AMD presents "The Bulldozer Blog"

[informal]

IMO,the conservative estimate AMD publicly gave comes from the current 32nm node condition @ GloFo(lower clock speeds than projected on the first batch of ES?).I think GloFo will certainly clear the issues by the time mass production starts next year,but for now AMD must use what they have.Remember ,this is all around new chip design with 8 monolithic cores (x 2 in server parts,via direct connect MCM) and they(AMD) must work within previously set power bands. As the node issues are cleared up AMD can better project the end clock speeds and thus the performance.

[Piotrsama]

They presented it?
I thought it was already done by Dresdenboy: http://citavia.blog.de/
j/k

I really hope this bulldozes the competition, will be fun to watch if it does.

[Solus Corvus]

I don't think there is anybody here that wouldn't like to have concrete bulldozer performance numbers right now. But I doubt that whining about it will change AMD's policy on the matter. They may not even have hard numbers themselves because there is still a lot of work on the silicon left (how would you come up with a hard number without the final clock speed?).

With all of the "doom and gloom" or "epic monster" scenarios being passed around it would be hilarious to watch all the backpedaling if it turns out to simply be a decent competitive product.

[hlopek]

JF can not comment on performance before the actual launch of the product. I guess thats clear to everyone besides the top screamers here (or is it just competition fishing for data?).

Real guy aren't afraid of competition only schemers are But then gain every marketing guy must have some scheme to sell out to indians (not pc enough?) wanting these shiny new pearls he's offering

Because only idiots get hyped up over pre-release marketing.

OMG. What kind of bluntness While in fact i agree with your principles, i think you used wrong approach to PR scheming. First you find out what that PR guys are really bubbling about, and does it really works or it's just JHH-type smokescreen (tribute to Jen-Hsun Huang) when delivering carboardware to the market. And only after that throw rocks and cedar wood onto him (well rotten cabbage will do the job also)

My point is simply that the "number", as you say, is virtually meaningless without knowing to what it refers, so there is no point to it in isolation.

Exactly. And that's why better to take these PR bubble talks with a jar of salt, rather then take it for a fact and try to explain to yourself that 50% (of nothing) performance gain (it still nothing for math experts). When these performance numbers would be presented against real product with another real product then we can talk about some performance gains.

You're losing customers over you're refusal to even release some meaningless numbers. Intel's next architecture is set in stone now, they can't change it. About all they could do at this time is release higher clock speed parts on their 22nm process. AMD is in the same situation, although they may be able to make minor tweaks to their silicon between now and the official launch. What could AMD possibly lose right now by releasing some Cinebench or some other benchmark numbers? I'm one of the few on this board who doesn't care about architecture or the design process. I just care about performance.

Excellent punch, some performance numbers (if this BD architecture is really near finish) might help them alot because AM3 offers seamless upgrade (except HT3.1?) and many high-end enthusiastic oriented people might jump on intel bandwagon if Bulldozer is delayed as K10 was.

If they really mean to start volume shipping in 2011/H1, and not to delay it to Q2 (or Q3) in fact, they should have pretty much working batch of "final prototype" and if that have some bugs they should be solved during August (this month) and we could see real performance of highly and extremely long anticipated Bulldozer architecture based on "prototype chip". They need some 4 month or more for every new tweak so if early 2011/H1 is exact date we should see some performance benches now and in early November at latest final product engineering samples of final product that will go onto market.

whos on that cycle?

i got a AM2+ board and its had duel cores, quads, and now hexies. if each one lasts just a year, im on a 3 year cycle for 1 motherboard. that reduces my cost by about 75$ a year compared to your 1.5 year cycle, which means i can spend 75$ more on a cpu, instead of replacing it with a motherboard that offers nothing new.

If you put all in economic terms. Some of us are never seen actual mobo that costs 75$ and that's good enough to be a solution for longer upgrades (over 3yrs). Where i buy, only obsolete nf560/nf720(8200igp)/amd720/amd740 cost that much (and it's well known how much manufacturers reduce costs on that entry level boards) While some boards that offer real upgrade path start at 170USD (equivalent) for some 3rd grade manufacturers or more for better ones like Gigabyte/Asus/MSI. So i'd say you wrong talking about reduced cost. (For server upgrades there's sowhat different story but still they start at 300USD)

Yep you can upgrade and put 95W Phenom X6 when released on some nf560 board but this board is designed to sustain 70W constant load and some (near 95W) CPU will kill it in few month (if it's not brand new, an if endure that long). Some higher quality nf590/570 boards usually didnt have that kind of upgrade path (only nice exception there is Gigabyte and Asrock). NTM those erratas AMD built to kill nV competition on HT2.0 based boards if you upgrade onto "Stars" architecture

So for me that "you dont have to change mobo" mantra doesn't really work in real life, just as it didnt in time of K7s KT266A chipset. And it's usually every 3yrs upgrade with new mobo cause old couldnt be upgraded and it wasnt even sub 100USD part

[slaveondope]

^^^^This guy working for Intel
Nothing but negative contradictions to anything anyone says positive about AMD.

