Rumour: Bulldozer 50% Faster than Core i7 and Phenom II.

[Abwx]

Not likely.

BD is consumer and server CPUs; the latter is $$$$.

Right, but the server version was scheduled for Q3.

Anyway, Llano and Bobcat will surely account for at least 90%
of their mobile and desktop Cpus production next year..

 

[pantsaregood]

Than I started reading what both sides were measuring e-penis with . At the time I had just put together my first PC The very box I am still using online . I started with the 2.4 p4C and overcloocked to 3.1ghz . Relized I need to water cool and built my first complete home built system . I ran Air alot more tho . I really at the time didn't like the water setup for the time. But this was the same time hammer came out and the AMD guys were complete asses. The Bench marks I got @3.1ghz were killing the best hammers out there. at the time I believe 2.2 hammer and not a clocker. In media I destroyed it . In game it won but I was way closer than review site I used 2 512mb sticks corsair bh5 @ 2 2 2 2 5 timing Idon't recall the what memory speed I used but it was realy high for those sticks at that setting . I am still using those sticks same setting except I am at lower V and running at 200.

I can't quite make out what you're trying to say, but it seems you're implying that Pentium 4s outperformed Athlon 64s, which is simply untrue on every imaginable level.

Netburst didn't "destroy" K8 in any way. In fact, the high end (EE and 3.4+ GHz) Pentium 4s were the ONLY models to outperform K8 in audio encoding. Performance in encoding wasn't consistently in Intel's favor, and at the times it was, it was generally by under 5%. That said, when in AMD's favor, it was equally negligible.

Arguing that Netburst had anything on K8 is just as dumb as arguing that K10 had anything on Conroe.

 

[RussianSensation]

Regardless if the thread title is based off a rumor, this argument is so wrong it hurts. Under most conditions we would expect the 8 core Bulldozer, under the pretenses of this rumor, to be 50% faster than the 4 core 2600k in applications which are heavily multithreaded scenarios.

Who is "we" exactly? Also, saying "50% faster in multi-threaded scenarios" and "50% faster" (as the title states) are totally different. If you think BD will be 50% faster than 2600k at $320, on average, you are a dreamer...In case you forgot, most apps don't use 8 threads anyway, which is why the 2600k is no better than 2500k for the majority. Therefore, to say that under most conditions BD it will be 50% faster and then throwing in "multi-threaded applications" is a loaded statement since most apps are not multi-threaded. The vast majority of applications in fact do not use more than 4 threads. So in essence you are suggesting that BD will be faster in most apps the majority of users do not care for anyway. In fact, most programs don't even scale well enough with 4 cores just yet as i3-2100 still beats the X4 955 in a variety of "common" tasks people actually care about and it doesn't have 4 true cores. If you really need the fastest processor for rendering since time is $ to you, you'll be upgrading to X79 chipset anyway where the performance will be untouchable.

Heavily multithreaded applications are exactly the area where the 980X beat the 2600K, sometimes significantly. So tell me again... what the hell were you thinking when you made this implication?

In which real world benchmarks outside of rendering or video encoding does the 980X beat the 2600k? I don't think you are reading the same reviews as I am. In Anandtech's review I linked, the 980X only won 13 tests but 2600k won 18 tests. That makes the 2600k the faster processor overall.

Xbitlabs had similar findings against the 990X.

At Bit-Tech, 2600k beat the 980X in every benchmark outside of Cinebench and WPrime. So that's 3 professional websites which showed that 2600k overall is faster than the 980X.

Unless all you do is calculate Pi to 32 million places or run rendering apps or x264 encoding 24/7, then a 2600k is faster than a 980X. By that understanding, claiming that BD will be 50% faster than a 2600k is implicit that it will also be at least 50% faster than 980X for 90% of users who could care less about Pi times or rendering scenarios (a professional application).

If the original title specifically said that BD will be up to 50% faster in Cinebench or WPrime or x264 HD 2nd pass, then I wouldn't have any issues with it. But right now the "50% faster" claim sounds like it's an "average" which makes it outright misleading for people who don't want to read 105 pages.

 

[cusideabelincoln]

Did you even read what I wrote or do you just enjoy making straw man arguments?

