8-core Zambezi confirmed to be priced approx. $300

[exar333]

Let me refresh your memory.

Pentium 4 3.2ghz vs. Athlon 64 3200+.

A64 was faster in games, but otherwise Pentium 4 3.2ghz HT competed well against it. With overclocking and onslaught of X2 processors, AMD really pulled ahead later. However, those X2 processors cost $400-500. The difference between a single core A64 and Pentium 4 "C" was actually fairly small at stock speeds. The gap in performance today is nothing like it was between A64 and Pentium 4.

Ironically, the situations where BD will likely be strong are exactly the ones where Pentium 4 outperformed A64 - Rendering, video encoding, etc. Yet, A64 proved to be superior for a lot of us since it was better for games - just like SB is today.

Also, Intel had 2 competing architectures at the time -- Pentium M and Pentium 4. Pentium M was superior to A64 in the mobile sector actually. Intel already knew that Netburst was end of the line since they miscalculated their ability to increase frequency. So the fact that A64 was better had a lot to do not only with AMD's excellent CPU, but Intel's failure to release a desktop Banias Pentium M processor.

Pentium 4 didn't really have great overclocking headroom compared to Athlon 64 (on a % basis) and it didn't have leading IPC either. With SB it is an entirely different situation. You have an Intel CPU with serious overclocking headroom (nearly 35-40%) + leading IPC at the same time. So even if BD can compete at stock speeds, that's still only half way there. For enthusiasts like us, in its overclocked form it would also have to be able to compete with an overclocked SB.

Great post.

I remember building my first A64 system and choosing that over the P4 775 system available. The A64 3200+ was ~$200 and the comperable P4 was the P4 3.0 or 3.2 at a similar price point. My A64 OCd to about 2.5ghz, while the P4 would OC to ~3.6ghz or so with equivalent cooling options.

The A64 was faster at what used my computer for back then (games) with about 1000mhz less speed. Encoding-wise, the processors both at stock or both at OC speeds would have been pretty similar. The A64 had about 25% OC headroom, while the P4 had 15-20%. I didn't want the heat from the P4 though...that was pretty bad.

When the dual-cores came out, the X2 was MILES ahead of the Pentium D's, but did cost a lot...I couldn't afford an X2 at the time, and didn't get a dualie until C2D came out a while later.

 

[Troll Trolling]

I think the primary point motivating all of this is the thread-title itself...the fact (confirmed per thread title) that Bulldozer will "top out" at around $300 gives us indication of where the performance must fall.

As others have rightly pointed out, if bulldozer performed such that it could command a $1000 price point in a market filled with $1000 westmere and $300 sandy bridge then without doubt AMD would price it as such. They are still a for-profit business.

Yes, sure.

I still hope that the $300 are just for the slowest octa core though.

Also, they could be trying to popularize eight core processors.

The real unknown now is the quad core's price. As someone said earlier, if it is priced the same as the i5 2400, then it's cores are performing at Sandy Bridge's level (at a higher frequency, most probably; nevertheless equally performing)

I am not going to say that it's impossible, but highly unlikely.


So you think it's more logically reasonable that AMD, which has been completely uncompetitive since C2D days, and has delayed BD for a countless number of times, will suddenly release a 3.9ghz (+ Turbo Mode) SB-like IPC 80% efficiency of a full 8-core processor for just $300? Is this like the R300 AMD HD6970 will smoke GTX580 by 30% hype all over again?

First of all, I've never said it was going to be as fast.
And yes, if AMD manages to beat Intel, when single-threaded, it is going to be almost a miracle.

But possibilities are... possible.
You know, AMD has been working on this for a long time. Since they introduced Barcelona, all the performance gained was from more cores/cache/clocks, I just think it would be utterly incompetent of them, in this giant meantime (4 years + the time they've been working on this before Barcelona), not making at least an measurable IPC gain while maintaining high clocks.

Yes, just wishful thinking. And yes, I'm being a hypocrite right now.

