Will AMD's Bulldozer plow through Intel's Sandy Bridge?

[sirmo]

From your link:
It says nothing about ClawHammer in particular and it was always the plan to release SledgeHammer first. That said, 4-6 months late is a fair estimate from the last delay. K8 was released in April 2003.

Why bring up Zacate? I've made no mention of it in this thread and I praised the low power Ontario in the Fusion thread a few posts down. The C-50 CPU is not as fast as the N550 when Ontario is scaled down to 1GHz/8W, but it's close enough. Ontario's GPU of course is a lot faster than the slow GMA 3150. I mentioned both things in that thread. If you meant my comment about Llano, it really uses a version of the K10.5 core, believe it or not. That was announced long ago and was done to mitigate risk. That puts Llano X2/X4 in Core 2 Duo/Quad territory at best, with the possibility of higher clock speeds to make up for clock for clock performance deficits. Again, really consider how late these products are and you'll start to see the problems AMD is facing.

I brought up Zacate because they only let them be benchmarked (engineering samples) a few months prior to release. You claim AMD is afraid to release Bulldozer benchmarks because they are terrible. Well there is still 4 months left until Bulldozer is supposed to be released. Zacate benchmarks weren't terrible (in fact I thought they were pretty stellar) and we only got to see them few months before release.

Llano is a weird hybrid CPU. That it uses K10.5 derived core it's a bit odd, but not late. K10.5 and whatever modifications AMD implements by the end of the year will probably be very competitive with mobile Core i3 segment. And you know that the GPU portion of Llano will smoke whatever Intel has in the TDP envelope and the price range.

By the time Llano comes out AMD will have 3 completely different architectures in pipeline: Bobcat. K10.5 and Bulldozer.

I think you are branding Bulldozer as a failure before you have any evidence to think that it will be a failure. There is more evidence to think that it will be a success. Zacate is the indication that they are on the right track. They delivered a netbook part that can game and have 10 hours battery life. They basically delivered an APU that matches Atom D525 + Ion2 performance at much less TDP and cost. I wouldn't say Zacate is late in terms of performance and I don't think Bulldozer will be either. They hit all the goals they aimed to hit with Zacate but they only showed us the benchmarks right before the release. The same is most likely the case with Bulldozer.

 

[cyclone3d]

I agree, while this is shaping up to be a good server chip, but it will only be great (and appeal the the enthusiasts) if they can IPC up.



Locking resources only when necessary isn't rocket science. I think that as the number of cores greatly increases, even the worse 'programmers' will have to think about taking threading seriously to get the desired performance.

It's not quite that simple though, there as are tasks that can't be broken down and must be executed sequentially. There's also a limit to how many cores can effectively be used before certain threads become bottle necks and all the message passing and idle waits brings everything to a halt.

But programming in such a way to eliminate needed locking as much as possible can get quite tricky.

I know some stuff has to be executed sequentially. I didn't say every program would increase linierly.

 

[pxc]

You can certainly choose to ignore AMD's own announcements and roadmaps if you want to believe something else. Peace.

 

[Dan_D]

Fair enough, I guess I am arguing from the position of somebody who is not a hardcore enthusiast



So it is a good thing we have AMD, otherwise we would still be using P4's

I never said AMD's existence wasn't a good thing. I simply refute the notion that AMD is the only reason why Intel has kept their prices so low in the last few years.

 

[ebduncan]

pxc what was the point of your post? seriously grow up and respect others opinions or just don't say anything at all. No need to flame war someone, just because they don't believe what you believe. We have jews for that.... J/k

anyways, Amd is going to release bulldozer on their own accord. People who talking like they already know what it is going to be. You do not have the cpu in your hand and a am3+ motherboard. Your getting all worked up for something that no one besides AMD employees know anything about. Personally if you want to debate which is going to be better sandy bridge or Bulldozer, wait till you have solid reviews and benchmarks of both compared to one another. If you want to discuss new features Bulldozer is going to bring to the table, so be it.

they are bringing 128bit computing, ddr3 1866, 32nm process, new instruction sets. since we already know how SB performs. Just wait till Bulldozer is out and benched and compared. Amd has no reason to hide what they plan on releasing.

