An Expression of Frustration over the Bulldozer Debacle

[Edrick]

Mr. Rory Read
CEO
AMD
One AMD Place
Sunnyvale, CA

Dear Mr. Read:

WITHOUT PREJUDICE

We the members of AnandTech forums have given your company more than enough latitude in the release of your Bulldozer line. Although we are fully cognizant of the fact that in order to trounce your competition in the most popular benchmarks you may need to reach stepping B74, we demand that you release Zambezi immediately and in any case no later than September 19. Should you fail to do so, we will hold our breaths, turn blue, and blame it all on you and your boyfriend Hector Ruiz. Do not take this threat lightly as we have accomplished Photoshop artists on this forum ready to ensure that TechCrunch will receive ample photographic evidence of you two in flagrante delicto.

Govern yourself accordingly.

The AnandTech Gang

Thats classic!!!! :thumbsup:

 

[LOL_Wut_Axel]

Just wait until the 19th of this month...

In any case, like I said before: expect higher multi-threaded performance than Sandy Bridge and lower single-threaded performance. How much or how little that performance difference is remains to be seen.

 

[sm625]

I just cant understand why they would make an 8 thread chip that has 16MB of cache, yet only be able to issue (intel-4) or (intel-6) or (intel-whatever) instructions per clock. I'm sure IPC will end up being higher than previous AMD, but I just do not get why they didnt make a wider core... With power gating it surely would be more efficient to go intel's route. It just bugs me that a hyperthreaded core could even remotely possibly be faster at processing two threads vs a bd module.

 

[AdamK47]

Movie studios that know their next release will be panned by critics usually opt out of advance screenings.

 

[zebrax2]

I just cant understand why they would make an 8 thread chip that has 16MB of cache, yet only be able to issue (intel-4) or (intel-6) or (intel-whatever) instructions per clock. I'm sure IPC will end up being higher than previous AMD, but I just do not get why they didnt make a wider core... With power gating it surely would be more efficient to go intel's route. It just bugs me that a hyperthreaded core could even remotely possibly be faster at processing two threads vs a bd module.

IIRC core complexity increases dramatically as you try of increase IPC. AMD just doesn't have the R&D budget to outrun Intel in the IPC race that is why they are looking for other ways to be competitive

 

[LOL_Wut_Axel]

IIRC core complexity increases dramatically as you try of increase IPC. AMD just doesn't have the R&D budget to outrun Intel in the IPC race that is why they are looking for other ways to be competitive

Even if it's just 10% higher IPC than Llano it'll be faster than Sandy Bridge in multi-threaded. If AMD hasn't mustered even that meager increase, Bulldozer is doomed to fail. Llano IPC just isn't gonna cut it.

However, I'm doubtful about a completely new architecture having the same IPC as one that has its roots in 2003 (K8). AMD should at least get a slight increase.

 

[Axon]

That may not be it. The length of the delay signals to me that they discovered a bug with one of the prototypes. (what has it been, 3 months? That is about the length of time it takes to refab a new prototype).

Fabbing prototypes is an expensive and time consuming process.

True, true. Who knows what the real reason is.

 

[AtenRa]

It just bugs me that a hyperthreaded core could even remotely possibly be faster at processing two threads vs a bd module.

There is no way in Hell that will happen, even if BD module will have the same performance as two Llano cores, witch will not happen, even then BD module will be faster than Single + HT SB core in two threads.

From The Sandy Bridge Pentium Review: G850, G840, G620 & G620T Tested review.

Llano A8-3850 (4 cores) at 2.9GHz is faster than Core i3 2100 (2 Cores + HT) at 3.1GHz.

more,

From AMD Phenom II X4 980 Black Edition Review

Athlon/Phenom II (4 cores) are faster than Core i3 2100 (2 cores + HT).
Phenom II X6 1100T is very close to Core i5 2500K at the same frequency and Core i7 2600K only gets 10-15% more than 2500K.
Core i7 X980 (6 Cores + HT) gets almost 50% more than Core i7 975 (4 Cores + HT).

There is no substitute for more threads than physical cores and as i have said before, i expect 8 core BD to come close to Core i7 X980 in those multithreaded benches

 

[jhansman]

*Reality Check*
Two wars going on, economy in the crapper, unemployment at double digits, higher gas prices than ever before, and you get your panties bunched over a CPU release date? Time to move out of Mom & Dad's basement.

 

[Idontcare]

And what's worse, AMD? You're shooting YOURSELVES in the foot! Don't give me this nonsense about "releasing performance data" would cannibalize your sales! I think that those who aren't willing to wait will go with the higher performing, reasonably priced Intel offerings (if they're looking at a "high end" price range), and those who are willing to wait would at least like something to keep them from going for the already-available Intel offerings that provide a very high level of performance.

It can't hurt to give us some performance and launch data beforehand unless Bulldozer will significantly under-perform your current offerings and/or offer a worse price/performance ratio than current Intel offerings. And if this is the impression you're willing to leave on us, then you're just sending us straight to 2500K/2600K/990x chips from Intel. So why don't you do the enthusiast community-- and indeed, yourselves -- a favor and give us SOMETHING.