 

[RussianSensation]

You know, AMD has been working on this for a long time. Since they introduced Barcelona, all the performance gained was from more cores/cache/clocks, I just think it would be utterly incompetent of them, in this giant meantime (4 years + the time they've been working on this before Barcelona), not making at least an measurable IPC gain while maintaining high clocks.

I hope so too. Trust me, I was extremely disappointed in Phenom I/II after A64 and I wanted it to do well. I just couldn't believe that AMD spent 4-5 years on Phenom I after looking at its performance.

 

[GaiaHunter]

Starcraft 2 is the poster child for measuring IPC differences in games. The quad-core i5-2500 and i7-2600 spank all of the current hexacore AMD CPUs. In fact, a dual core HT i3 2100 is as fast as a hexacore 1100T Phenom.

At 1920x1080, Phenom II's minimum frames are almost 2x lower than they are of SB processors in SC2.

More cores is the new MHz myth.

Since Sc2 only use 2 cores, I don't see the point of comparing a dual HT vs an hex core - heck the i3 2100 is also almost the same speed of Intel 6 cores. And a Phenom II X2 is as fast as an X6 too.

I'm not sure for what SC2 is a poster, but certainly doesn't seem to be IPC or MHz.

21% increase in clocks yield 8% boost in FPS for AMD but for Intel 38% clock increase yields 35% performance increase?

 

[formulav8]

Third, how much stock will there be at BD's launch?

That is key for AMD. AMD will sell all they make anyways when they are released.

 

[nonameo]

21% increase in clocks yield 8% boost in FPS for AMD but for Intel 38% clock increase yields 35% performance increase?

think: turbo core

 

[Vesku]

I wonder the same thing when I see cpu charts for certain games like SC2, Gaiahunter. What exactly is going on code-wise that an Intel chip sees expected MHz->% perf gains but AMD chips see such a reduced gain rate?

 

[frostedflakes]

Let me refresh your memory.

Pentium 4 3.2ghz vs. Athlon 64 3200+.

A64 was faster in games, but otherwise Pentium 4 3.2ghz HT competed well against it. With overclocking and onslaught of X2 processors, AMD really pulled ahead of Pentium-D. However, those X2 processors cost $400-500. The difference between a single core A64 and Pentium 4 "C" was actually fairly small at stock speeds. The gap in performance today is nothing like it was between A64 and Pentium 4. Pentium 4 didn't really have great overclocking headroom compared to Athlon 64 (on a % basis) and it didn't have leading IPC either. With SB it is an entirely different situation. You have an Intel CPU with serious overclocking headroom (nearly 35-40%) + leading IPC at the same time. So even if BD can compete at stock speeds, that's still only half way there. For enthusiasts like us, in its overclocked form it would also have to be able to compete with an overclocked SB.

Ironically, the situations where BD will likely be strong are exactly the ones where Pentium 4 outperformed A64 - Rendering, video encoding, etc. Yet, A64 proved to be superior for a lot of us since it was better for games - just like SB is today.

Also, Intel had 2 competing architectures at the time -- Pentium M and Pentium 4. Pentium M was actually superior to A64 in the mobile sector. Intel already knew that Netburst was end of the line since they miscalculated their ability to increase frequency. So the fact that A64 was better had a lot to do not only with AMD's excellent CPU, but also with Intel's failure to release a desktop Dothan/Banias Pentium M processors.

If Pentium M had been available on a desktop socket as a mainstream offering, then A64 wouldn't have been as special as it turned out.

Yup, and this is something that a lot of people seem to forget. It's not so much that A64 was amazing, it's just that P4 was so underwhelming that A64 seemed amazing by comparison. Even back then they had a great architecture with Pentium M, though, which is why they used it as the basis for Core.

 

[GaiaHunter]

think: turbo core

Think: It is set clocks so turbocore is disabled.

 

[Arzachel]

As others have rightly pointed out, if bulldozer performed such that it could command a $1000 price point in a market filled with $1000 westmere and $300 sandy bridge then without doubt AMD would price it as such. They are still a for-profit business.

