Confirmed: AMD will delay BD launch again now to Oct. 2011

[AtenRa]

You do recognize that these "fair comparisons" you are seeking are artificial constructs, right?

Sans clockspeeds, sans price, sans power-consumption...what are you attempting to compare? Normalized by die-size? OK, I see utility in such an artificial construct for microarchitectural efficiency analyses but not I don't see much utility in it beyond that.

What do you think AMD intends the 4170 to compete against? 990X? SB-E? i3-2100?

Compering a smaller (Fewer transistors) processor against a bigger (More transistors), I believe anyone agree is not fair (performance wise) and will make you come to wrong conclusions about IPC, power Usage, performance/w and more.

To give you an example, let’s compere Intel Core i3 against Core i5. Both CPUs are of the same architecture but, Core i3 has two cores + HT when Core i5 has 4 cores no HT.
It will take a nice OverClock of the Core i3 part in order to be able to catch Core i5 in performance and it will most certainly that will raise Core i3 power consumption to be higher than Core i5.
Fewer transistors will need to “work” harder to match the work done by more Transistors.

So, If I will say that it will need a higher clocked Core i3 to compete against Core i5, I will come to a conclusion that Core i3 will need higher frequency, it will use more power and that will make its IPC lower than Core i5 when in fact we are talking of the same architecture for both of them.

Since Dual module four core Bulldozer will be smaller (if it would be a dual module die)and will have fewer transistors and it is expected to be cheaper than Core i5, comparing the two and coming to conclusions about Bulldozer IPC, performance and power usage in general is invalid.

Again if you will compere a Quad Module 8 core Bulldozer against Core i5 you will come to the opposite conclusions that core i5 has to be OCed higher in order to match 8 core BD and the conclusion will be that Core i5 needs higher clocks and more power than Bulldozer.

So I believe that FX-4000 series will go against Core i3, FX-6000 series will compete against Core i5 and FX-8000 series will compete against core i7 (perhaps even against some 6-core), both in price, performance and for the first time in years, in power usage too.

Edit: I do believe that the best comparison in order to see the architecture strong and weak points of BD vs SB will be to compare one BD Module against one SB Core.

 

[poohbear]

and not just october but mid october! that's over a month away.

SLAP

that's intel bitchslapping AMD, poor poor AMD. i really feel for them....

 

[BlueBlazer]

:thumbsup: AMD was competitive in <$100 segment. But what kind of strategy is that? How much $ do you think AMD made off those chips?

You may want to check out this >> Q2 CPU market share. Can easily see which are AMD best sellers.

 

[bryanW1995]

At least now we know the real reason that dirk got fired. BD is gonna be a flop of phenomenal proportions.

 

[AnandThenMan]

Bulldozer can't flop if it's never released. :\

 

[Meghan54]

You may want to check out this >> Q2 CPU market share. Can easily see which are AMD best sellers.

What really bothers me about that chart, which includes all their processors like their server chips, is this: "asp of $50.31."

That ain't good.

 

[formulav8]

Sept. 13 overlaps Microsoft's BUILD conference. So Win8 related?

Maybe their new gpu's will support win8 with dx12? Or is Win8 still going to be dx11?

 

[Rvenger]

Judging by those clock speeds in the article. AMD was clearly tweaking the frequencies to match/beat intels SB CPUs. FX-4170 is just a joke, Im just hoping we will be able to unlock the FX-4170's extra cores (If any) and it would be worth the money.

 

[Black96ws6]

Anand just posted this: http://www.anandtech.com/show/4749/amd-ships-bulldozer-for-servers-desktops-to-follow-in-q4

With AMD saying that Zambezi will ship in September but launch in Q4 it's safe to say that rumors of a September launch date aren't true. Q4 encompasses October, November and December. My money is on an early-to-mid Q4 launch.

 

[nemesismk2]

this is disappointing news from amd I have already replaced my phenom ii x2 560 cpu with a intel core i3 2100. it looks like the phenom ii x4 955 is going to get replaced with a intel core i5 2400 or maybe a 2500 instead of something from amd

 

[Stuka87]

this is disappointing news from amd I have already replaced my phenom ii x2 560 cpu with a intel core i3 2100. it looks like the phenom ii x4 955 is going to get replaced with a intel core i5 2400 or maybe a 2500 instead of something from amd

Out of curiosity, what are you doing that makes the 955 feel too slow for use? An SB i5 is certainly faster, but is it worth the cost? You could just OC the 955 and make it last longer. It wont be quite as fast as the i5 2400, but its about $350 cheaper than a new CPU/Mobo.

According to Anands post, the AM3+ part will ship before the end of the month. But that means we won't see until mid October. And most likely is smaller quantities. But I am looking forward to benches of the Opterons.