Ever hear of The Osborne Effect?

Adam Osborne, pre-announcement of next Osborne computer

The year was 1982. British computer pioneer Dr. Adam Osborne, a man who has been universally credited with creating the portable computer industry announces the “Executive” OCC-2, the the successor to his current shipping product, the CP/M-based Osborne 1. In fact, over the next year, he also publicly discusses a second, smaller model, the “Vixen”, one which would follow on after that.

Not many people will remember Adam Osborne and the significant contributions he made to help establish the personal computer industry. Many people reading this article weren’t even born when the Osborne 1, let alone the Vixen was shipped.

However, there is one particular event in computer history in which Mr. Osborne’s name will forever be associated with:The Osborne Effect.

What happened to the Osborne Computer Company after the announcements of the “Executive” and the “Vixen” is now classic business school material. Due to the pre-announcement of the newer, better products while the current inventory in the reseller channel was still full, buyers were no longer interested in current products.

Despite the fact that the company had a number of advantages, one of those being that it bundled application and OS software with its computers, Osborne was also facing heavy competition from companies like Kaypro, Apple and IBM, so the timing couldn’t possibly have been worse.

By November of 1983, the company went bankrupt, and Osborne Computer Corporation was no more.

Its fairly obvious that AMD has.

 

[lifeblood]

Some of us are missing the point. Intel can piss off the world and not worry about it effecting sales because they are the 800 lbs gorilla. AMD needs every friend and sale it can get. Obviously if they annoy a couple of enthusiast they won't notice, but if they annoy the whole community they will take notice and maybe perform some kind of damage control. They are well aware that some of us are just kids living in our parents basements, but others of us work in IT for medium to large companies. Some of us make meaningful purchase decisions, both server and workstation, and it is unwise to really piss us off.

They, like most companies, pay attention to their user base and try to keep them satisfied. They know that the delays are pissing off their customer base because they have eyes and ears and can see it happening. JF-AMD is one of those sets of eyes and ears. So do they care about this post in particular? Hell no! Do they care about a bunch of posts like this in numerous forums? I think so.

Obviously if the product isn't ready or their not making enough they can't improve it any more than they are already trying to. But at some point, if enough mud is getting thrown at them, they will be forced to say something.

 

[d4a2n0k]

Not meant to be flamebait....

Is anybody still buying Phenom chips? Cant cannibalize sales that arent there can you?

 

[Spikesoldier]

Not meant to be flamebait....

Is anybody still buying Phenom chips? Cant cannibalize sales that arent there can you?

they are selling much better in the oem channel than most believe.

 

[Accord99]

Ever hear of The Osborne Effect?

Its fairly obvious that AMD has.

But strictly speaking, the Osborne Effect comes into play simply by pre-announcing the new product. AMD has already Osborned their current lineup years ago by announcing Bulldozer, saying it has 8 "real cores" and has 50% throughput increase as compared to existing server products.

 

[LOL_Wut_Axel]

Not meant to be flamebait....

Is anybody still buying Phenom chips? Cant cannibalize sales that arent there can you?

If you want a cheap system with great all-around performance the Phenom II X4 955 is hard to beat at $112. The Athlon II X3 445 is pretty decent at $75 as well.

The reason why the "no benchmarks before launch means low performance" argument makes no sense is that it's how AMD has worked for years. AMD tried to make the Radeon HD 5870 as secretive as they could, and look how that turned out. The Phenom II X4 was competitive in performance with the Core 2 Quad as well, and they were quiet about it.

 

[Idontcare]

Some of us are missing the point. Intel can piss off the world and not worry about it effecting sales because they are the 800 lbs gorilla. AMD needs every friend and sale it can get. Obviously if they annoy a couple of enthusiast they won't notice, but if they annoy the whole community they will take notice and maybe perform some kind of damage control. They are well aware that some of us are just kids living in our parents basements, but others of us work in IT for medium to large companies. Some of us make meaningful purchase decisions, both server and workstation, and it is unwise to really piss us off.

They, like most companies, pay attention to their user base and try to keep them satisfied. They know that the delays are pissing off their customer base because they have eyes and ears and can see it happening. JF-AMD is one of those sets of eyes and ears. So do they care about this post in particular? Hell no! Do they care about a bunch of posts like this in numerous forums? I think so.

Obviously if the product isn't ready or their not making enough they can't improve it any more than they are already trying to. But at some point, if enough mud is getting thrown at them, they will be forced to say something.

Unfortunately for AMD, they saw just how fickle this "enthusiast" crowd was when they launched the Quad-father and it sank faster than a 1.3GHz Willamette.

AMD can't afford to be anything less than pragmatic about the situation.

JFAMD's post was pretty succinct, and quite frankly left nothing for the masses to misinterpret or misunderstand.

At this point if a person finds themselves at odds with AMD's position here regarding information relating to bulldozer then there really is no helping that person, their expectations have exceeded the bounds of reality and rationale.