I think that if a two module BD can get within 10% of a i5 2400, then AMD will release some high-end SKU's just after SB-E to steal Intel's thunder.

 

[bryanW1995]

I have often skipped over that detail in my head. You are right. If Bulldozer doesn't have the Sandy Bridge handicap of not allowing overclocking except for "K" (or "unlocked") units, then certain BD SKUs may certainly prove to be a boon for enthusiasts as a throw-back to the good-old days of overclocking - cheapest reasonable performer that can then overclock to the levels of the high-end SKUs.

I hope this comes true, I guess we'll see in about two months.

That will be awesome if true, just get the lowest-clocked BE 8 core and you're gtg. My only concern is that with all the launch issues we might not be able to OC it enough to be competitive with SB.

 

[bryanW1995]

That's not what he said. He implied that if AMD needs an 8 core 3.9ghz processor to compete with a 2600k, then it can't possibly have a similar IPC per core. Why does IPC matter? Because SB ships at 3.4ghz but has a nice overclocking headroom to 4.7ghz+.

Hyper-threading only adds about 4-5% to the average performance of the 2600k. In other words, AMD is going to use an 8-core processor vs. a 4 core Intel processor.

If each BD core is as fast as a single SB core, then you will essentially get an 8 core "SB" (from AMD) vs. a 4 core SB (from Intel). But not only that, but the top CPU is expected to have 4.5ghz Turbo frequency? How are people expecting an 8 core "SB" like processor clocked at 4.5ghz? :whiste: Let me get this straight for a second......Santa Clause exists.

There are only 2 realistic possibilities:

1) BD is way behind in IPC compared to SB. As a result it needs both high clock speeds and more cores to remain competitive.
2) The 2-core module setup is nowhere near as fast as 2 full fledged cores.

Otherwise, if Bulldozer were truly 30-40% faster than SB (which surely would be if it was an 8-core processor clocked at 3.9ghz @ 4.5ghz Turbo with IPC comparable to SB), then prices would reflect this.

I don't feel like digging back through all the BD rumor threads, but I recall reading in several different threads that BD is expected to require higher clocks for the same single-threaded performance. Personally, I would be shocked to see any BD sku this year that beats a 2600k on average across a typical benchmark suite. Of course, for you and me BD will probably still be the better choice as the 8 cores will almost certainly overwhelm a 2600k in DC apps, and markfw900 is probably going crazy with anticipation right now... But for anybody not taking advantage of at least 6 + cores then I think it's going to be a tough row to hoe for AMD...again.

 

[Blue Shift]

That will be awesome if true, just get the lowest-clocked BE 8 core and you're gtg. My only concern is that with all the launch issues we might not be able to OC it enough to be competitive with SB.

That's a good point; given the process-related issues, BD's release stepping may not have that much overclocking headroom.

 

[tweakboy]

As stated above, initial pricing means nothing in terms of performance.


If it is better on programs that actually take advantage of 8 cores, but is not as good on the majority of programs.....I can't see that price holding for long.

I use DAW. It says on their site its mutlithreaded app because if you keep adding instruments then eventually that 40 percent goes to 100 percent which wont come to that. When X1 has 100 usage or when games fall below 60fps then Im have no reason to upgrade wont see a difference....happy man ... cpu usage is avg 40 percent on my projects , My gaming is all stuck at 60fps @ vsync on,@1080p 4XAA 16AF16x high quality.

not a break thru like Core 2 ,,, rite now at least.



Sonar X1 Producer

 

[Idontcare]

I wonder the same thing when I see cpu charts for certain games like SC2, Gaiahunter. What exactly is going on code-wise that an Intel chip sees expected MHz->% perf gains but AMD chips see such a reduced gain rate?

Could be code, could be microarchitecure.

Intel cpu's may have less of a bottleneck somewhere such that with this code it scales better with clockspeed.

 

[bryanW1995]

I hope so too. Trust me, I was extremely disappointed in Phenom I/II after A64 and I wanted it to do well. I just couldn't believe that AMD spent 4-5 years on Phenom I after looking at its performance.

