Indeed, eveyone shouting 4 module BD is not 8 cores just says the exact opposite of what that head markteing guy of the server section is saying. :ROTF:
Printable View
And I agree since Zambezi has 8 cores. They are just organized in the way that a basic building block is actually a very optimized dual core(a module).
As for "next year intel trouncing" and "3d transistor revolution",how about we wait and see what both companies have 1 year from now and what kind of performance and perf./watt they both get out of those products.
with AMD using GF now, who would we expect to hear such technology come from? my first assumption is that its going to be GF since they have the fabs and just offer such things for AMD to use.
Informal,
I am a consumer that likes computer technology. Thatīs about it.
I think Zambezi is an 8 core CPU but something comes to mind and, if possible, Iīd like to know how both Intelīs Sandy Bringe (2600k) and AMDīs Bulldozer (Zambezi) would behave if they have to handlle more than four 256bits instructions at the same time (is it even possible?).
Edit: 4 x 256bit floating point instructions
Thanks.
No, BD has 4 modules each consisting of 2 int cores and 1 256bit fp core and a shared frontend. There's no logical cores in BD although it's not really a full 8 core CPU either. Tbh describing BD with the traditional 1 int core = 1 cpu core is not really accurate but on the other hand using a different nomenclature without confusing the avg consumer is also impossible.
I thought I'd pop this up here to encourage even more stimulating debate :D
http://i716.photobucket.com/albums/w.../bulldozer.jpg
My understanding is the L2 is not shared between modules, but the L3 is shared between modules.
I'm not really the guy to answer this but my understanding is even though AVX provides 256-bit executions a single-precision FP command is 32-bit and a double-precision FP command is 64-bit.
So, essentially, the module above could execute 4 64-bit commands per cycle or 8 32-bit commands per cycle.
Here is the JF-AMD 'blog' on AMD Flex FP that explains it better ...
Where is it? I'm sick of this here it comes debate. I want it now. For god sake. They are good at GCard's why not CPU's. They beat NVIDIA to releases why not Intel. Intel are starting to scare me.
Pimp my CPU lawl
same as any other operand. If it cannot be sheduled at the same cycle it will be sheduled the next cycle.
So if cpu 1 needs 6cycles for one AVX instruction it will need 7cycles to do 2 AVX ops
Also with FMAC BD can do 8. SB can do 12 if they are different types but that rate is not sustainable.
Guess that means I should put an LCD monitor inside my case door, and swap out all my Noctua for LED fans :P
Exactly, at 32nm now AMD finally caught up on the design side but they didnt have a fab to actually produce the design into physical chips until now...:rofl: and then Intel will move on to 22nm :rolleyes:
I mean, Core 2 at 45nm was kind of the step AMD took with Phenom II...in all reality AMD was trying to market essentially a competitor to Core 2 Quad against Core i7 @ 45nm...some how they did it successfully...no doubt, they are fun chips to play with and are great for most people, but they were behind the competition as far as a pure performance standpoint. I think Bulldozer will be AMD's "Core i7" competitor, but can they compete with Sandy Bridge? If you look at the past, 2006 to now, you could almost safely say the chips will be at Nelahem/Westmere level. Sandy Bridge, as much as I love AMD I believe will be a challenge...Sandy-E even more.
Yes there are 4 256bit FP units.But why is this important? AMD also supports 256bit FMA IIRC,so if both chips are running software which supports all of their capabilities(FMA and AVX) then benchmarks will show which approach will return better performance. It's performance and perf./watt what matters.
The throughput of a core does not define the number of cores.
Core is an abstract concept. No core is equal so it cannot be used as a measuring stick.
Also it would be very simplistic because a single core (1Thread) will have mispredicts, bubbles, dependant operands (have to wait on others) so not all the resources will be used. This is the concept where HT tries to improve, they use leftovers to execute another thread. So having 2threads over a certain unit will most likely give a higher throughput than 1thread would.
That is the wrong way to think about it. You have integer execution and FP execution.
90% of what happens in processing is integer. Especially in games.
Where, exactly, are you looking to use these 256-bit instructions? Software needs to take advantage of them and it needs to be rewritten. If they can't take advantage then they will not recompile.
Those applications that will be recompiled will probably take advantage of the FMA4 instructions as well, which will present an even more efficient way to handle complex math.
Suffice to say that if you need 256-bit, then you can bet your apps will be there, but without the need, you'll be in 128-bit mode. And in 128-bit mode, we have double the capability in Bulldozer.
Hi JF.
Like I said before, Iīm just a curious costumer that likes computer technology.
I think AMDīs approach with Bulldozer is really nice.