Who is "we" exactly? Also, saying "50% faster in multi-threaded scenarios" and "50% faster" (as the title states) are totally different. 1. If you think BD will be 50% faster than 2600k at $320, on average, you are a dreamer...In case you forgot, most apps don't use 8 threads anyway, which is why the 2600k is no better than 2500k for the majority. 2A. Therefore, to say that under most conditions BD it will be 50% faster and then throwing in "multi-threaded applications" is a loaded statement since most apps are not multi-threaded. The vast majority of applications in fact do not use more than 4 threads. 2B. So in essence you are suggesting that BD will be faster in most apps the majority of users do not care for anyway. In fact, most programs don't even scale well enough with 4 cores just yet as i3-2100 still beats the X4 955 in a variety of "common" tasks people actually care about and it doesn't have 4 true cores. If you really need the fastest processor for rendering since time is $ to you, you'll be upgrading to X79 chipset anyway where the performance will be untouchable.

1. I never said that and the OP never did either.

2A. Who ever said "under most conditions?" Who was talking about the 2600k? You are just making up things now. And "multi-threaded" is not a loaded statement, at all. It is a specific statement with a specific meaning. Being vague and making things up is what you are doing in creating "loaded statements".

2B. What. The. Hell. No, I'm not suggesting that. I don't know where you got that. Even the OP doesn't suggest that. This is one perfect example of a straw man argument.

In which real world benchmarks outside of rendering or video encoding does the 980X beat the 2600k? I don't think you are reading the same reviews as I am. In Anandtech's review I linked, the 980X only won 13 tests but 2600k won 18 tests. That makes the 2600k the faster processor overall.

Xbitlabs had similar findings against the 990X.

At Bit-Tech, 2600k beat the 980X in every benchmark outside of Cinebench and WPrime. So that's 3 professional websites which showed that 2600k overall is faster than the 980X.

Wow did you even read what I wrote? I specifically said multithreaded benchmarks and I was not referring to the overall average. DO YOU NOT KNOW WHAT IT MEANS WHEN SOMEONE SPECIFIES THE CONTEXT ABOUT WHICH THEY ARE TALKING? It means I'm not talking about the context in which you specify (the average). So here is a fine straw man.

Unless all you do is calculate Pi to 32 million places or run rendering apps or x264 encoding 24/7, then a 2600k is faster than a 980X. By that understanding, claiming that BD will be 50% faster than a 2600k is implicit that it will also be at least 50% faster than 980X for 90% of users who could care less about Pi times or rendering scenarios (a professional application).

The logic here is infallible. How do you not see how wrong you are? Besides which, rendering, encoding, and the like are valid uses for a processor.

If the original title specifically said that BD will be up to 50% faster in Cinebench or WPrime or x264 HD 2nd pass, then I wouldn't have any issues with it. But right now the "50% faster" claim sounds like it's an "average" which makes it outright misleading for people who don't want to read 105 pages.

This is a RUMOR! Rumors don't have to be specific; they are to be taken with a grain of salt. Even when taken seriously, you have to try and understand the nature of the rumor, which is what has been done in this thread until people got tired of talking about it.

Actually it's been so long since I've read the original OP... it says nothing about the 2600k. Go ahead. Read the god damn OP again. They mention the previous generation Core i7 950 as Bulldozer potentially POTENTIALLY being 50% faster than.

So it's not hard to imagine an 8 core processor being 50% faster than last generation's 4 core processor. The 980X is already 50% faster in select applications than the 950.

So god damn stop making up arguments I get sick of seeing you go on about your straw mans. The title of a thread can only be so specific, and this thread started 5 months ago. Stop whining about the title of a rumor article.

edit: Now I'm actually laughing about how you're going on about the 2600K, when the original article that started this thread was talking about the previous generation Core i7 950.

 

[jvroig]

I thought servers don't need to run at high clock speeds to be effective though? Or does this whole stepping issue stem from the fact that BD isn't energy efficient enough?

We can't generalize about "servers not needing high clockspeed". It certainly depends on "what the server is serving", so to speak.

Well, I'm not really sure, but I'm pretty sure general consumers are buying Llano way more than Bulldozer. For desktops AND laptops I think it'll be a hit with OEMs or budget conscious consumers.

