team.

[Sn0wm@n]

nothing is stopping you to gather their financial information ... and find out about the market segment you talked about and see if it makes sense .. send it to them .. with your resume of course and maybe you could score a job and then use your influence and make good things happen???

[Dimitriman]

nothing is stopping you to gather their financial information ... and find out about the market segment you talked about and see if it makes sense .. send it to them .. with your resume of course and maybe you could score a job and then use your influence and make good things happen???

indeed youre right and this very topic would make a really interesring thesis

[JF-AMD]

Anyone that correlates the $1000 client CPU market with market share doesn't know much about the actual market purchases of those products. I'd be willing to bet that those processors equal less than .1% of the market.

Now, cue the "halo effect" crowd. "But if AMD had the top product in the $1000 spot they could sell so many more lower end products."

I call BS on that theory.

Let's assume that company 1 has the fastest processor in the world, it scores 6987 on the one benchmark that matters. And that system costs $5000US

And you have $1000 to spend on a system (total). You two choices are:
Company 1 system. $1000. Benchmark score of 2140 on the one application that really matters to you
Company 2 system. $950. Benchmark score of 3027 on the one applicaiton that matters

Since you have $1000 in your budget do you buy the system from the halo company that is 30% slower, or do you buy the faster system that is cheaper. Now, just for fun, run that exercise over and over. At what point does the company 1 system need to get to before you buy that one over the company 2 system.

If you are like 99.9999% of the world, the answer has something to do with the benchmark being above company 2. Probably by at least enough to compensate for the higher price.

[Dimitriman]

Anyone that correlates the $1000 client CPU market with market share doesn't know much about the actual market purchases of those products. I'd be willing to bet that those processors equal less than .1% of the market.

Now, cue the "halo effect" crowd. "But if AMD had the top product in the $1000 spot they could sell so many more lower end products."

I call BS on that theory.

Let's assume that company 1 has the fastest processor in the world, it scores 6987 on the one benchmark that matters. And that system costs $5000US

And you have $1000 to spend on a system (total). You two choices are:
Company 1 system. $1000. Benchmark score of 2140 on the one application that really matters to you
Company 2 system. $950. Benchmark score of 3027 on the one applicaiton that matters

Since you have $1000 in your budget do you buy the system from the halo company that is 30% slower, or do you buy the faster system that is cheaper. Now, just for fun, run that exercise over and over. At what point does the company 1 system need to get to before you buy that one over the company 2 system.

If you are like 99.9999% of the world, the answer has something to do with the benchmark being above company 2. Probably by at least enough to compensate for the higher price.

for the sake of your argument, it would be a joy already if bulldozer competes agressively on performance for the upper mainstream market. i for one am not willing to spend 1000$ on a cpu since i would be much better off buying a 350$ one and spending the rest on a top graphics card.

but i still think that youre underestimating the high end market. of course its low volume although if you take into account all intel socket 1336 its probably more than .1%, but its still a very large profit margin and it also benefits amd as whole having a top product. look at how much hype ati gets simply by marketing the fact that they have the top gpu int the world.
if the mainstream is 30% faster then maybe you dont need the top product. but lets face it the reality will not be that ideal.
i believe the high end market is larger than you publish and it incorporates a latger slice of profits. you cant isolate it as the 2p desktop alone and if you instead take all of intel socket 1336 as well as socket c32 that would represent the market were talking about. and if that high margin low volume market were so easily ignorable tell me then why does nvidia STILL make more profit than ati even during the 5000 series slaughter? you know why - the low volume high margin workstation cards.

[scorpidragon]

if the mainstream is 30% faster then maybe you dont need the top product. but lets face it the reality will not be that ideal.
i believe the high end market is larger than you publish and it incorporates a larger slice of profits. you cant isolate it as the 2p desktop alone and if you instead take all of intel socket 1336 as well as socket c32 that would represent the market were talking about. and if that high margin low volume market were so easily ignorable tell me then why does nvidia STILL make more profit than ati even during the 5000 series slaughter? you know why - the low volume high margin workstation cards.