On a side note i have the 1055 T hex core processor. I do video decoding and encoding all day. So i use all 6 cores, Its also a home server, which all other pcs/ tablets connect to. (4 pcs) with out hard drives. Your average PC user, will not need more than 4 cores ever. 6 core+ processors with todays applications are a niche product.

 

[eastmen]

I don't know why people are comparing prices. Intel just launched new chips , amd has just responded 16 hours ago with new chips

http://www.engadget.com/2011/01/04/...ocessor?icid=sphere_blogsmith_inpage_engadget

we will see price changes again and I'm sure the new six core will be priced the same as the current fastest 6 core and will change up opinions again

 

[ScratchDrive]

isn't someone going to compare them already? are these not out as of now? someone do some gaming benchmarks already!

 

[ebduncan]

there is nothing fancy about them, the 975 is just 100mhz faster than the 970. However it is AMD's new fastest quad core.

the 1065T is just another 6 core processor. So they have the 1045t, 1055t, 1065t, 1075t, 1090t, 1100t

 

[pxc]

isn't someone going to compare them already? are these not out as of now? someone do some gaming benchmarks already!

Rule 1:
When it has something worth showing, AMD sings it from the mountain. All comers are welcome.

The silence is not a good sign.

 

[Deleted member 174368]

Rule 1:
When it has something worth showing, AMD sings it from the mountain. All comers are welcome.

The silence is not a good sign.

You are aware people said the exact same thing right before the Radeon HD 4800, 5800, 6800 and 6900 series launched?

 

[Cat1yst]

You are aware people said the exact same thing right before the Radeon HD 4800, 5800, 6800 and 6900 series launched?

Ehhh.... i wouldnt use that last one as an example...

 

[jeremyshaw]

Ehhh.... i wouldnt use that last one as an example...

Why? Nobody even really know how many shaders there really were, and AMD spent an extra month shaving off a corner of every PCIe power connector. I mean, one entire month. That would be leak central, yet nothing solid until the day of release.

 

[Fantasyalive]

Think he was implying AMD's 6900 isn't that spectacular.

 

[cyclone3d]

The way I see it, the sooner to launch they release performance figures the better. That way, they shouldn't really hurt sales of current products.

From what I can tell and from what makes sense is that AMD is going to be able to pack a lot more performance into a very respectable power envelope due to how they are handling threading when compared to Intel's hyperthreading.

If they have nailed the memory controller to be at least as fast as Intel's LGA1366 setup they should have a really good product on their hands. And they are supporting quad channel which should hopefully handily beat the 1366 triple channel setups currently available.

Their memory controller is one of the things that has been lagging behind since the i7 first came out. Before that, AMD's memory controller spanked Intel's offering because Intel was still not using an on-die memory controller.

Back in the day when the A64 was first released, I was able to get in between 8-9GB/s on DDR1-400.

An Intel system didn't catch up to that speed until you were running close to DDR2-800 on a core2 based system.

 

[jeremyshaw]

Think he was implying AMD's 6900 isn't that spectacular.

Really? lol...

 

[pxc]

You are aware people said the exact same thing right before the Radeon HD 4800, 5800, 6800 and 6900 series launched?

Weird that people were talking about future AMD CPU launches during all those video card launches. Which of those were 2 years late?

 

[jeremyshaw]

Weird that people were talking about future AMD CPU launches during all those video card launches. Which of those were 2 years late?

Where did you figure Bulldozer was two years late? I mean, they never gave a solid launch time (then again, AMD never gave a solid launch time for the HD2900XT....)

Relatively, the HD4800 series were some 6 months late, AFAIK. The HD6800 series arrived just before the real bleeding began.

 

[Cat1yst]

Where did you figure Bulldozer was two years late? I mean, they never gave a solid launch time (then again, AMD never gave a solid launch time for the HD2900XT....)

Relatively, the HD4800 series were some 6 months late, AFAIK. The HD6800 series arrived just before the real bleeding began.

Well manufacturing processes take time to get running at decent yields and such....

 

[eastmen]

Think he was implying AMD's 6900 isn't that spectacular.

gtx 580 performance at $200 less , using less power to do with DP 1.2 support , 3,4,5 monitor gaming on a single card all while being a much smaller chip than the gf110

It may not be spectacular but its pretty damn good considering that 32nm/28nm got canceled on them. I wouldn't expect anything big from nvidia till the end of the year either.