 

[chihlidog]

JF-AMD has made comments over on [H] that could seem to imply that it wont even release in October. He affirmed that it's a Q4 launch. Someone stated that fiscal Q4 is over in 7 weeks, and JF-AMD responded: "Where did the 7 weeks come from? Our fiscal quarters match calendar quarters."

So it looks as though it may well be past October. Take it as you will.

 

[Terzo]

So, according to the article Anand listed as a source, Interlagos began production in August, and recently began shipping. Availability is expected as Q4.

Initial production of the world's first 16-core x86 processor, codenamed "Interlagos," began in August and shipping to customers is already underway. Compatible with existing AMD Opteron™ 6100 Series platforms and infrastructure, "Interlagos" is expected to launch and be available in partner systems in the fourth quarter of this year.

They actually mention nothing about Zambezi, so I'm not sure what Anand's source is when he says it will ship in September.

Doing some extrapolation, if the server product is shipping now and is expected to be available in Q4, that's basically at least a month from shipping to availability. If what Anand says is true and AMD begins shipments for Zambezi in September (probably later in the month), we wont see it until November at the earliest.

At this point, I'll believe the desktop variant has shipping when I can go to Microcenter and buy one off the shelf. I'm going to go pessimistic on this and say I don't expect to see Zambezi until January, a full year after Sandy. At that point Ivy is only a few months away.

Considering (for me) a processor upgrade is not a need but a desire, I think 2012 is going to see my buying my first blue chip.

 

[BlueBlazer]

JF-AMD has made comments over on [H] that could seem to imply that it wont even release in October. He affirmed that it's a Q4 launch. Someone stated that fiscal Q4 is over in 7 weeks, and JF-AMD responded: "Where did the 7 weeks come from? Our fiscal quarters match calendar quarters."

So it looks as though it may well be past October. Take it as you will.

Speculation, its possible that AMD needs another stepping. :\

 

[chihlidog]

Speculation, its possible that AMD needs another stepping. :\

Ive seen that, but dont necessarily buy it. I mean, how could the results not be favorable when compared to an 1100t??? That doesnt make sense to me. I doubt it will have the IPC of Sandy Bridge, but I'm willing to bet it's at least a bit improved over Phenom II. No way AMD would release something that goes BACKWARDS.

Unless that poster just wasnt satisfied with the jump in performance, and expected it to be more.

 

[Stuka87]

No way AMD would release something that goes BACKWARDS.

At this point, I don't think AMD has any choice but to release BD, regardless of performance. They have so much time and money tied up into its development, its going to have to be released to get something out of the investment that they have in it.

I am hoping that it turns out to be better than most are saying. But if it does turn out to be not much of/any gain over Thuban, it could be very bad news for the future of AMD.

Although, Intel managed to pull off releasing a CPU that was slower than its predecessor when it released the Pentium. Clock speed is the only thing that made it faster than the Pentium 3. And that same thing is what gave AMD the opportunity to jump ahead of Intel. But today, Intel is already well ahead of AMD.

 

[Idontcare]

Ive seen that, but dont necessarily buy it. I mean, how could the results not be favorable when compared to an 1100t??? That doesnt make sense to me. I doubt it will have the IPC of Sandy Bridge, but I'm willing to bet it's at least a bit improved over Phenom II. No way AMD would release something that goes BACKWARDS.

Unless that poster just wasnt satisfied with the jump in performance, and expected it to be more.

65nm Athlon X2 went backwards from 90nm Athlon X2. IPC dropped and top clockspeeds could not exceed the then top-90nm X2 clockspeeds.

180nm Willamette was internally delayed until they got the clockspeeds high enough such that it would beat their own P3 chips already on the market.

90nm Prescott was a step back from the top-speed capabilities of 130nm Northwood at the time.

These are NOT the result of dumb mistakes or silly project management. These are simply the results of calculated risks.

The design envelope for these processors is forecast some 3-4yrs in advance of them exiting the fab.

If they minimize risk and take the "tried and true" path then the incremental performance gains will be quite low, and still won't be selling for 4 yrs.

So they have to make aggressive decisions and pile on the risk. Sometimes it works well and they get a jump on the competition - K7 and C2D - and sometimes it lands with a flop like Phenom and Prescott.

So I wouldn't say it is impossible for AMD's bulldozer to turn out to simply be a bulldoozy.

But I also wouldn't blame AMD management for it if it turns out to be that way, they had to take risks, no risk means guaranteed out-of-business, lots of risk merely means you stand a chance of staying in business.

Look at DEC, MIPS, Qimonda, Agere, Spansion, Hynix, etc.

 

[chihlidog]

True enough, IDC.

I'll put this here too, though. JF-AMD stated that he had seem many of the "benchmarks" that have been "released", and that some were representing its performance as lower than what it really is, and some higher than what it really is.

I havent seen a benchmark that showed it to be slower than PHII, so I'm somewhat hopeful.