Talking like the average Joe that I am, what I need is to be able to do a lot of stuff at the same time (listen to music, work on a document, browse the web,...). I donīt need to do one thing super fast (i mean, half a second is irrelevant).
I know you are a server marketing guy, but just to say... If all goes well, Iīm buying Llano.
offtopic an redundant:
What we have to understand is that, at least where I live, at the client side the majority of costumers (95%) donīt buy nanometers, power consumption or noise. It all comes down to price and brand.
At the same price and even if a little more expensive, people will go Intel.
You have a good product... Advertise it!
* Sorry for the bad english. Hope you can undurstand me.
Again, it's not a logical vs physical debate. If your workload is pure integer, you can execute 8 threads on 8 real integer cores at the same time. If your workload is pure 128-bit floating point, you can again execute 8 threads at the same time. The FPU is a little more akin to HT in this instance in that there's one unit that is doing multiple threads, but ultimately it's more like two real cores. It splits in half and executes each thread independently. If your workload is pure 256-bit floating point, you can only execute 4 threads at the same time. This is not like HT at all. To make it even more interesting, a given cycle could have the CPU look like five, six, or seven "cores" depending on the mix of instructions. Using the "core" nomenclature for something that isn't built using that traditional model isn't very precise, but if you want to call it anything you should go by the integer cores since the vast majority of arithmetic in computer programs is integer. Eight real "cores".
You can't run 256-bit FP on a 980X, right?
So those aren't real cores, right?
It was just a question for my personal knowledge (like I said before, Iīm just a curious consumer).
And I believe Zambezi is an 8 core CPU.
But then you (AMD guys) should not let people call a module "a core and a half" just in front of you.
http://www.youtube.com/watch?v=wxbG2AmdMNY (2:05)
Believe me. I like AMD.
So for clarification a core refers to a set of two integer units and one floating point unit? So then an octocore bulldozer would have 16 integer units?
I would suggest to have a look at that nice picture of BD module in that video ;)
take that 'core and a half' expression and analyse it with the help of that picture or what we know about the BD module.
yes they are saying that it is core(integer) and a half(shared FPU). But that half(shared FPU) can be split into two FULL(compared to current Phenom design) FPUs(cores). If needed BD module can be one core and really-really-powerful-half(256bit FPU) or that module can be two regular cores like we are used to have in todays design(integer+128bit FPU). That FPU can be split into two units, thus Flex FPU. So you get the best of both worlds.
You guys need to stop looking at BD arch in comparison to current design, as you will never come to the apples to apples comparison :)
You can say that if FPU is split then and we have two conventional cores per module and nothing has changed compared to current(phenom design), then again it would be not true as AMD spent like half a decade designing and building BD, thus most of the stuff is beefed up and upgraded and quite a few of the arch bottlenecks removed and thought out.
Edit: OT, damn, what happened to new comment notifications :/ It is like reading a new thread, even though I had it subscribed long time ago :D
Leaked performance, pleaseeeeeeeeee .............
IMO, people who are concerned with speed per core, or how many.. don't understand API's, or their own OS. That whatever they want to do with their system (game, watch movie, etc) has no bearing whether it's single-threaded, as many things are vying for resources.
It almost always never matters how fast ur system is, only how slow it is not.
Coincidentally, if you are concerned^, then shop tdp as any modern cpu will OC itself for max thruput when needed. And again, if u generally don't use a computer, but specifically do, then ur specific needs don't apply to the populace, therefore have little bearing as "end-user" needs. As those needs would be more scientific, or business.
So what if this is called xtremesystems, do you have to have an Intel and an OC of more than 4ghz to access the forums? Freakin' elitist attitude that's no different than the apple fanatics. It's what you do with your hardware, rather than what your hardware is.
Right because no one overclocks on this forum.also when did overclocking ever get brought up? If you read what I first quoted you can see why xtremesystems was brought up.
Typical response from an amd user though.the difference is that I will buy a bf if it performs better than my setup now.
Are we not allowed to compare sandy to bd? I would be saying the same thing even if I had a amd rig.its funny that we can't bring up intels performance.
Last I looked we are not in the amd section.its fair game to compare bd to sandy bridge.
The 22nn 6ghz was brought up from the post I quoted saying that we don't know our os and or what multi cores do or how they perform.
The overclock was mentioned for sake of knowing what a bd core is about and wanting to push intels next shrink to the limits.I should of put I can't wait to overclock amds 22 nm core to 6ghz this is not an apple forum.