 

[Fantasyalive]

gtx 580 performance at $200 less , using less power to do with DP 1.2 support , 3,4,5 monitor gaming on a single card all while being a much smaller chip than the gf110

It may not be spectacular but its pretty damn good considering that 32nm/28nm got canceled on them. I wouldn't expect anything big from nvidia till the end of the year either.

Well not quite gtx 580 at $200 less but yes I agree with you that it's great. I was just saying that I think that was what the guy was trying to imply with his statement.

 

[TheBlueChanell]

Initial CPUs in socket LGA 2011 will also be Sandy Bridge models later this year. In 2012, it's likely that Ivy Bridge will use the same socket.

Llano was made using a "K10.5" core, and since it lacks a L3 cache, should have performance similar to Athlon II X2 and X4 processors, possibly with minor performance improvements. Performance will not be competitive with even Core 2 Duo or Core 2 Quad clock for clock.

It's a tweak/refined and shrunk K10.5 core, and I think we will see an IPC increase with Llano when compared to Athlon II and maybe even Phenom 2.

 

[Pieter3dnow]

There's no point in arguing this with someone who's 100% illiterate in the subject.

+1 internets for creative use of "crockpot"

What part of me denying windows is multithreaded can you quote ?
The difference between multithreading on platforms like Linux and BeOS vs windows is something you cannot understand because basically it is to complex for you to understand.
Find someone on this forum that can explain it to you I won't bother.

 

[Pieter3dnow]

I never said we should stick to single core systems.

Who in this thread proposed 20 GHz machines? Delusional much?

What part of "Windows is multi-threaded" don't you get? It's been stated several times, and you've refused to refer to anything other than your ongoing crockpot of fanboi misinformation.

Please return to trolling AMDZone.

Multithreading is not a checkmark feature , it's implementation differs per platform. Maybe you should have realized when I talked about BeOS that I was talking about something you don't understand.

Some people on [H] do more with their computers then play games , maybe install linux and all that daring stuff but don't worry there will be a day that you might meet someone that can explain to you what the benefits are of an operating system that has been dedicated to multithreading and a gaming OS which supports multithreading .

 

[Michaelius]

It's a tweak/refined and shrunk K10.5 core, and I think we will see an IPC increase with Llano when compared to Athlon II and maybe even Phenom 2.

Which more or less is in line with what he said.

Phenom II IPC is at level of 65nm core 2 and slightly below 45nm Penryns.

 

[rflcptr]

What part of me denying windows is multithreaded can you quote ?
The difference between multithreading on platforms like Linux and BeOS vs windows is something you cannot understand because basically it is to complex for you to understand.
Find someone on this forum that can explain it to you I won't bother.

Multithreading is not a checkmark feature , it's implementation differs per platform. Maybe you should have realized when I talked about BeOS that I was talking about something you don't understand.

Some people on [H] do more with their computers then play games , maybe install linux and all that daring stuff but don't worry there will be a day that you might meet someone that can explain to you what the benefits are of an operating system that has been dedicated to multithreading and a gaming OS which supports multithreading .

Oh man. Instead of explaining anything, you dodge out with some ad hominem.

 

[pxc]

Which more or less is in line with what he said.

Yeah, I wasn't understanding that nit pick either. "similar performance.., possibly with minor performance improvements." It's still a K10.5 with no L3 cache, and miracles are not expected.

Llano is not going to become much more competitive clock for clock, but as I also mentioned, a speed bump may help close the gap. Still, Llano will only be competitive with Core 2 Duo and Core 2 Quad and I only posted that to dispel the fantasy speculated that Llano was going to compete with Sandy Bridge. Llano isn't much of a mystery on the CPU side. It's apparently nothing to brag about either since AMD has been oddly quiet about performance (contrast with handling of Zacate 3rd party testing and information, which is SOP when AMD has something competitive or better).

It's worth repeating: in the year 2011, AMD's mainstream CPUs are still at Athlon II X2 and Athlon II X4 performance. Doesn't anyone see the problem here? Maybe that's what's so hard to understand, how this is still happening, a nightmare to those who hoped AMD would have turned it around by now.