Consumers, yes. For the consumer market, you are no doubt correct. What Exar is saying, however, is that the server market commands a much higher premium, so higher margins/greater profit per unit sold.

I cannot remember the exact figures, but last I remember, AMD's margins from servers are still significant, and their share is so small so there is a tremendous growth opportunity, and it's theirs if only BD delivers - something Llano won't be able to compete in.

 

[Vesku]

Yeah, don't think RussianSensations read your post correctly. He basically says "no you are wrong" and then goes on to back up your assertion with his references. It's true 2600K is the king of the 4 cores but certain specific applications can be threaded enough to utilize more. As for the actual thread title, I think it's only in these heavily threadable circumstances that the prediction can at all make sense. Pretty sure that was the consensus near the very beginning. Though in a thread this large I can see how we might go in circles.

 

[Abwx]

Yeah, don't think RussianSensations read your post correctly. He basically says "no you are wrong" and then goes on to back up your assertion with his references. It's true 2600K is the king of the 4 cores but certain specific applications can be threaded enough to utilize more. As for the actual thread title, I think it's only in these heavily threadable circumstances that the prediction can at all make sense. Pretty sure that was the consensus near the very beginning. Though in a thread this large I can see how we might go in circles.

Basically, RussianTroll is...trolling !! around AMD s products.
Indeed, he created his own thread for this purpôse, a great idea
since all BS and subsequent trolling is kept out of this one..

As for BD launch, AMD s said at a Computex conference
that it will be for next month, that is, july..

 

[qurious76ss]

Basically, RussianTroll is...trolling !! around AMD s products.
Indeed, he created his own thread for this purpôse, a great idea
since all BS and subsequent trolling is kept out of this one..

As for BD launch, AMD s said at a Computex conference
that it will be for next month, that is, july..

AMD had a slide that said 60-90 days for Zambezi, so more then likely Aug-Sept release for BD launch. The question now is why the delay.

 

[AtenRa]

With a new 32nm process and only one fab it is very difficult to produce both Llano and Bulldozer at the same time and have enough dies for both CPUs, I believe the delay has to do with a manufacturing limit and not with BD itself. I could be wrong.

 

[qurious76ss]

I would think that the higher ASP product would become a priority to increase margins but then again not sure how the GF deal is structured and how much or how exactly they pay per wafer or die etc..

 

[darkswordsman17]

I would think that the higher ASP product would become a priority to increase margins but then again not sure how the GF deal is structured and how much or how exactly they pay per wafer or die etc..

They've already started shipping Llano to OEMs, which means constraining supply of it right now could be harmful. It might actually be good news for AMD as well, as it might mean they're getting high feedback/demand for Llano and so they need to produce all they can of them. Then again, it might mean manufacturing constraints (GF not able to produce enough or other issues).

With a new 32nm process and only one fab it is very difficult to produce both Llano and Bulldozer at the same time and have enough dies for both CPUs, I believe the delay has to do with a manufacturing limit and not with BD itself. I could be wrong.

Difficult to say. I'd like to see some leaked benchmarks of current Bulldozer so we can see if it really is a performance issue or if there's something else going on. AMD saying Bulldozer isn't performing where they want it doesn't necessarily mean its performing bad, they might have wanted it to clock higher than Sandy Bridge for bragging rights (imagine their marketing department salivating that they both have more cores and higher gigahertz). It could mean that its competitive now, but they want a decisive victory. I think they need that to win over OEMs (especially since they can go, look Llano is good, but then what? Intel's already clearly laid out their plans, we need to see results or we'll just stick with Intel).

If its already clocking 3.8-4.4, I don't imagine that they'll be going that much higher, and that's almost certainly not going to make too much difference in performance anyway. So the reports that they think a new stepping for better clocks (which will apparently make them happy with its performance) to me indicates that Bulldozer shouldn't be a failure, even if its not the messiah chip either.

They might just be taking extra steps to get the most out of it, which I think would be important considering that it is their future (as their APUs are also going to Bulldozer). Having Bulldozer and Llano be a success, could really pave the way for AMD's next couple of years. I could see them viewing Bulldozer as betting the farm, and so they better get it right or else they could be in big trouble. It would look good to OEMs if AMD can say, look, we already have the better GPU, and now we have a more competitive CPU as well.