Where did I complain against comparison against Intel? Then again your post suggests we're not "xtreme" if we're not using Intel. I bought my X6 for 175USD at launch so yeah, a typical AMD user is what again? I'm reacting to your elitist attitude regarding to what "extreme" really is. I can build a celeron at stock for whatever use and have an insane case mod, and that would still be extreme.
And like most sane person, I will wait for performance reviews before choosing anything, but I won't sneer at people who choose differently.
edit:
except if they choose a mac just because it's pretty, I'd make fun of that
My god some of this is hard to read.
Hi muziqaz,
In a limited way, I understand Bulldozerīs concept. My doubts were in execution.
I already saw those pictures of Bulldozer (a lot of times). That was not the question. My concern was with people that does not come to forums and listen to a marketing video, where some guy call a module "a core and a half". I think, at that moment, the AMD representant should clarify and say.. "No, a module is two real cores......"
Pretty sure theyre gonna use cores.From a marketing standpoint its better to have 8 than 4 :P.
Im not sold on this however, but that is a debate, what actually IS a core.Anyhow, it took good few years before laymans started to understand what cores are, i dont think AMD is going into explaining what a module is ,mess.
madcho, they will use cores only. Modules are just for presentation of a concept only ;) End user will know only cores.
memmem, but John(JF-AMD) said numerous times that BD module is two cores. Numerous. Even then some intel shills would come and say oh it is not real cores, etc and etc. But John would repeat over and over again, that BD has real cores.
Those 1 and a half cores in that video was just a figure of speech of some sorts, again for presentational purposes.
XRL8, I think a lot of people cannot get used to AMD changing CPU design so radically. Now CPU core has integer unit and FP unit. BD can have that conventional core if needed, but also it can be something a bit different :)
So close to launch and no performance numbers yet, hmmm......
you guys must have some psycho capabilities of guessing the future i'm not aware of :P
Are you that depressive?
i only read "fail fail fail" ... you almost start short circuiting ... :P It's a freaking cpu that no one knows about ... yeah sure... you must be frustrated with past launches ... but just hold your horses a little bit ...
Jeeeezz ... the hormonal levels must be hitting the limits :D
But that is AMD of todays. Look at the couple of recent GPU launches. They were very silent about those. GPUs came out fine.
... AMD CPU's as nothing to do with AMD GPU's ... two different teams, different marketing approaches ... and ... the 69xx series came late... they were supposed to "battle" with the 4xx series ... but Nvidia did such a good job on the 4xx series that they had to launch the 5xx series :-)
We can't compare ... even if they have the same company behind :)
How is it not the same? They all are under AMD. AMD decides to be tight lipped about their products, so that will include GPUs, CPUs, APUs. Being silent does not mean they have :banana::banana::banana::banana: products for us. and it does not mean it has something great for us. It is just some of you guys are not very patient and start spreading FUD.
AKM... yeah i saw it :) of course i was not referring to a single post ... in this case, yours :)
As for AMD now owning ATI ... of course there's technological sharing between the GPU R&D and CPU R&D.
What i mean is that we can't make a direct comparison between CPU & GPU marketing and business model.
Although they have the same "company vessel, AMD" they have different teams taking care of business development.
The 69xx were neither late or meant to battle GTX 4xx. It was designed to be AMD's next gen videocard on a 32nm process. But with TSMC canceling that and going straight to 28nm(which won't be up for a while still) AMD decided to take as much from that 32nm design as they could and cram it into a 40nm chip. The 69xx is AMD making the best out of a bad potion they were put into by something completely out of their control.
i'm pretty sure they will be.
but from what i've heard, transitioning from one foundry's process to another is difficult when you care about performance(leakage, etc.). :banana::banana::banana::banana: just takes time. there may or may not be contract issues involved as well.
The bandwidth gained from the VLSI is where fusion will gain some if it's prominance. PCI-e is like 64 Gbit/s (16GB/s), & HT is like 330 Gbit/s (41GB/s), no? So, AMD has alot of advantages to work with when you consider the integration (cache sharing, faster main memory access for GPU, etc). Obviously, AMD gets alot of their performance & agility (Llano vid) with the tight intigration in the APU.
Secondly, nobody notices how fast a system is, just how slow it is. Fusion will eliminate that, thus setting a much higher minimum.
The choice evenutally comes down to: Is your software slowing you down (lag, hitching, pauses, etc), or is the output not enough? (Crunching numbers, deep mining numbers in shorter time, or magnitude more in same time, Pi, benchmark, etc)
Thus, determining what path to take.
I am sure Bulldozer will give a good glimpse at whats achievable @ smaller size, next year. Such, as a 1080p computer, that can handle BF3 in DX11 for $500 ? (& yes, if u stop and think about it, that is comming next year for xmas, no video card required!)