 

[Idontcare]

But strictly speaking, the Osborne Effect comes into play simply by pre-announcing the new product. AMD has already Osborned their current lineup years ago by announcing Bulldozer, saying it has 8 "real cores" and has 50% throughput increase as compared to existing server products.

If you search my posts you'll note I have long argued that AMD "Osborned" their phenom II, thuban, and bulldozer products by evangelizing that "the future is fusion". It bothered me intrinsically for a long time.

But...obviously there is a difference between providing a simple roadmap versus handing out the full monte.

The 50% more performance with 33% more cores really didn't tell us jack-crap, hence you see the forums filled full of threads and threads debating the nuance of what that statement even meant. It was tantamount to saying "next year there will be new products, and just like every year preceding it, those new products next year will outperform the products we want you to buy this year".

Every industry has this cycle, be it aerospace, auto, or computers. We all expect next year's products to outperform this year's products.

But AMD is quite intent to avoid having a JC tip their hand. Intel, thanks to their monopoly I am convinced, is the only outlier in this trend of hiding your hand that spans many industries.

 

[Cogman]

Ever hear of The Osborne Effect?

Its fairly obvious that AMD has.

Wow, that is pretty sad really. I kind of wish that the early days of computing didn't see so many casualties. It would be nice if there were more players in the game today.

 

[BlueBlazer]

Llano A8-3850 (4 cores) at 2.9GHz is faster than Core i3 2100 (2 Cores + HT) at 3.1GHz.

But AMD's current quad core does not share the FPU, fetch, decode and L2 cache. Each core is a full blown core.

Unfortunately for AMD, they saw just how fickle this "enthusiast" crowd was when they launched the Quad-father and it sank faster than a 1.3GHz Willamette.

Quad FX was altogether a fiasco from the beginning, from the "demos" (thru video screens, behind closed doors, no benchmarks) to the "launch" (?!). The reviews shows why in the end. Then there's the 3GHz Phenom FX which went MIA. :\

Wow, that is pretty sad really. I kind of wish that the early days of computing didn't see so many casualties. It would be nice if there were more players in the game today.

Besides that, others that fell away from computer hardware are DEC, Data General, Honeywell, Commodore, Atari, Tandy Radio Shack, Coleco, Sinclair, Texas Instruments, etc. Silicon Graphics almost joined this group at one time. We lost quite a number in the x86 CPU business, like Cyrix (gobbled up by VIA), IBM and IDT in the last "shakeup".

 

[AtenRa]

But AMD's current quad core does not share the FPU, fetch, decode and L2 cache. Each core is a full blown core.

Dont make me analyze BD architecture again, with better Front end, better Integer execution units, 2MB L3 cache and more, BD module will have more performance than two Llano cores.

 

[GaiaHunter]

Wow, that is pretty sad really. I kind of wish that the early days of computing didn't see so many casualties. It would be nice if there were more players in the game today.

Patents are an awesome idea in some ways, but in other ways they limit competition.

 

[Cogman]

Patents are an awesome idea in some ways, but in other ways they limit competition.

I don't know that patents are TOTALLY to blame for this, but they certainly haven't helped things. The fact that the x86 instruction set just barely came out of patent protection is retarded. We need to limit tech patents to 2->5 years max.

 

[lifeblood]

Unfortunately for AMD, they saw just how fickle this "enthusiast" crowd was when they launched the Quad-father and it sank faster than a 1.3GHz Willamette.

AMD can't afford to be anything less than pragmatic about the situation.

JFAMD's post was pretty succinct, and quite frankly left nothing for the masses to misinterpret or misunderstand.

At this point if a person finds themselves at odds with AMD's position here regarding information relating to bulldozer then there really is no helping that person, their expectations have exceeded the bounds of reality and rationale.

Thats the great thing about being a customer. As fickle as we are, they just can't ignore us. We're the ones who buy their products and without us they would have nothing.

Having said that, I don't think we are at the point of AMD having to say something. They (AMD) are balancing many factors and right now they think silence is the best course, and I tend to agree with them. But at some point, if the delays go on too long, they will need to say something. When is "too long"? I don't know.

 

[Maximilian]

not really, because for the average Joe-buying-a-new-CPU, there is a "new thing" all the time and he can't tell the difference except that he needs new M/B and RAM. Average Joe doesn't know anything about release dates, target segments, process technology and so on...

Average joe isnt gonna buy a CPU, hes gonna open a bud and maybe call dell or go to PC world and ask for a computrmobile or something :\

 

[Arg Clin]

Not meant to be flamebait....

Is anybody still buying Phenom chips? Cant cannibalize sales that arent there can you?

Why wouldn't they? Just pulled the latest prices. I'm europe based, so bear with me if it's totally different in the US.

Phenom II X4 965 and i3 2100T costs about the same (~ $ 170). As the same price point I'd go with 4 PhII cores over the i3 dual core any day. In addition the AM3 mainboards offer way better bang for the $ than their LGA1155 counterparts.

Sure it doesn't beat the 2500K, but I'm talking builds where that is not required/relevant. I'm hesitant to call it low end, because both are quite decent for just about anything - even gaming.