 

[Idontcare]

True enough, IDC.

I'll put this here too, though. JF-AMD stated that he had seem many of the "benchmarks" that have been "released", and that some were representing its performance as lower than what it really is, and some higher than what it really is.

I havent seen a benchmark that showed it to be slower than PHII, so I'm somewhat hopeful.

What we know to be true is that JFAMD has stated IPC did not/does not go down.

That means at worst the chips will perform the same as PhII on a core-normalized and clockspeed-normalized basis.

I believe John would not make such a statement unless he knew for fact it was true.

So I have to conclude the delay of Zambezi is entirely an economical one - yields are too low at this time once you factor in both functional yield (GloFo's job to improve this) and parametric yield (binning, AMD decides the brackets here).

 

[chihlidog]

What we know to be true is that JFAMD has stated IPC did not/does not go down.

That means at worst the chips will perform the same as PhII on a core-normalized and clockspeed-normalized basis.

I believe John would not make such a statement unless he knew for fact it was true.

So I have to conclude the delay of Zambezi is entirely an economical one - yields are too low at this time once you factor in both functional yield (GloFo's job to improve this) and parametric yield (binning, AMD decides the brackets here).

What I took it (hopefully) to mean, since I'm not aware of any benchmarks floating around that show it to be slower than PHII, is that even if not by much, it will be faster clock per clock. I could be reading way too much into it, but I feel like we're safe in thinking it's faster than the slowest faked benchmarks floating around. IF that's the case, the slowest benchmarks I've seen arent that bad, and I'm hoping for a win for AMD here.

 

[Tuna-Fish]

You may want to check out this >> Q2 CPU market share. Can easily see which are AMD best sellers.

What really suprised me on that chart is that Ontario is one quarter of their unit shipments and one fifth of their revenue. Given that the chip less than 1/4th of a athlon X4 in size and has the asp of 33$, I think it's entirely reasonable that Ontario is much more than a fifth of their profits. That's a bet that paid off.

 

[Vesku]

So I have to conclude the delay of Zambezi is entirely an economical one - yields are too low at this time once you factor in both functional yield (GloFo's job to improve this) and parametric yield (binning, AMD decides the brackets here).

I'd say it's mostly yields, apparently they are even having to specifically prioritize the 16 core server chips over the 8 core so that they can deliver in the volumes required by Cray and the like. Interesting considering we are probably talking a few hundred thousand chips. Highlights once again how Intel gets business not only for making fast chips but also for it's relative delivery reliability.

How severe the other issues are can be gleaned from how many months we see between BD launch and the release of a respin.

 

[Black96ws6]

here's another new article that may shed some more light on things:

http://semiaccurate.com/2011/09/07/bulldozer-finally-shipped-last-week/

...Bulldozer delays come down to a lot of little things, and three big ones.

The first one is that there is a new stepping coming, SemiAccurate is hearing mid- to late Q1/2012 for the next rev. That rev is said to bump performance, specifically integer performance, up by quite a bit, and possibly improve clocks too. Either way, it looks like that stepping is the one to keep an eye out for. It isn't a Barcelona type fiasco, but it isn't an HD4870 launch either.

Next on the list is yield, or lack thereof....

 

[gmaster456]

Continued:

12) January 2009 - AMD finally launches Phenom II X4 @ $234 and $275. Core i7 920 sells for $284. "The results of our tests show that the top Phenom II X4 processors can only be worthy rivals to the Core 2 Quad CPUs from the “junior” Q8000 series. Unfortunately, Phenom II X4 cannot yet do better than that.". So not only was Phenom II unable to compete with its direct competitor i7, but it wasn't even able to compete with Q9xxx series CPUs from 45nm C2Q generation.

13) February 2009 - Phenom II X4 Overclocking - can't beat overclocked Q6600 / Q9300 processors. Pretty underwhelming considering Intel's "current" generation at this time is now i7. This is really a sad state of affairs since Q6600 was a <$300 CPU in August of 2007....

14) March 2009 - The beginning of "New AMD Strategy" --> If we can't compete on price or performance, let's throw in more "free" cores at the same price. Suddenly, Phenom II X3/X4 finds itself competing with C2D and lower-end Q8xxx series from Intel. AMD is really struggling here.

15) August 2009 - AMD launches Phenom II X4 965. Performance can only match a Q9550 @ 2.83ghz. Phenom II X4 965 costs $245. I quote: " As we have seen during our tests, the new Phenom II X4 965 working at 3.4 GHz frequency is pretty much as fast as Core 2 Quad Q9550 at 2.83 GHz nominal speed and fall behind Core i7-920 with even lower nominal frequency of 2.66 GHz. So, AMD CPUs lose to Intel competitors quite significantly in IPC (instructions per clock)"

16) September 2009 - S1156 launches: The end of Phenom II X4's relevance for a vast majority of users: $199 Core i5 750. "This platform immediately turns Core 2 and Phenom II CPUs into outdated solutions that can only be of interest in the sub-$200 category."