OK then AMD. How about letting us know how many BD modules the AM3+ socket can support?
If it cannot go over 4 BD modules, then you will loose future market share.
I'm more interested in a 1355 socket and a 1207 socket but then again it depends on how many BD module the C32 supports?

[Chrono Detector]

I know some people have already stated earlier that going tri or quad channel does not offer much performance, but I'm very disappointed that AMD is still sticking to dual channel, come on, this isn't 2005 anymore. Eventually programs will require more RAM, and I wouldn't have mine paying a premium price for AMD hardware anyway.

[stuffme]

I know some people have already stated earlier that going tri or quad channel does not offer much performance, but I'm very disappointed that AMD is still sticking to dual channel, come on, this isn't 2005 anymore. Eventually programs will require more RAM, and I wouldn't have mine paying a premium price for AMD hardware anyway.

To me this looks like amd is artificially keeping some performance headroom for the future. They know that BD is significally slower than SB, so there will be bigger performance boost(what we usually have seen) for the future models. And all the hails and praisers goes to amd once again for being again performance/watt/$ king.

[Oliverda]

OK then AMD. How about letting us know how many BD modules the AM3+ socket can support?
If it cannot go over 4 BD modules, then you will loose future market share.:

Are you a market analyst?

[FlanK3r]

4 Bulldozer modules, thinking, will better in overal performance than SB to LGA1155 clock for clock. My theory is massive multithread performance and about 85-90% single performance Sandy Bridge.Buts, still is a theory :-D.
About price, we seen higher (but very good!) price at Thubans, because this CPUs are good. No problem for customers, if someone need high performance can buy with simillary price x6 1090T or i7 930/950. But multithread performance is better for Thuban, lower real consumption (-30-50W in load) and whole platform is cheaper.
If Bulldozer will the best, can real price a bit higher than Thubans (example 500-550 dollars) and still it will ok for us!

[JF-AMD]

but i still think that youre underestimating the high end market. of course its low volume although if you take into account all intel socket 1336 its probably more than .1%

Please share your data. I know it makes me sound like a jerk to say "I am in this business and I think I might know more about market shares since I get the analyst reports each quarter", so I won't say it.

Wanting something to be true does not make it so.

When you consider "all of 1336" you are not doing the fair comparison. I am strictly talking about that $1000 SKU that everyone fixates on but nobody buys.

[RaV[666]]

When you consider "all of 1336" you are not doing the fair comparison. I am strictly talking about that $1000 SKU that everyone fixates on but nobody buys.

Well, for one, the talk was about high end platform.Not ONLY the 1000$ SKU.And people do buy them, just look on people SIGs onf various forums.But thats beside the point, that large quantitty of people buy the 250-500$ CPU`s.
And if youre saying thats ridicolous.
your company sells 5970 graphic cards, some partners are making VERY extreme editions of those.Thats 700-1000$ cards.
So your company is willing to make (much more complicated and harder to produce) extreme editions GFX, at low quantitys.But somehow selling just cpus (which piece for piece are MUCH easier and less costly to produce) isnt sensible ?
AND you did that before when you where on top ?
Thats just insanely not logical.Thats contradiction.
And about doing SMP consumer skus, all phenom x6 and magnycours chips are made form the same die.Same mask, same costs, there is little difference by binning, but not that much.It would cost nothing AMD to have a phenom FX in c32 package.

[pokipoki]

For server/workstation customers, unless Opterons rock their world, it is NOT in the best interests of AMD to use those chips for anything pertaining to overclocking. Allowing this to happen only makes the battle with the competition more difficult.

Compromising server chip revenue, a bigger slice than desktop, does not make good economic/business sense. Unless AMD fans want to see the green company go down in flames, like Cyrix.

[Dimitriman]

Please share your data. I know it makes me sound like a jerk to say "I am in this business and I think I might know more about market shares since I get the analyst reports each quarter", so I won't say it.

Wanting something to be true does not make it so.

When you consider "all of 1336" you are not doing the fair comparison. I am strictly talking about that $1000 SKU that everyone fixates on but nobody buys.

I understand I do not possess the market data regarding how large this segment is and even less so how large AMD defines it to be.