So, no matter how slow AMD's BDs are, it still won't matter. They won't be slow enough to notice.
btw, I don't believe in fanboism, onlic logic & reasoning.
thing i'm most excited about is I do believe in a few years.. (say next gen consoles) The process of adding GPU to processor will be much more refined.. and will make PC's able to compete with consoles on a price level.
I really hope it does happen cos' for the average gamer its about price :)
you sure its not just about 'living space'? most people i know aren't comfortable with the idea of having their pc hooked up to their tv.
it is difficult, especially after adding so many custom circuits for power saving. those will have to be completely redone, along with some other things. it is not easy to do and takes a lot of time.
also, when GF split from AMD they agreed that the gpu division's fab needs would transition to GF over time so they will be using global foundries.
both tsmc and gf have absurdly high wafer output for 28nm. i dont see that being a major issue.
Until the New York Fab 8 is up and running in 2012, GloFo is limited to the Dresden for 32nm and it will be pushed to the max making Llano and Bulldozer(which are much more important to the long term viability of AMD than any GPU). When Fab 8 is up and running 28nm, we'll see AMD at least considering switching to GloFo for GPUs, but even then it'll be a drawn out process. Making high-end GPUs is something GloFo has no experience with, so there will be growing pains that mean it'll take that much longer.
If GF were actually being pushed to the limit by AMD CPUs alone they wouldn't have spun the company off in the first place....
I agree with pretty much all you have to say except one thing. You are confusing Bulldozer with Llano. ;) You are talking about Llano, but you say Bulldozer. Interesting point about people not noticing performance, only lack thereof.
He was talking about using Glofos 32nm process for GPUs. Glofos 32nm capabilities is limited. And they aren't producing much 28nm yet.
Hehe, whats real and whats not lol.
Its all getting really confusing.
So 8 core cpu would be just below Nehalem clock-per-clock, and clock higher?
Not too bad at all.
I assume poor multicore speed-up is because when only one core per module is used it has more resources?
ahh what?
Those results are WAY off lol.
They suggest a 6 Core Phenom II is 45-60%% faster clock/clock :lol:
Phenom II x6:-
Fritz Chess @ 3.0Ghz= 21.4
Cine 11.5 64b @ 3.0Ghz= 5.07
More time wasting screens unfortunatly.
It's a K44 sample, so B0 stepping. Anything below B1 isn't worth bothering with.
One thing we know for sure, the early eng. samples are crap.
RussC
Yeah ... it's like that Ivy Bridge sample selling on Ebay ... 1.8Ghz ... :P
the dude that sells an ivy bridge now is gonna be f***ed i suppose
I've been loosely following this thread for updates and kicks and giggles but with no solid conclusions since I haven't closely followed it. Does that pretty much sum up this thread thus far?
So if I get this right... we're not even sure if BD will compete with BD clock per clock in IPC? Or..I should say it doesn't look very promising? Man if that is the case then I will be disappointed as I've been anticipating launch.
Given the secrets, wait for the E3 in early June.
It makes sense of course that the older samples are the ones that find their way into the hands of those less concerned with abiding by their NDA. I'm sure there are better samples out there already.
Yes the entire thread is about the solid proof we have of how horrible BD is, in fact it is slower than current Phenom CPUs and it will be a horrible failure. Make sure to repeat this as fact as often as possible, you wouldn't want to waste all of your valuable research.
Is the Release aim Q2 or Q3?
But if its a big release in June then they need to start shipment pretty soon and then its pretty odd there is no benchmarks out yet? Shipping out that many CPUs is not done in one 1-2 weeks I would think. Normally it seems like shops have the products some days before release as well, so 3 weeks from launch I would guess it would need to ship from the foundrys?
Exclusive AMD Bulldozer Details From Gigabyte
http://www.rumorpedia.net/exclusive-...from-gigabyte/Quote:
Today is a good day as we have just received a great picture from Kirllos which was acquired by mega-d botnet.
Where is it from? According to email, it was borrowed from Gigabyte’s corporate network.
Machine is running AMD FX 8110 processor with the maximum TDP of 95W @ 2.8 Ghz and Socket AMD3b 942 package. If you are wondering why it says 3.8 GHz as well, it’s because o the AMD turbo boost.
http://www.rumorpedia.net/wp-content...dbulldozer.jpg
Is it just me or it totally owns in Wprime?? :D
Never mind superpi result,it rocks tehre too but it's irrelevant one.
How does this compare to i7 2600k, for example? How high would it have to be clocked to get similar results?
Is the voltage info legit?