[Manicdan]

If you put all in economic terms. Some of us are never seen actual mobo that costs 75$ and that's good enough to be a solution for longer upgrades (over 3yrs).

its 75$ per year, so over 3 years it totals 225$ which is a decent price for a very good long lasting mobo.

if the cycle is 1.5 years per mobo 225+225 = 450$
if its a 3 year cycle, just 225$

(450-225)/3 = 75$ per year saved by having a longer platform. and i checked my purchase date, turns out my motherboard is 3.5 years old already, and i plan on keeping it until AM3+

[Dimitriman]

It's not too hard to understand why no numbers are up yet.

If you take marketing and customer reactions into account:

If BD is released on Servers first, then AMD should keep performance secret as server products often involve million dollar contracts for high volumes. Magny Cours is a competitive product and sells very well, and releasing BD performance numbers could affect millions in MC sales. Intel often releases desktop chips before migrating them to server, hence MC sales should continue to do well until Bulldozer.

If BD is released on the desktop first, then AMD should release some benchmark numbers, as they are only competing on price/performance segment in this market, expectations of near future high end products that could beat the competition could hold customers from buying them. Sandy bridge is just around the corner and AMD customers on the desktop would prefer to see whats coming now rather than later.

My understanding is that BD is coming on Server markets first, hence all the quietness from AMD

[STaRGaZeR]

It's not too hard to understand why no numbers are up yet.

AMD's marketing department is teh , but even they realize is stupid to advertise a product that is at least 1 year away.

[freeloader]

AMD's marketing department is teh , but even they realize is stupid to advertise a product that is at least 1 year away.

If Bulldozer is truly 1 year away and that's a very big IF, then my next purchase would definitely be SB. I was hoping for a January to March release in 2011.

[STaRGaZeR]

If Bulldozer is truly 1 year away and that's a very big IF, then my next purchase would definitely be SB. I was hoping for a January to March release in 2011.

You can believe what you want to believe, the rumours saying GF is having issues with 32nm or the usual marketing BS saying "still on track" for 2011. This sounds a lot like the typical bullcrap you get when a product is still far away (or it's crap) and you have to maintain interest. The fact is that SB is Q1 2011, BD is in no way Q1 2011. Plus no word on perfomance.

[FlanK3r]

Star:SB only to 1155, this LGA is outperform LGA1366 (i7 960/965/975/970/980).
BD is in no way Q1 2011 How do u can know? It can be Q1, it can be Q2....

[Calmatory]

You can believe what you want to believe, the rumours saying GF is having issues with 32nm or the usual marketing BS saying "still on track" for 2011. This sounds a lot like the typical bullcrap you get when a product is still far away (or it's crap) and you have to maintain interest. The fact is that SB is Q1 2011, BD is in no way Q1 2011. Plus no word on perfomance.

+50 % with 33 % more cores from existing product to new product. So the increased throughput is achieved by more cores.

Not in games though.

[STaRGaZeR]

+50 % with 33 % more cores from existing product to new product. So the increased throughput is achieved by more cores.

Not in games though.

That claim comes from the marketing team. Never, ever believe marketing teams, or even presentations. You'll see how the real thing performs when it launches. And yes, single thread perfomance is not looking good at all.

1 year of Bulldozer nonsense until it launches,

[JF-AMD]

That claim comes from the marketing team. Never, ever believe marketing teams, or even presentations. You'll see how the real thing performs when it launches. And yes, single thread perfomance is not looking good at all.

1 year of Bulldozer nonsense until it launches,

Yeah, marketing always overhypes things. I said Magny Cours was going to be 50-60% higher than istanbul. I was really wrong.

80-120% was more like it.

What people forget about any predicitions is that it does no good to overhype your products. If I said it was going to be 300% faster and it wasn't a.) I blow all credibility and b.) the product is viewed as a failure.

Fanbois will overhype (or downplay) any technology without any evidence of actual performance, but from a marketing side there is more on the line.

Overhyping a product could lead to charges that you are trying to pump up your stock price and nobody wants that. I have grown accustomed to sleeping in my own house and the luxury of eating normal food. I wouldn't look good in stripes.

I would argue that marketing people are probably the most reliable sources of information on these boards because they are the only ones actually bounded by the truth. And legal teams that get nervous about what they are saying.

If you take a look at everything that I posted about Magny Cours prior to the launch, you'll see that I was not misleading. Except about the total performance, but we had some upside with the final silicon and we kept that to ourselves for obvious reasons.

I can't speak to the other marketing people here, but if you identify yourself and clearly align yourself to the forum at to who you are and who you work for, there is a high probability that you are going to be honest, in most cases, conservative with your claims because you are seen as speaking for the company, even though you are posting on your own as an individual.

My 2 cents.

Also, don't forget that we have a boost technology. The 50% number is based on a fully utilized processor. If you were only running a single thread you would be looking at a very different uplift.