Didn't they spend a year or two on k9 before realizing that it was a dead end and switching to k10? I think that was part of the problem, barcelona was rushed out in reaction to c2d and just fell flat on its face. At least with BD we know that they've been working on it for quite a while. And the good news is, the longer the delay the more time they've had to perfect it!

 

[SlowSpyder]

If I remember correctly, after having lost the enthusiast market, AMD priced the 4870 @ $299 and the 4850 @ $199. We all would have looked pretty silly if we had tried to guess 9800GTX performance at best, as we would have done, based on the initial pricing.

I'm certainly not saying Bulldozer will be the runaway success the 48x0 cards were in the eyes of enthusiasts. But I don't know how much we should look at pricing as a performance indicator right now.

 

[ocre]

Didn't they spend a year or two on k9 before realizing that it was a dead end and switching to k10? I think that was part of the problem, barcelona was rushed out in reaction to c2d and just fell flat on its face. At least with BD we know that they've been working on it for quite a while. And the good news is, the longer the delay the more time they've had to perfect it!

I actually think AMD self inflicted the devastating blow. They were swimming drunk on their success and just wasnt paying attention. They were so comfortable with the cash flow and riches that they decided to skimp out on R&D. they even pushed back their future CPU schedules all while the conroe was nearing the runway. Once Intel dropped the C bomb AMD was almost a year behind on their own schedule because of their ignorance and the Intel architecture was more than enough to destroy AMD. AMD was really in a bad position. They rushed and scrambled to gain any ground but the were plagued with more setbacks. the original Phenom sealed the deal. As amd scrambled intel kept tick/tock'n away. This is still the case today.

If AMD had not slacked up when they were shinning then the story couldve been different. There is no doubt that conroe wouldve still been a force for AMD but they wouldve been in a much better position. it would be 6months before AMD even had 65nm A64s out by this time intel was stacking their cores and making quads. Had AMD not pushed their schedule in their comfort the 65nm die shrink wouldve been out months before conroe. Things couldve been very different. Intel may have still been ahead but AMD really shot themselves in both arms when they got high on their glory and this sealed their fate as being a budget alternative for the past several yrs.

Sadly i was one of those ppl who actually thought AMD would come back and top intel. I was their biggest fan for many yrs....even settling for the slower performance just to support them. Only recently have i let go of my fanboy crush and can now use an intel cPU without feeling like i am "cheating".....Lol

 

[frostedflakes]

I actually think AMD self inflicted the devastating blow. They were swimming drunk on their success and just wasnt paying attention. They were so comfortable with the cash flow and riches that they decided to skimp out on R&D. they even pushed back their future CPU schedules all while the conroe was nearing the runway. Once Intel dropped the C bomb AMD was almost a year behind on their own schedule because of their ignorance and the Intel architecture was more than enough to destroy AMD. AMD was really in a bad position. They rushed and scrambled to gain any ground but the were plagued with more setbacks. the original Phenom sealed the deal. As amd scrambled intel kept tick/tock'n away. This is still the case today.

If AMD had not slacked up when they were shinning then the story couldve been different. There is no doubt that conroe wouldve still been a force for AMD but they wouldve been in a much better position. it would be 6months before AMD even had 65nm A64s out by this time intel was stacking their cores and making quads. Had AMD not pushed their schedule in their comfort the 65nm die shrink wouldve been out months before conroe. Things couldve been very different. Intel may have still been ahead but AMD really shot themselves in both arms when they got high on their glory and this sealed their fate as being a budget alternative for the past several yrs.

Sadly i was one of those ppl who actually thought AMD would come back and top intel. I was their biggest fan for many yrs....even settling for the slower performance just to support them. Only recently have i let go of my fanboy crush and can now use an intel cPU without feeling like i am "cheating".....Lol

I don't know if it I would chalk it up to arrogance. AMD would probably argue that Intel's rebate programs and other anti-competitive practices robbed them of sales and money that could have been put back into R&D to help keep them more competitive with Intel. Even back when AMD had the superior desktop architecture, they never really gained much traction with OEMs due to exclusivity deals that Intel had with big companies like Dell. Many years later the courts and regulatory agencies ruled in AMD's favor and they got settlements, but the damage had already been done and AMD was now so far behind Intel in technology that it's very hard for them to catch up.