Too bad Anand doesn't get inside info on AMD as much as he used to, so we just have to speculate.

 

[3DVagabond]

If BD is faster single thread than SB I'll be more than surprised. (Intel will crap their corporate pants too.) I think most people don't think that. On the other hand, don't buy an 8 core processor if you don't run apps that will take advantage of the capability.

I'm hoping that BD is appreciably faster than SB in apps that can max out all 8 cores. If not, I'll be getting a SB, or possibly IB, instead. Even working with 3D modeling apps single thread performance is important. If SB thrashes BD single thread, and BD is only a bit faster maxed out, SB will still be a better option. Most modeling apps still only use 1 core for a lot of their operations. Editing, UV unwrapping, textures/materials, GUI, etc.

Edit: 50% faster when rendering would make BD worth it for me though. Rendering is terribly time consuming.

 

[Abwx]

AMD had a slide that said 60-90 days for Zambezi, so more then likely Aug-Sept release for BD launch. The question now is why the delay.

Late summer, so august as a worse case scenario.

 

[AtenRa]

AMD Computex Taipei 2011 Press Conference Part 1

http://www.youtube.com/watch?v=mSG_F-WjjRA&feature=BFa&list=ULYjPpPXK84wQ&index=19

 

[AtenRa]

Ok my theory is this,

Initially AMD wanted to release Llano in Q1 2011 and launch BD at Q2. They had some issues with the 32nm process earlier on and they missed that window by one Quarter.

OEMs know and AMD knows that Llano APUs occupies a larger market share than High End CPUs like Bulldozer, so they chose to ramp up Llanos production.

What happened could be pressure from OEMs, ODMs for more Llano APUs in order to have a lot o products for Summer and the Back to School season in September, so AMD chose to use all the 32nm production capacity to manufacture more Llano APUs. As i have said before GloFo’s 32nm SOI HKMG process is new and they only have one fab and releasing both Llano and BD at the same time could hurt Llanos production.

Those 60-90 days will give a more mature 32nm process for BD manufacturing and even if BD will be released in late August or early September it could take the performance crown until Intel’s SB-E 6-Core will be introduced in Q4.

Server BD will still be a Q3 release so no worries there.

Ps: If AMD already have Trinity APUs manufactured (BD core APUs) then it seams that BD core don’t have technical issues.

http://www.youtube.com/watch?v=NbIboX36Lhs&feature=BFa&list=ULYjPpPXK84wQ&index=17

 

[Abwx]

Ok my theory is this,

Initially AMD wanted to release Llano in Q1 2011 and launch BD at Q2. They had some issues with the 32nm process earlier on and they missed that window by one Quarter.

OEMs know and AMD knows that Llano APUs occupies a larger market share than High End CPUs like Bulldozer, so they chose to ramp up Llanos production.

What happened could be pressure from OEMs, ODMs for more Llano APUs in order to have a lot o products for Summer and the Back to School season in September, so AMD chose to use all the 32nm production capacity to manufacture more Llano APUs. As i have said before GloFo’s 32nm SOI HKMG process is new and they only have one fab and releasing both Llano and BD at the same time could hurt Llanos production.

Those 60-90 days will give a more mature 32nm process for BD manufacturing and even if BD will be released in late August or early September it could take the performance crown until Intel’s SB-E 6-Core will be introduced in Q4.

Server BD will still be a Q3 release so no worries there.

Ps: If AMD already have Trinity APUs manufactured (BD core APUs) then it seams that BD core don’t have technical issues.

http://www.youtube.com/watch?v=NbIboX36Lhs&feature=BFa&list=ULYjPpPXK84wQ&index=17

That sound right, and probably that AMD was talking about "Llano first"
when they said that they were directing in the short term to price/perfs competition rather than pure performances one.

Of course, some have interpreted it as BD not being as good as SB....:biggrin:

 

[BigDH01]

Late summer, so august as a worse case scenario.

Given AMD's track record regarding major CPU releases, I'd say August is best-case scenario. I would be surprised if you could actually go to Newegg and buy one of these things before September.

 

[Phynaz]

Basically, RussianTroll is...trolling !!

Pot....Kettle.

