Conroe... are you fo real?

As Mr. Spock would say; "I find this information fascinating "

"Athlon FX 62 to have big L3 cache
What does it take to compete with Intel's next-gen architecture? Cache. Lots of cache.
--------------------------------------------------------------------------------
posted 3:42pm EST Thu Feb 16 2006 - submitted by RickGeek
BLURB


The rumor site is at it again. Not only is The Inquirer bringing up the Dell/AMD rumor again, but it's also reporting that AMD has found an answer to the upcoming Intel next-gen architecture onslaught due to hit the shelves in the second half of this year.

What is the answer? A very large L3 cache made using ZRAM (see our ZRAM coverage), a low-latency, high-speed, extremely dense cache architecture (one-fifth the space requirement of the SRAM typically used for high-speed caches). In fact, from what I've been able to read on ZRAM, it might end up being that the L3 cache has less latency than the existing L2 cache, making it an extremely fast performance enhancer. Plus, according to Charlie at The Inquirer, it could be at least 4 MB in size, meaning that even with its on-die memory controller and all of the advantages to be had there, it will still make Athlon's memory appetite more satiable using this configuration.

ZRAM only works on SOI (Silicon on Insulator), and this is a very potent advantage for AMD because it is using SOI for its manufacturing processes. The incredibly high density will allow AMD's L3 cache to take up very little additional die space, allowing the company to continue to be cost-effective.

AMD on 90 nm may be able to stave off Intel's next-gen architecture on 65 nm, including a new core design that is supposed to be the one Intel will use from this point forward as a replacement for both the former Pentium-M and now Cores Duo/Solo lines, as well as the Netburst chips. Dual-core Athlon FX-62, 2.8GHz, dual 1 MB L2 caches, dual (??) 4 MB L3 caches, or even a shared 4 MB L3 cache ... could be quite the little screamer.
To everyone who may have been thinking otherwise, Intel's upcoming next-gen architecture may not be the end of AMD--as if there were ever any doubt. :)

Read more at The Inquirer."

It should be interesting to see if this actually pans out...Discuss among yourselves.
 
would be embarassing if all amd had to do was add cache... mighty embarassing.
 
Jason711 said:
would be embarassing if all amd had to do was add cache... mighty embarassing.
Click to expand...
it'll do a lot, but not a ton, seeing as the type of cache they'd put in it is "z-ram", which isn't as fast as normal sram :p
 
Intel Just needs to impliment a 32MB L2 Cache w/ 1 Cycle Latency... Im sure they can afford to do it for us :) It would be nice.

I mean shoot, the way we are going, its not to far off.

My First computer had as much ram as conroe has cache... thats pretty nuts.
 
Isn't Intels Conroe or Laptop counterpart using a "card" so to speak for extra cache for the CPU - so technically Intel is already added more cache.. this zram sounds familar,
 
MrGuvernment said:
Isn't Intels Conroe or Laptop counterpart using a "card" so to speak for extra cache for the CPU - so technically Intel is already added more cache.. this zram sounds familar,
Click to expand...
No, it's on-die SRAM. Intel hasn't used off-die cache since the P2/P3 era.
 
SLee said:
No, it's on-die SRAM. Intel hasn't used off-die cache since the P2/P3 era.
Click to expand...
Well you are correct, they are going to be using Flash RAM as a Buffer between the CPU and RAM. This has already been stated. It will be like another level of cache. More or less.

As far as I know, this will only be for Laptops? however it may come to the desktop. Its a good way to beef up another cache level. We have large 4MB L2 caches now. On Die... full speed.... but we have the abilitly of having a 128MB L3 etc off die.... this is how it has always worked until years down the road the tech is there to integrate to the Die itself. Then we will have 128MB L2 On die.... and the cycle will repeat.
 
USMC2Hard4U said:
Well you are correct, they are going to be using Flash RAM as a Buffer between the CPU and RAM. This has already been stated. It will be like another level of cache. More or less.

As far as I know, this will only be for Laptops? however it may come to the desktop. Its a good way to beef up another cache level. We have large 4MB L2 caches now. On Die... full speed.... but we have the abilitly of having a 128MB L3 etc off die.... this is how it has always worked until years down the road the tech is there to integrate to the Die itself. Then we will have 128MB L2 On die.... and the cycle will repeat.
Click to expand...

I sure as hell hope not. Flash RAM is incredibly slow compared to dram/sram/zram.
 
