!!!The Ultimate K8L Thread 2007 & Beyond!!!

[[XC] Teroedni]

So if AMD tells the hardware press about a new CPU that they intend to launch in 2001 and issues roadmaps showing it, it doesn't count unless they put an official press release on their website and never take it down?

Face it, K8 launched years later than AMD planned and announced, official press release or not. I never said that K8L will arrive years late like the K8, it's hardly revolutionary after all. IMO K8L may arrive on schedule in Q3, it may be delayed by a few months or it might even arrive early. I don't know and I don't have a crystal ball to see the future so I'll leave those of you with them so you can get back to discussing how great K8L must be since AMD fans have been talking about it for a long time.

Well if you could dig up some from Amd themselfes. Otherwise i would say it dosent count as rumors always will float around in The Hardware community.

The point is that Amd Barcelona and Budapest is confirmed in papers made by Amd themselfes, and that make me a lot more sure that they will come out in H2. If you see on the latest roadmap i posted , you can see it will come in H2 rather than H3. When it comes to Desktop im not sure. Im only talking about Opterons, especially Budapest since it will come out as 1xxx series and work in Am2

By the way:Nothing stops you from Creating a Intel Yorkfield thread in the Intel Forum. In fact i would be very interested to read whats Intel is up to so please go ahead or you could add to this oneYorkfield Thread

[oldblue]

Theres a newer now , but its a little bit more unclear, but following the newer im pretty sure it comes out in h2 2007. So hopefully i can trust this information.

Those slides are from AMD's December 2007 Analyst Day. One of the analysts asked whether this meant we would see Barcelona in H1 instead of H2, and AMD said the target is still mid-2007.

[Serge84]

Budapest is what I'll be getting for my AM2 so I get a FX for cheap.

[GoThr3k]

So if AMD tells the hardware press about a new CPU that they intend to launch in 2001 and issues roadmaps showing it, it doesn't count unless they put an official press release on their website and never take it down?

Face it, K8 launched years later than AMD planned and announced, official press release or not. I never said that K8L will arrive years late like the K8, it's hardly revolutionary after all. IMO K8L may arrive on schedule in Q3, it may be delayed by a few months or it might even arrive early. I don't know and I don't have a crystal ball to see the future so I'll leave those of you with them so you can get back to discussing how great K8L must be since AMD fans have been talking about it for a long time.

stop whining about K8 delays, thats history,look forward instead of looking back
you just keep on whining and whining,jeeezz

[Fred_Pohl]

The only people whining about the 2 year K8 delay are Teroedni, who apparently doesn't want to admit that it happened, and now you. The only reason I even posted in this moronic fanboy speculation thread is that when I saw the ridiculous argument presented that "K8L must be a truly great CPU because people have been talking about it for a long time", I just couldn't bite my tongue hard enough to remain silent...

[ted3]

Too much emotions in this thread already, and who cares about the delay back then anyway. X2 wasnt delayed i think, so schedule is looking good for K8L.

Q: Read some stuff about the K8L architecture last autumn, Xbitlabs (i think) was worried about the L1 bandwidth, it seemed to be the same as K8. Any new info on this? Anyone got any links to newer indepth looks at K8L arcitechture?

[arisythila]

nice pic, nn you just gave me my new wallpaper

I'm eagerly waiting to see what K8L brings to the table - not like I'll be able to afford to upgrade but I still want to see it.

Its my wall paper too

[Lightman]

Unfortunately there is no new info on K8L. That is why speculation at the moment is running rampant.

But it was confirmed that L1 will at least duble bus width from dual ported 128bit to dual ported 256bit .

[Serge84]

This rant about K8's delay has to stop. >_> Grow up and find something better to talk about for once and not history. This is about the future not the past.

Budapest has 256bit wide ports in L1 L2 and L3. I known that for a long time.

[Shadowmage]

I never said that K8L will arrive years late like the K8, it's hardly revolutionary after all.

And neither is Conroe. Computer Architecture academia have been simulating wide machines for quite some time now for an 'agressive design', and Conroe is no different. There's nothing in Conroe that can be considered revolutionary. It's just about time Intel listened to reasoning instead of marketing.

That's not to say that Conroe is bad. Conroe is a very good agressive superscalar design. However, to say that K8L is not "revolutionary" is downplaying another very good aggressive superscalar design.