I don't think I need to keep going. Basically outside of Free / $10 mobo + Phenom II X4 / X6 MicroCenter deals and AMD's only shining bright star: Athlon II X4 6xx series processors, the last 5 years saw complete Intel domination.

Your forgetting something. AMD literally did all they could, but Intel had tons of patents keeping AMD from doing anything more than what they did. Only recently have the doors been re opened and AMD is finally able to do something that can finally counter Intel which has been the $150+ CPU juggernaut for the last 5 years.

 

[inf64]

Continued:

12) January 2009 - AMD finally launches Phenom II X4 @ $234 and $275. Core i7 920 sells for $284. "The results of our tests show that the top Phenom II X4 processors can only be worthy rivals to the Core 2 Quad CPUs from the &#8220;junior&#8221; Q8000 series. Unfortunately, Phenom II X4 cannot yet do better than that.". So not only was Phenom II unable to compete with its direct competitor i7, but it wasn't even able to compete with Q9xxx series CPUs from 45nm C2Q generation.

13) February 2009 - Phenom II X4 Overclocking - can't beat overclocked Q6600 / Q9300 processors. Pretty underwhelming considering Intel's "current" generation at this time is now i7. This is really a sad state of affairs since Q6600 was a <$300 CPU in August of 2007....

14) March 2009 - The beginning of "New AMD Strategy" --> If we can't compete on price or performance, let's throw in more "free" cores at the same price. Suddenly, Phenom II X3/X4 finds itself competing with C2D and lower-end Q8xxx series from Intel. AMD is really struggling here.

15) August 2009 - AMD launches Phenom II X4 965. Performance can only match a Q9550 @ 2.83ghz. Phenom II X4 965 costs $245. I quote: " As we have seen during our tests, the new Phenom II X4 965 working at 3.4 GHz frequency is pretty much as fast as Core 2 Quad Q9550 at 2.83 GHz nominal speed and fall behind Core i7-920 with even lower nominal frequency of 2.66 GHz. So, AMD CPUs lose to Intel competitors quite significantly in IPC (instructions per clock)"

16) September 2009 - S1156 launches: The end of Phenom II X4's relevance for a vast majority of users: $199 Core i5 750. "This platform immediately turns Core 2 and Phenom II CPUs into outdated solutions that can only be of interest in the sub-$200 category."

I don't think I need to keep going. Basically outside of Free / $10 mobo + Phenom II X4 / X6 MicroCenter deals and AMD's only shining bright star: Athlon II X4 6xx series processors, the last 5 years saw complete Intel domination.

Let's see:
Average of a dozen desktop benchmarks across 3 generations of AMD and intel CPUs.

1) Phenom I vs C2Q (65nm both). Q6600 2.4Ghz 100pts - X4 9750 2.4Ghz 94.1pts. 4C Merom is 6.2&#37; faster on average than 4C Agena at the same clock in desktop workloads.

2)Phenom II vs C2Q (45nm both). QX9650 3.0Ghz 131.4pts - X4 945 3.0Ghz 123pts. 4C Penryn is 6.8% faster on average than 4C Deneb at the same clock in desktop workloads.

3)Phenom II X6 w/Turbo vs Nehalem 4C/8T w/Turbo (45nm both and both supporting more than 4 threads per CPU). Core i7 960 3.2Ghz 185.5pts - Phenom II X6 1090T 3.2Ghz 161.6pts. 4C/8T Nehalem with Turbo is 14.7% faster on average than 6C Thuban with Turbo at the same clock in desktop workloads.

4)Phenom II X6 w/Turbo vs next gen Core (SandyBridge) 4C/8T w/nextgen Turbo (45nm AMD vs nextgen 32nm intel;both supporting more than 4 threads per CPU). Core i7 2600K 3.4Ghz/3.8Ghz 213.4pts - Phenom II X6 1100T 3.3Ghz/3.7Ghz 165.8pts. 4C/8T Nehalem with Turbo is 28.7% faster on average than 6C Thuban with Turbo but not at the same clock, in desktop workloads. SB usually runs at close to 3.5Ghz or 3.6Ghz since 4C Turbo allows for this to happen.

5) next gen 8C/8T Zambezi(Bulldozer core) w/nextgen Turbo vs next gen Core (SandyBridge) 4C/8T w/nextgen Turbo (32nm for both CPUs;both supporting 8 execution threads per CPU). Core i7 2600K 3.4Ghz/3.8Ghz 213.4pts - Zambezi FX8150 3.6Ghz/4.2Ghz ???pts. Zambezi needs to be around 30% overal faster,on average, than 3.3Ghz X6 (in desktop workloads) in order to be faster under the same conditions than Core i7 2600K.