L3 Zram, sounds like a plan, but donno if that alone will make up for 20% of core performance
 
I think you'll find that when the true enthusiasts start testing Conroe and the FX-60/62 the performance will be closer than the 20% we were seeing. Part of the reason is the way in which the Intel benchers oc'd the FX-60. Instead of adjusting the HT cpu clock (fsb - whatever you want to call it) they adjusted the cpu multiplier instead. While this will provide a higher clock (snce the cpu multi is unlocked upwards) it is less efficient than leaving the cpu multiplier at the stock setting and increasing the cpu clock frequency. I'm not saying this would account for the 20% difference, all I'm saying is it wasn't an optimal test for the AMD cpu. Would the oc'd FX-60 still have lost? By all accounts indeed it would have, but it may have been a little closer. Notice I said "may have been."
Plus we don't really know what the true capabilities of the FX-62 will be. We can only guestimate based on what we know about the FX-60.
Right now all we have is one bench run by Intel and then Anand showing a very pre-production Intel mobo, chipset and cpu. While it looks extremely promising for Intel, once we get production systems in the hands of the enthusiast base we'll have a better idea of who is faster, by how much, and why. The Intel may be even stronger than the preview showed since they will have time to optimize and tweak the setup. We won't know until Conroe is released.
Really, we are wasting a lot of breath, emotion and vitriole on something that hasn't even debuted yet, Intel fans are finally getting something to cheer about and that's a good thing. AMD has been on top and its fans are now taking a bit of a beating. Its all good. I'm not going to lose any sleep over it and neither should any of you, unless of course you have a controlling interest in either company. ;)
 
^^ then i cant wait when people get them in their hands and use properly timed ram settings, nice and tight and not loose as they did on the intel test which was a downfall for intel and a certain loss of more performance.
 
there has to be a reason intel wont just release samples to get properly "previewed".

if its that damn fast (and i do believe it is mighty fast) then whats the deal? it would only be more publicity. :confused:
 
Well, a preview chip in the wrong hands could do more harm than good. They played it smart - give everyone a taste and get the buzz going. Intel has always been market savvy. I would certainly expect Conroe to be a good running system when it finally debuts.
 
^^^ exactly

If someone from a fanny boi site got a hold of it they would botch the results - now, we all know over time proper reviews would come out, but of course your going to have those people who will always beleive the false and fake benchmarks to hold onto their hope.
 
MrGuvernment said:
^^^ exactly

If someone from a fanny boi site got a hold of it they would botch the results - now, we all know over time proper reviews would come out, but of course your going to have those people who will always beleive the false and fake benchmarks to hold onto their hope.
Click to expand...

I hope you'll have a really bad day when Conroe ships ...
 
savantu said:
I hope you'll have a really bad day when Conroe ships ...
Click to expand...

Why would I?

I am not buying a new system anytime soon, I hope the conroe kicks some butt, and even "if" it does, and it is proven time and time again you are still going to always have your people in denial that AMD is better, just as even to this day people think a P4 or a 5/6/8** series is better for gaming over AMD...

I then hope the M2 socket offers something to compete with conroe, either to drive down Intel prices more, or to get AMD on the road to the top again , "if" conroe knocks out AMD.

i have no brand loyalty, as many here, I get what does the tasks i need to get done, fastest.
 
BigMacAttack said:
AMD is already working on quad core (and above) cpu's, plus the AMD A64 architecture was laid out with multi-cores in mind from the very beginning. From an engineering standpoint I've read that AMD's architecture is expandable to the multi-core environment much more efficiently that Intel's architecture. To go to quad core will take another entire generation of chips on Intel's part while AMD won't have to radically redesign. Anyone who thinks Intel is going to run away and hide from AMD better think again. 45nm (and smaller) architecture is in AMD's road map by 2010. Intel may have a theoretical lead now (we'll see how real world Conroe results pan out) due to the 65nm architure vs. AMD's present 90nm architecture but it won't last long. AMD may not get get to quad core first (this is debatable) but their design will be more economical due to their architectural superiority.
I feel sorry for the software designers. They can't keep up as it is. :rolleyes:
Click to expand...

I am a little confused or maybe you are. You are aware the P6 pro cores were designed with multi cores in mind were you not. That is why the Israeli team choose that core to do Donthan . Befor dothan was even started the Israeli team was planing on the Yonah dual core. Your statement About AMD having superior tech is laughable at best.

Amd has ondie memory controller that wasn't even their idea. DDR2 on the AMD processor isn't looking to good.

As for you guys that haven't researched Merom there is noway you should even talk about Penryn.

AMD has superior tech. LOL LOL LOL.
 