 

[Michaelius]

Well on positive side to Llano for people who use computers to browse net and write something in ms office Llano will offer enough performance (after all so does my notebook with 2x1,7 Ghz CULV) while being dirt cheap for oems.

And it should be more powerfull than X360/ps3 anyway hopefully bringing more people to pc gaming.

So for us on [H] it's only HTPC material but for mass market it should be fine.

 

[Dan_D]

Which more or less is in line with what he said.

Phenom II IPC is at level of 65nm core 2 and slightly below 45nm Penryns.

And given how high Intel processors are scaling now clock speed wise AMD will either have to bring the clocks on Bulldozer WAY up or they'll have to improve IPC by a large margin. Ideally they'd do able to do quite a bit of both but that seems unlikely.

 

[ebduncan]

honestly the 45nm amd parts right now scale over 4ghz.

when 32nm parts come out 4ghz+ is going to happen. I think alot of people are judging AMD's performance on its performance rather than its specs. AMD has 45NM parts, competting against Intel 32nm parts.

The fact that AMD has been competitive at price points is amazing. Bulldozer will solve a lot of problems for AMD remember ground up design, 128bit computing.

Last time we saw a bit increase was the Athlon 64, and it stomped Intel. While i don't predict it to exactly happen, there is VERY high possibility that Intel will be sitting 2nd for a little bit.

From the orginal phenom to the phenom II 65nm process to 45nm process, along with a bit bigger L3 cache, AMD picked up roughtly 15% performance boost at the same clock speeds. While just an example, if you add 15% to the current Phenom II's they will be neck and neck with the I7 parts. I think this go around with the improved IPC we are going to see around 20-25% improvement clock for clock. Which will make it very competitive to sandy bridge. While all speculation Amd will remain competitive.

 

[Nicocat]

I've been a bit out of the loop and my google-fu is weak today. Has AMD been any more forthcoming on when the Bulldozer cores are going to be released, even just as engineering samples for people to get a look at the performance? I'm currently rocking a Phenom II 940, and I'm thinking an update is soon to be in order.

Edit: Nevermind, found answers in the sticky. Leaving my comment up as a warning to anyone else who sees it =P

 

[jeremyshaw]

Q2 desktop, Q3 server, according to AMD's sever marketing guru. (JF-AMD).

 

[Rauelius]

Why all the AMD hate? I changed my mind on getting the 2600k, I checked out some reviews with games running at 1920x1080 (the rez I game at) and the performance difference between a 1100t and a 2600k isn't that huge, and in Handbrake the performance is comprable...so I guess I'll be getting an AM3+ board whenever AMD decides to release them and a Vision Black FX-8 or Vision Ultimate FX-6 CPU when they are released. Yeah, scouring the web it seems like AMD is dumping the Phenom/Athlon name and going with Vision across the platforms. It really looks like the Bulldozer vs Sandybridge will be reminiscent of the original Athlon vs Pentium III....after the k6-3, which was barely comparable to the Pentium II, no one expected the brutal blow the original Athlon did to Intel, and AMD kept improving upon it...Bulldozer will end up like that...trading blows with the Sandy Bridge.

 

[Dan_D]

Why all the AMD hate? I changed my mind on getting the 2600k, I checked out some reviews with games running at 1920x1080 (the rez I game at) and the performance difference between a 1100t and a 2600k isn't that huge, and in Handbrake the performance is comprable...so I guess I'll be getting an AM3+ board whenever AMD decides to release them and a Vision Black FX-8 or Vision Ultimate FX-6 CPU when they are released. Yeah, scouring the web it seems like AMD is dumping the Phenom/Athlon name and going with Vision across the platforms. It really looks like the Bulldozer vs Sandybridge will be reminiscent of the original Athlon vs Pentium III....after the k6-3, which was barely comparable to the Pentium II, no one expected the brutal blow the original Athlon did to Intel, and AMD kept improving upon it...Bulldozer will end up like that...trading blows with the Sandy Bridge.

I'm not sure what benchmarks you've been reading but with few exceptions the Core i7 2600K is definitely faster than the best Phenom II X6 out there. Then you need to factor in overclocking. (Unless you simply don't do that.) Once that's factored in the game changes again and the gulf between them gets larger. The Phenom can't remotely compete with a processor doing 4.8GHz on air cooling. As for Bulldozer being akin to the Athlon vs. the Pentium III, we'll see. I seriously doubt it but you never know.