 

[Gigantopithecus]

Ok my theory is this,

Initially AMD wanted to release Llano in Q1 2011 and launch BD at Q2. They had some issues with the 32nm process earlier on and they missed that window by one Quarter.

OEMs know and AMD knows that Llano APUs occupies a larger market share than High End CPUs like Bulldozer, so they chose to ramp up Llanos production.

What happened could be pressure from OEMs, ODMs for more Llano APUs in order to have a lot o products for Summer and the Back to School season in September, so AMD chose to use all the 32nm production capacity to manufacture more Llano APUs. As i have said before GloFo’s 32nm SOI HKMG process is new and they only have one fab and releasing both Llano and BD at the same time could hurt Llanos production.

Those 60-90 days will give a more mature 32nm process for BD manufacturing and even if BD will be released in late August or early September it could take the performance crown until Intel’s SB-E 6-Core will be introduced in Q4.

Server BD will still be a Q3 release so no worries there.

Ps: If AMD already have Trinity APUs manufactured (BD core APUs) then it seams that BD core don’t have technical issues.

http://www.youtube.com/watch?v=NbIboX36Lhs&feature=BFa&list=ULYjPpPXK84wQ&index=17

I think this analysis makes a lot of sense. Facts (not rumors) are, early benchmarks with beta drivers show Llano wiping the floor with current-gen Intel HD on-chip graphics, and the A8 APU at least trades blows with the i3-2100. The CPU capabilities of an i3-2100 with lower-end discrete GPU capabilities takes care of the computing needs of what? 90% of desktop users? Aside from high-end gaming rigs, powerful research computers, and hardcore HD encoders, I can't think of anyone I'll build an Intel-based rig for once Llano hits the streets. And none of those people are typical desktop users.

I agree with AtenRa, AMD's constrained by production capacity and they're making more Llanos because they're going to sell more Llanos.

 

[Dribble]

If you have constrained production capability you make the chips with the highest profit margins first - that would be the bulldozer opterons. LLano is a low end consumer product and bound to have wafer thin margins hence to make much money you're going to have to sell a lot of them - not something you can do if you're production constrained.

 

[AtenRa]

Llano is 226mm2 when BD is more like 260-290mm2, GloFos 32nm is a new process and yields will not be at the optimum level. Like Graphic Cards, you always produce the smaller die first in order to raise yields.

Llano will go to both Desktop and Laptop so that's more volume than BD would do and one more thing, Laptop CPUs are more expensive and AMD will make more money from them than BD.

I could be wrong

 

[Gigantopithecus]

If you have constrained production capability you make the chips with the highest profit margins first - that would be the bulldozer opterons. LLano is a low end consumer product and bound to have wafer thin margins hence to make much money you're going to have to sell a lot of them - not something you can do if you're production constrained.

Sure, you do that without consideration of other factors.

As I said, Llano is clearly superior to Intel's low-end Sandy Bridge offerings, and stands to make major inroads against Intel's dominance of the mobile market. Very, very few people need more than an i3-2100 equivalent in their laptop, so AMD is confident that they will sell a lot of Llano chips. Furthermore, the mobile market is growing, while the desktop market is at best stagnant, maybe shrinking. Still further, AMD holds its own in the budget and even low-to-mid range of desktops, but it's been getting trounced in the mobile market. Back to school shoppers buy laptops, not desktops, and AMD wants to be eating that largesse come fall.

I think it's clear at this point that Bulldozer will not be hands-down better than the higher-end SB chips, so why make a bunch of chips that might or might not sell, even if the margins are better? There are no corporate upgrade cycle timings like back to school, either, so whether Bulldozer is available in July or October is not as important as whether Llano hits the market in July or October.

 

[psolord]

I am sorry if this has been posted before, but if it is, I missed it!

Is it real?

 

[Arachnotronic]

A chip with a max TDP of 186W? That's funny.

 

[Gundark]

If you have constrained production capability you make the chips with the highest profit margins first - that would be the bulldozer opterons. LLano is a low end consumer product and bound to have wafer thin margins hence to make much money you're going to have to sell a lot of them - not something you can do if you're production constrained.

Maybe taking biger market share will have its benefits on stock market. This is important as well, at least for investors.
..........
If my English is bad, I'm sorry.