USMC2Hard4U said:
Well you are correct, they are going to be using Flash RAM as a Buffer between the CPU and RAM. This has already been stated. It will be like another level of cache. More or less.
Click to expand...
I think you may be confusing the idea of having a flash based cache between RAM and the HDD with what you wrote. Flash is very slow compared to most solid-state volatile storage (SRAM, DRAM...), however it has some advantages when it comes to random access over a HDD.
 
drizzt81 said:
I think you may be confusing the idea of having a flash based cache between RAM and the HDD with what you wrote. Flash is very slow compared to most solid-state volatile storage (SRAM, DRAM...), however it has some advantages when it comes to random access over a HDD.
Click to expand...
Yeah, short random accesses on a hard drive are relatively slow. An 8ms average seek (or even 1 or 2ms track to track seek) is slow if you only need to access a few clusters. Fast flash memory could perform comparably. But I think writes to flash won't be the normal operating mode for the drive.
 
MrGuvernment said:
...

i have no brand loyalty, as many here, I get what does the tasks i need to get done, fastest.
Click to expand...

Yeah , you're not having brand loyalty , but fanatical loyalty to AMD , blindly refusing to accept that Chipzilla will release something better , cooler ancheaper than equivalent AMD solution.

By having a blind faith that somehow AMD will pull wonders by simply moving to another socket you clearly are over the loyalty thing to something more dangerous , reality denial....
 
The Z-RAM license deal was made 2 months ago. It's not going to show up in the FX-62. Development and validation of a new core is not immediate. Z-RAM will not show up AMD CPUs for a while, even by the end of 2007 is extremely optimistic. If there is any truth the rumor of a L3 cache in the FX-62, that decision would have been made well over a year ago.
----

Both AMD and Intel have put their cards on the table in roadmaps. AMD hasn't exactly stood still. They have done some very good work on desktop CPU power consumption (35W-65W for mainstream processors, including dual core) coming in June and mobile Turions are very power competitive already. The socket change isn't going to do much for performance on desktops. And AMD has made their own misstep in the successor to the K8 (how many have been cancelled, 1, 2 or 3?), forcing the K8 and K8L to fill in for the meantime.

Conroe and Merom performance gains are not a recent surprise. Hints of it have been trickling out for almost a year (and a 65W desktop TDP for over 2 years). If Conroe/Merom dominates performance vs the K8, don't act angry or surprised. It's a competitive market and it's bound to swing in favor of either company. Both have competent engineers/designers.

It's very possible for AMD to come back with their next generation core, but that's at least a year and a half away. Some people will wait but I upgrade far more often than that. I really don't care if I buy Intel or AMD. Bang for the buck is most important and from the 2H06 pricelists I've seen, the Conroe 2.4GHz/4MB looks like it will fill that requirement.
 
I can honestly say If I had a FX60 I would wait If I wanted AMD . Sense I don't buy gpu's on every release I sure as hell wouldn't jump on every cpu release. FX60 is a good cpu and should be strong performer for at least 2 more years.

Some people have to have the very fastest to fullfill their e-penis fantasies. My old PC with 3.2c Intel is serving my wife very well. But she wants the lastest greatest upcoming release . I see nothing wrong with that 3years is a long time. My dothan is another story and I really don't like replacing this that soon, but I will run that untill Vista is released and than upgrade.
 
It a cycle here people.
AMD Athlon took the performance crown where the was concerned Pentium 3, then came to trump the Athlon with the Pentium 4. AMD decided enough is enough and then the Athlon owned all. Now we have intel here to regard the thrown.
I have no worries thanks to the pressure one company is putting on the other to make great chips I see this cycle continuing.
 
savantu said:
Yeah , you're not having brand loyalty , but fanatical loyalty to AMD , blindly refusing to accept that Chipzilla will release something better , cooler ancheaper than equivalent AMD solution.

By having a blind faith that somehow AMD will pull wonders by simply moving to another socket you clearly are over the loyalty thing to something more dangerous , reality denial....
Click to expand...


lol - um, i think you got me wrong.... i never said AMD will pull wonders - i said it would be nice if they did, iam sorry if you miss-read things, or if i miss-worded them. My statement that you quoted shows that - i buy what is faster for my task - how is that fanatical loyality to any single brand ?

I have a preference for Intel, always have - i use all laptops now usually, my last $2k gaming rig was using a prescott 3.2ghz, my system before that was a northy 3ghz, the ONLY AMD system i have are work systems i have put together and the fastest one was a xp2600+ cause it was cheaper. I have never used and AMD system about a 3800+ which my uncle owned and couldnt get to work with a DFI board so i tried to fix it - i dont like AMD for the many complicated issues with just getting them to work.


EDIT: Reading over my posts i see NO indication i am an AMD fanny, where did you get that idea?
 
MrGuvernment said:
lol - um, i think you got me wrong.... i never said ..


SNIP!!!


EDIT: Reading over my posts i see NO indication i am an AMD fanny, where did you get that idea?
Click to expand...



He's an Intel fanny, therefore anyone who even says the letters is a fanny ;).
 
You must log in or register to reply here.
Back
Top