However it is a fair comparison to look at the 1336 intel segment since that segment is the platform with which we are comparing a possible c32 Bulldozer platform with. It incorporates not only all of Intel's high end processors, but also 1p/2p Xeon server processors as well. Hence its a perfectly fair comparison. I suppose we could look at how large AMD's total C32 market is, together with Intel's 1336 and define that as the Workstation/Enthusiast market. Now the question we're asking here is, how much is this market worth and what opportunities and threats are there for AMD.

AMD has been in this market before and the opportunities are clear:

-Re-establish AMD's brand as technological leader in the industry
-Enhance brand perception as "the choice for gamers" once held by the Athlon FX line
-Profit from low volume high margin sales
-Being present in every segment of the competition, enhance product portfolio and economics of scope (somewhat as this platform can be defined as a distinct product)

The threats we all know and we know what Intel will be bringing to the table regarding the high end market: 6-8 cores with 12-16 threads and quad channel memory.

Now I know I'm being very obnoxious here with this analysis that you must perform in your head every hour, but as a consumer with experience in marketing, I just don't understand how AMD can justify not being present in the high end segment considering all the direct and indirect benefits derived from being a performance leader. Think of every average consumer that walks into a store and knows nothing of pcs but asks "which is faster AMD or Intel?" storekeeper will refer to the high end segment dominated by Intel and answer "Intel has the fastest processors", consumer buys an Intel.

My final point is, when you compete in the high end segment, it's not just about that 0.1% market. It's about product portfolio and enhancing your brand image and the spill effects of being a industry leader, not a follower.

Sorry John for being so questioning of your product, but I do want AMD to continually improve as well, I'm not being a Terrace here :P

Thank you for your attention. It's much appreciated.

[Manicdan]

OK then AMD. How about letting us know how many BD modules the AM3+ socket can support?
If it cannot go over 4 BD modules, then you will loose future market share.
I'm more interested in a 1355 socket and a 1207 socket but then again it depends on how many BD module the C32 supports?

there is no simple limit like that since core count has almost nothing to do with the socket. we already know 8 modules are coming to the servers, and those chips are wider, so space wise i think we can get 6 modules on desktop using the AM3 size. but if they add more pins and make it bigger, then its based on how big they go. but then your looking at 2 chips being connected cause yields of monolithic 8 module chips would be lower than they want, and it wouldnt sell enough to pay for them to make it that big.

[JF-AMD]

And about doing SMP consumer skus, all phenom x6 and magnycours chips are made form the same die.Same mask, same costs, there is little difference by binning, but not that much.It would cost nothing AMD to have a phenom FX in c32 package.

Actually, it would be probably $5M just to get started. Different testers, different engineering processes, etc.

That is like saying bread and cake both use flour, so they are the same. The die is an ingredient, but it is hardly the only cost driver.

We have done the math on dual socket client systems and there is no way to make money with that product. The market is too small, and getting smaller every quarter.

Everyone here has emotions, but I have facts. Facts on the size of the market, facts on the costs to bring a product to market. It just doesn't make sense financially.

[JF-AMD]

I understand I do not possess the market data regarding how large this segment is and even less so how large AMD defines it to be.

However it is a fair comparison to look at the 1336 intel segment since that segment is the platform with which we are comparing a possible c32 Bulldozer platform with. It incorporates not only all of Intel's high end processors, but also 1p/2p Xeon server processors as well. Hence its a perfectly fair comparison. I suppose we could look at how large AMD's total C32 market is, together with Intel's 1336 and define that as the Workstation/Enthusiast market. Now the question we're asking here is, how much is this market worth and what opportunities and threats are there for AMD.

AMD has been in this market before and the opportunities are clear:

-Re-establish AMD's brand as technological leader in the industry
-Enhance brand perception as "the choice for gamers" once held by the Athlon FX line
-Profit from low volume high margin sales
-Being present in every segment of the competition, enhance product portfolio and economics of scope (somewhat as this platform can be defined as a distinct product)

The threats we all know and we know what Intel will be bringing to the table regarding the high end market: 6-8 cores with 12-16 threads and quad channel memory.