[MrMojoZ]

That claim comes from the marketing team. Never, ever believe marketing teams, or even presentations. You'll see how the real thing performs when it launches. And yes, single thread perfomance is not looking good at all.

1 year of Bulldozer nonsense until it launches,

You say the information is wrong/un-reliable then decide that single thread information is going to be bad based on that same information.

You can be a fan without being a fanboy, try it.

[Manicdan]

I would argue that marketing people are probably the most reliable sources of information on these boards because they are the only ones actually bounded by the truth. And legal teams that get nervous about what they are saying.

i wasnt happy with the marketing behind ipad, calling it a revolution, but that is a company based on sales driven by marketing. the fact amd does not market much at all, leads to having to do it the old fashioned way, with benchmarks and facts.

[Nintendork]

1155 do not worth for any with 1156/1366 platform. The octocores from SB come in late 2011(or Q3 AFAIK, the 1155 are useless for this chips). So both SB and Bulldozer will probably be on market at the same time.

And really, wanting six/octocores and then talking about single thread performance? That's a thing of the past.

[Manicdan]

And really, wanting six/octocores and then talking about single thread performance? That's a thing of the past.

People who like SuperPi, probably also like to have an orgy with 4 women, but only let one touch, and tells the other 3 to sit in the corner, watch, and dont even act like your enjoying it.

[Calmatory]

SB octocore with 16 threads vs. BD octomodule with 16 threads...

[cegras]

As Calmatory said, all I'm looking for is a socket level performance. If it delivers there's no point in trying to downplay its accomplishments. You sell CPUs, not cores or architecture.

[JF-AMD]

As Calmatory said, all I'm looking for is a socket level performance. If it delivers there's no point in trying to downplay its accomplishments. You sell CPUs, not cores or architecture.

Bingo, we have a winner. People don't buy architecture and cores aren't available a la carte.

You buy a processor and that is the most discrete level that you really want to get down to. Arguing below that is less productive.

What is the performance, price and power consumption of the processor, that is what matters.

[Shadov]

Bingo, we have a winner. People don't buy architecture and cores aren't available a la carte.

You buy a processor and that is the most discrete level that you really want to get down to. Arguing below that is less productive.

What is the performance, price and power consumption of the processor, that is what matters.

JF thank you for interacting with us and although you are sometimes dealing with "pure fanboism" you seem to handle it very well!

I personally enjoy every bit of information that you share with us. That way keep up the good work and don't get discouraged...

[STaRGaZeR]

Bingo, we have a winner. People don't buy architecture and cores aren't available a la carte.

You buy a processor and that is the most discrete level that you really want to get down to. Arguing below that is less productive.

What is the performance, price and power consumption of the processor, that is what matters.

Exactly. John, you know your work better than me, so please. Your goal is to sell CPUs, fooling the customers if necessary. You load your CPU with all kinds of workloads. Your perfomance figures are only valid when you load the CPU to the max. Nobody is saying that the CPU will be bad is this situation, but things don't work like that most of the time and you know it.

I'll tell you what's productive. I use app X, what processor gives me the best perfomance, perfomance/price, etc? If my application doesn't load the CPU to 100% I don't care about what it does when it is at 100%. No 100% in this case means no app with 12/16/whatever number of threads the CPU has. If my app is restricted to 2 threads, or any other number below the max, your CPU will underperform compared to the competition. You know that there are more CPUs running below and way below 100% load than fully utilized. Your competitor is not the leader in everything except maybe price without reason, some people will never get it. You can attack them with perfomance where it matters or with words for the fool, and I don't see a lot of the former. Athlon64 was such a success because the core was a lot better than Netburst, not because its "socket level" (what a term...) perfomance was better. Unless BD's integer "clusters" are significantly faster than the ones present in the current generation of AMD CPUs you'll still lag behind Intel. We'll talk again when BD is released. If I remember about it, which is going to be difficult.

[Calmatory]

Stargazer. Don't worry. If one wants to be a pessimist, 8 cores + 12 % IPC increase should be enough for even you.

For the third time please, oh someone PLEASE provide SOMETHING which does not scale to multiple threads well, and which is other than SuperPI or games. Or software older than 3-5 years. I can't think of anything serious apart from games. And even in that case it is developers decision not to use multiple threads.

[ajaidev]

Before a real processor is made the company has to sell the architecture to the investors and say what you will the architecture of a processor says a lot. Its no surprise that stock buy/sell organization often employ several chip architects in an attempt to find out how good the processor will turn out and how well will it deal with the other side which will offcourse affect the price of the companies stock. Informed speculation on insider trade are tools of the trade be it legal or not.

When K10 was 2 months from release an stock investment firm bought quite a few shares in AMD because according to their resident chip architect the architecture was very well made and it was suppose to be a run away success. But you all know how that turned out. The point being Intel will keep building on and enhancing the nehalem architecture for years to come till AMD forces them to think differently or they exhaust the architecture out.