 

[Rizen]

It really looks like the Bulldozer vs Sandybridge will be reminiscent of the original Athlon vs Pentium III....after the k6-3, which was barely comparable to the Pentium II, no one expected the brutal blow the original Athlon did to Intel, and AMD kept improving upon it...Bulldozer will end up like that...trading blows with the Sandy Bridge.

Considering that there are no performance statistics or comparisons leaked on Bulldozer yet, what are you basing this conclusion on?

 

[pxc]

Considering that there are no performance statistics or comparisons leaked on Bulldozer yet, what are you basing this conclusion on?

Hope.

I posted it earlier, but several people confuse the idea of AMD needing to hit a home run with what it will actually deliver. It's a bad sign that AMD is not releasing performance numbers so "close" (snicker) to launch.

 

[Dan_D]

Considering that there are no performance statistics or comparisons leaked on Bulldozer yet, what are you basing this conclusion on?

Exactly.

Hope.

I posted it earlier, but several people confuse the idea of AMD needing to hit a home run with what it will actually deliver. It's a bad sign that AMD is not releasing performance numbers so "close" (snicker) to launch.

Agreed.

 

[griff30]

Hope.
I posted it earlier, but several people confuse the idea of AMD needing to hit a home run with what it will actually deliver. It's a bad sign that AMD is not releasing performance numbers so "close" (snicker) to launch.

I'm no expert but f I had any stock left in AMD right now I would sell. The last decade has been Intel's and it's getting worse.

 

[sirmo]

Hope.

I posted it earlier, but several people confuse the idea of AMD needing to hit a home run with what it will actually deliver. It's a bad sign that AMD is not releasing performance numbers so "close" (snicker) to launch.

This is what I don't get. Usually we get numbers a few months before the launch. We are about 4 months away from the Bulldozer production (not even launch). Expecting numbers this early is surprising to me.

Again AMD only showed Zacate's performance few months before it is being released (around the time they started production). And this was on a proven 40nm fab process. Why would we expect any different for something as important as Bulldozer on a brand new 32 SOI fab process at GloFlo?

I wouldn't be drawing any conclusions just yet. It is way too early. We most likely won't see any benchmarks until April or May.

 

[Mackintire]

It's a tweak/refined and shrunk K10.5 core, and I think we will see an IPC increase with Llano when compared to Athlon II and maybe even Phenom 2.

Its a very small IPC increase, the big deal is the die shink and the fact that Llano is very inexpensive to produce.

The 2 modular (4 core) bulldozer with the 7000 series APU due out 2nd quarter of 2012 is the real lower midrange product 28nm.

Until then... AMD will try to beat Intel based on cost in the desktop market.

My best predictions, assuming no big issues appear:

AMD should grab some significant traction in the workstation and server markets.

AMD is going to dominate on the low end laptop market until Intel has a decent response.

AMD is going to fight Intel based on cost in the midrange market. Intel offers 4 cores = AMD offers 8 for the same price. When AMD's system is on loading 4 cores the clock will be higher so that the performance is comparable. But I fully expect Intel's chips to pull away during single and possibly dual thread activity.

AMD FX chips could be cherry picked wildcards

Given Intel's previous track record they will not give up the overall speed crown.

 

[Rizen]

This is what I don't get. Usually we get numbers a few months before the launch. We are about 4 months away from the Bulldozer production (not even launch). Expecting numbers this early is surprising to me.

Again AMD only showed Zacate's performance few months before it is being released (around the time they started production). And this was on a proven 40nm fab process. Why would we expect any different for something as important as Bulldozer on a brand new 32 SOI fab process at GloFlo?

I wouldn't be drawing any conclusions just yet. It is way too early. We most likely won't see any benchmarks until April or May.

It's not just showing benchmarks or performance figures, they haven't even said anything about the performance at all, except one comparison to Magny Cours processors that was extremely vague. There is literally no information at all.

Obviously you can draw your own conclusions as to what this might mean, but I would think that they would be dropping hints in order to stave off some of the Sandy Bridge purchasing if they had something worth showing off, no?