Now I know I'm being very obnoxious here with this analysis that you must perform in your head every hour, but as a consumer with experience in marketing, I just don't understand how AMD can justify not being present in the high end segment considering all the direct and indirect benefits derived from being a performance leader. Think of every average consumer that walks into a store and knows nothing of pcs but asks "which is faster AMD or Intel?" storekeeper will refer to the high end segment dominated by Intel and answer "Intel has the fastest processors", consumer buys an Intel.

My final point is, when you compete in the high end segment, it's not just about that 0.1% market. It's about product portfolio and enhancing your brand image and the spill effects of being a industry leader, not a follower.

Sorry John for being so questioning of your product, but I do want AMD to continually improve as well, I'm not being a Terrace here :P

Thank you for your attention. It's much appreciated.

Here is the problem: C32 is packaging, just like AM3+. Putting client bulldozer in C32 would not increase performance. Or increase capabilites. All it would do is increase prices. There is zero benefit to moving client to C32.

[crazydiamond]

awaits the next reply as to why you are still wrong

[Dimitriman]

you are right about features and i correct myself when i mentioned that c32 could bring quad channel ddr3 to bulldozer on the desktop as i understand that cannot happen with c32 as it would require more pins (hence g34)

What it could do however, is bring AMD back into a segment it has not been in since the days of athlon fx where the very best 1p server processors had a desktop equivalent (core count included) with higher clocks and shared socket.

But i suppose if AMD does not want to be merging desktop with workstation ever again, I hope at the very least they will consider the question: What is the best we can do within 140 watts? Even if the answer only interests .1% of the market.

Anyway I rest my case as I have nothing further to add. thanks again John for the discussion and I cross my fingers for am3+

[zanzabar]

c32 is not quad channel the g34 is, and its dual dual channel so u need 2 memory controller. the c32 is just socketF on ddr3 so its the am but with links for 2x HT for the cpu.

i would love to see a 2p c32 or g34 that could clock but amd says that its not worth the money and they dont have r&d money to throw around for pr and records like intel dose.

[scorpidragon]

c32 is not quad channel the g34 is, and its dual dual channel so u need 2 memory controller. the c32 is just socketF on ddr3 so its the am but with links for 2x HT for the cpu.

i would love to see a 2p c32 or g34 that could clock but amd says that its not worth the money and they dont have r&d money to throw around for pr and records like intel dose.

2x HT links could feed another north bridge chip; like having 2x 990FX NB with 1x or 2x 950 SB.

My personally I'm not going to bother upgrading to AM3+! Looks like Intel may get my business in the future? That all depends on AMD though

[zanzabar]

2x HT links could feed another north bridge chip; like having 2x 990FX NB with 1x or 2x 950 SB.

My personally I'm not going to bother upgrading to AM3+! Looks like Intel may get my business in the future? That all depends on AMD though

i guess or they could move the SB to the HT buss as its lacking throughput now and i dont see what u would gain to justify it, but i would buy one if they keept a similar price range to the current boards and high end amd chips.

the only platform socket change that could make any sense is the g34 to get the 2 die chips on an overclocking platform, as if i got a mangy 12 core up to 3.6-4ghz it would be sweet or there is supposed to be a BD one or something with a 16 core one next year and that would also be great if it clocked higher even though most of the time only 2-3 cores would be used.

[scorpidragon]

i guess or they could move the SB to the HT buss as its lacking throughput now and i dont see what u would gain to justify it, but i would buy one if they keept a similar price range to the current boards and high end amd chips.

the only platform socket change that could make any sense is the g34 to get the 2 die chips on an overclocking platform, as if i got a mangy 12 core up to 3.6-4ghz it would be sweet or there is supposed to be a BD one or something with a 16 core one next year and that would also be great if it clocked higher even though most of the time only 2-3 cores would be used.

You're not the only one, if Phenom FX came out on the G34 platform well I'd buy that and a board (providing it could OC)!

[freeloader]

All I want to know before I purchase my next high end system (2nd half of next year) is whether AMDs Socket AM3+ will handle more than 4 modules (8 cores)? If not, I'll see whether Intels next socket can. Whoever supports more than 8 cores is getting my money.

I would settle on 8 cores as long as they're at a decent speed of 3.5ghz or more.