You want to know what IS revolutionary? Sun's Niagara is ahead of its time by a few years. AMD's Fusion and non-homogenous cores will be following the trends in computer architecture. As far as I know, Conroe and K8L are not ahead of their times at all.

[savantu]

And neither is Conroe. Computer Architecture academia have been simulating wide machines for quite some time now for an 'agressive design', and Conroe is no different. There's nothing in Conroe that can be considered revolutionary. It's just about time Intel listened to reasoning instead of marketing.

That's not to say that Conroe is bad. Conroe is a very good agressive superscalar design. However, to say that K8L is not "revolutionary" is downplaying another very good aggressive superscalar design.

What was revolutionary was Netburst.Sadly it was too ahead of tis time.Intel engineers tried some options which didn't pan out as expected.

You want to know what IS revolutionary? Sun's Niagara is ahead of its time by a few years. AMD's Fusion and non-homogenous cores will be following the trends in computer architecture. As far as I know, Conroe and K8L are not ahead of their times at all.

Really ? Since when are 8 Sparc 2 derivates slapped onto a single die revolutionary and ahead of its time ? Compared with Conroe that CPU is archaich.

[nn_step]

What was revolutionary was Netburst.Sadly it was too ahead of tis time.Intel engineers tried some options which didn't pan out as expected.

actually netburst was nothing more than a Clock speed attempt, not actually trying to improve performance or anything. And that replay system they included, was just a waste of Die space. heck if they dropped OoO for a second, they could have had a 10Ghz proc by this time last year. Heck IBM is making a 5Ghz processor before Intel and with less stages than Northwood to boot.

[Turtle 1]

Come stay on travk enough with intel stuff or this thread will get locked to.

So we can exspect the first K8L's around June than. Is that about right?

[Serge84]

If you want to talk about Intel thats what the intel forums are here for, not this.

As far as I know all they said was K8L will be around in the middle of 2007.

As for fusion, they are not for mid or low end ranges. They will not replace the GPU on a video card. Half the stuff they do won't even be emplied on graphics. Fusion isn't about replacing video cards its about expanding there power as well as the cpus power in doing tasks that are impossible for a cpu to do alone. People fail to see the big picture about fusion, think of it more like a cell processor with dif cpu cores acting as a GPU on a spisific task accept the fusion core will be 2 gpu's and 2 cpu cores on one processor.

It will not just be used to power up on-board video, it will pretty much expand on many other things allowing for games to have better physics by useing one of the fusion cores for that task as a example or anything you can think of that will help the main video card be more. And when its not in use the fusion cores will still be used for more complex tasks and allow the cpu to have no stress on it at all graphically and allow for much more things to happen at the same time. The GPU's in fusion will act like cpu's at some point switching from one thing to the next to allow greater speed the cpu cores can't do alone.

Fusion is like a attempt to make a totally new class of hybrid processor more powerful then any cpu we have today capable of doing cpu and GPU related tasks through the processor itself. Think of how much more powerful and faster our games would be with a set up like this. if you do a single sli you have 4 GPU cores, if you do a quad SLI you have 6 GPU's, and this can be upto 8GPU's on one box with 2 X2-7950's and 2 one slot cards, for qual SLI. It would be more then quad.

Fusion isn't going to be onboard video. lol Its going to be far more then just gpu's on a cpu. What makes you think its just to replace onboard graphics? That would be a waist of cores when you put a video card in, think the gpu cores won't be used? You all have fusion mixed up all together if you think its that limited. XDD

[Fred_Pohl]

And neither is Conroe. Computer Architecture academia have been simulating wide machines for quite some time now for an 'agressive design', and Conroe is no different. There's nothing in Conroe that can be considered revolutionary. It's just about time Intel listened to reasoning instead of marketing.

That's not to say that Conroe is bad. Conroe is a very good agressive superscalar design. However, to say that K8L is not "revolutionary" is downplaying another very good aggressive superscalar design.

You want to know what IS revolutionary? Sun's Niagara is ahead of its time by a few years. AMD's Fusion and non-homogenous cores will be following the trends in computer architecture. As far as I know, Conroe and K8L are not ahead of their times at all.

Who's talking about Conroe in a K8L thread? I simply stated the obvious fact that K8L is an evolutionary design rather than a revolutionary design as Netburst was. As for how good K8L will be, we really don't have any way of knowing that yet. IMO it will depend greatly on what clock speed AMD can manage because it is very unlikely that K8L will be so efficient that it can best Conroe/Penryn if it only runs at 2.5-2.9GHz.