 

[AnandThenMan]

AMD settled for $1.25 billion, the settlement should have been 10 times that at least. For that reason alone I can see why Dirk Meyer is no longer AMD's CEO. How much did Intel bribe Dell alone, it was in the billions.

 

[Skurge]

Since Sc2 only use 2 cores, I don't see the point of comparing a dual HT vs an hex core - heck the i3 2100 is also almost the same speed of Intel 6 cores. And a Phenom II X2 is as fast as an X6 too.

I'm not sure for what SC2 is a poster, but certainly doesn't seem to be IPC or MHz.

21% increase in clocks yield 8% boost in FPS for AMD but for Intel 38% clock increase yields 35% performance increase?

Yeah, im pretty sure SB has about 50% IPC advantage over Core 2, but look at this.

 

[soccerballtux]

I wonder the same thing when I see cpu charts for certain games like SC2, Gaiahunter. What exactly is going on code-wise that an Intel chip sees expected MHz->% perf gains but AMD chips see such a reduced gain rate?

Could be code, could be microarchitecure.

Intel cpu's may have less of a bottleneck somewhere such that with this code it scales better with clockspeed.

http://arstechnica.com/hardware/reviews/2008/07/atom-nano-review.ars/6
http://www.agner.org/optimize/blog/read.php?i=49

vesku: wouldn't surprise me if more of this happens at a higher level.
this sort of cheating really frustrates me. The smaller performance improvements in matlab and mathcad where it's only 1% I don't mind so much, but completely ignoring the presence of SSE registers is...unethical, IMO.

 

[bryanW1995]

AMD settled for $1.25 billion, the settlement should have been 10 times that at least. For that reason alone I can see why Dirk Meyer is no longer AMD's CEO. How much did Intel bribe Dell alone, it was in the billions.

That amount was most likely a board decision, not just dirk's. His ouster was more likely due to his refusal/inability to speak out of both sides of his mouth while consolidating his power base (yes I'm looking at you hector!). Also, I strongly suspect that early BD revelations convinced the board that they were highly unlikely to compete with intel (again) for any reasonable length of time, if at all, this round.

 

[ocre]

I don't know if it I would chalk it up to arrogance. AMD would probably argue that Intel's rebate programs and other anti-competitive practices robbed them of sales and money that could have been put back into R&D to help keep them more competitive with Intel. Even back when AMD had the superior desktop architecture, they never really gained much traction with OEMs due to exclusivity deals that Intel had with big companies like Dell. Many years later the courts and regulatory agencies ruled in AMD's favor and they got settlements, but the damage had already been done and AMD was now so far behind Intel in technology that it's very hard for them to catch up.

What???? AMD had over 40% of the market! This is unbelievable considering how AMD came to be in the x86 market. AMD was the shiz and virtually every real tech head new this. I mean the FX brand was the top dollar CPU. AMD made some poor decisions when they were at the top of their game and this was a devastating blow. The anti-competitive practices are proof that AMD had even intel shaking in their boots.

 

[ocre]

That amount was most likely a board decision, not just dirk's. His ouster was more likely due to his refusal/inability to speak out of both sides of his mouth while consolidating his power base (yes I'm looking at you hector!). Also, I strongly suspect that early BD revelations convinced the board that they were highly unlikely to compete with intel (again) for any reasonable length of time, if at all, this round.

without a doubt. I agree on both. The settlement wasnt something hat dirk did on his own and BD had to play a part in this.


-As far as the $$$ in the settlement, ppl only started complaining about how small the cash was after they seen how big nvidias settlement was. Nvidias was huge cause of patents/licenses agreements included. These we intellectual properties that intel payed a hefty price for cause they thought they couldnt live without such property and was willing to pay. Its not like AMD was ripped off or got less money on the same deal. Its really nothing like that!