[red]

http://www.hkepc.com/bbs/itnews.php?tid=709944
2.1-2.5GHz more likely.

[alucasa]

I have to agree with Fred_Pohl. There isn't enough information to judge K8L. There simply isn't enough information.

Of course, if you are a fanboy, you receive certain visions that we, mere mortals , never do ....

[nn_step]

ultimately the really interesting question is about AMD's implementation of Z-Ram. Since it has a known performance of 3ns on 90nm SOI, but the question is how well it will do on a 65nm process

[Fred_Pohl]

Red: If true, AMD is in a lot of trouble.

Alucasa: Try as I might, visions of the future elude me... I must be a mere mortal.

nn_Step: When was it confirmed that K8L will implement ZRAM on K8L?

[nn_step]

Red: If true, AMD is in a lot of trouble.

Alucasa: Try as I might, visions of the future elude me... I must be a mere mortal.

nn_Step: When was it confirmed that K8L will implement ZRAM on K8L?

well it hasn't been confirmed but we KNOW and have confirmed that AMD has licensed Z-Ram and then licensed the second generation as well.

[Fred_Pohl]

So AMD has licensed ZRAM and K8L might use it in 2007 or it might not. Since the first K8 benchmarks surfaced in January 2002 and the first K8s shipped in September 2003, I wonder how long it will be before we see some K8L benchmarks and how much longer before K8L ships.

[nn_step]

So AMD has licensed ZRAM and K8L might use it in 2007 or it might not. Since the first K8 benchmarks surfaced in January 2002 and the first K8s shipped in September 2003, I wonder how long it will be before we see some K8L benchmarks and how much longer before K8L ships.

Actually I am really hoping they use Z-Ram, since it competes with SRAM in Capacities greater than (4-8MB I forget which) when it comes to latency and since it is twice the density of DRAM, and five times that of SRAM. They could Have 20MB of L3 in the same space that normally only 4MB of SRAM would be.

[Shadowmage]

I have to agree with Fred_Pohl. There isn't enough information to judge K8L. There simply isn't enough information.

Of course, if you are a fanboy, you receive certain visions that we, mere mortals , never do ....

Oh jeez, you guys just aren't thinking hard enough. I was talking about computer architecture academia for a reason, you know.

For example,

B. Calder and G. Reinman, "A Comparative Survey of Load Speculation Architectures," Journal of Instruction-Level Parallelism 1, 2000.

http://citeseer.ist.psu.edu/328645.html

This paper shows some of the benefits of a variety of Load-Store Queue performance enhancements. As shown from the benchmarks, performance increases from 5-30%-ish are seen with different levels of aggressiveness. Unfortunately the benchmarks were simulated on a 16-issue processor, which would aggravate the LSQ bottleneck. Real results would be lower, but not by much probably, since a 16-issue processor would face large branch misprediction penalties. However, it's easy to see that a large performance gain can be achieved through OoO loads (I would assume AMD is doing some of the other speculative optimizations as well).

This isn't the perfect paper to use as an example, but there's loads of papers out there with tons of benchmark information. You just need to know how to read it.

[Shintai]

Actually I am really hoping they use Z-Ram, since it competes with SRAM in Capacities greater than (4-8MB I forget which) when it comes to latency and since it is twice the density of DRAM, and five times that of SRAM. They could Have 20MB of L3 in the same space that normally only 4MB of SRAM would be.

Currently ZRAM can do 400Mhz at its second generation. I think people have misunderstood ZRAMs real value. Its not on highend CPUs, but rather embedded items. ZRAM is another fanboy dreams thats blown out of proportions. Even if ZRAM is to be used in highend CPUs. Its several years and breakthroughs away. There is a really long way from 400Mhz to 3-4Ghz.

[Serge84]

http://www.hkepc.com/bbs/itnews.php?tid=709944
2.1-2.5GHz more likely.

That is only for servers. Desktops will run upto 2.9ghz at 1st then go higher later on in 2008. How many times do we have to flash this stuff around. And how do we know this won't be enough, because we have no benchmarks of the thing yet. How would AMD be in alot of trouble if 40% faster then K8 at the same clock speed eh? That wouldn't make any sense...

http://badhardware.blogspot.com/2006...y-rainier.html