K8 To Counter Conroe?

[EQTakeOffense]

Just wondering if you guys think AMD's anwser to Conroe lies in K8 or whether they will have to design a new core to beat the Conroe by a considerable amount.

I'm under the impression that AMD can modify the K8 to get equal preformace to Conroe, but to actually have a Conroe Killer, they'll need a new core.


Mod Edit: Sadly, the thread died in a fiery flaming death. ~ Papa-Ming

 

[(cf)Eclipse]

by modifying K8.. it's not K8 anymore
for now, they'll probably have to rely on higher clock speeds, then when K8L comes out, some things should change up a bit

 

[DemonDiablo]

by modifying K8.. it's not K8 anymore

thats not necessarily true. each change of the A64's socket has been a modification to the K8 yet its still the K8.

 

[xFlankerx]

Rumors have K8L as the architectural change in 2007.

 

[(cf)Eclipse]

thats not necessarily true. each change of the A64's socket has been a modification to the K8 yet its still the K8.

but the base architecture hasn't been touched, and overall performance is still basically identical to when K8 first came out on s754, except the slight boost from s939 (and soon another wee boost from AM2), and addition of sse3 support with revE

 

[Asian Dub Foundation]

conroe has K8 beat. its done and done. AMD will need a new achitecture to take back the crown. we as enthusiats hope it dosent take them as long as it took intel

 

[Met-AL]

The thing about this that matters is the price per performance. If Intel follows it's current path, the Conroe will cost a fortune. For example, a low end Conroe may be faster than a mid range Athlon, but what is the point when the low end Conroe costs 100 bucks more than the mid range Athlon?

Yaa, yaa, I know, the fans need something to be proud of and that is what the flagship CPU's are for.

 

[texuspete00]

conroe has K8 beat. its done and done. AMD will need a new achitecture to take back the crown. we as enthusiats hope it dosent take them as long as it took intel

Agreed. I'm getting ready to go Conroe. I need a case, so I'm getting a stacker so I can convert it to BTX. Lol, I haven't had an intel rig in a loooooong time.

 

[FreiDOg]

The thing about this that matters is the price per performance. If Intel follows it's current path, the Conroe will cost a fortune. For example, a low end Conroe may be faster than a mid range Athlon, but what is the point when the low end Conroe costs 100 bucks more than the mid range Athlon?

Yaa, yaa, I know, the fans need something to be proud of and that is what the flagship CPU's are for.

Conroe ranges from $210 for the 1.833/2MB to $530 for the 2.67/4MB. From today, AMD would have to cut their X2 prices by up to 1/3rd to be as cheap as Conroe. (probably a lot more to be competitive in the price/performance vein of the argument, as I don't see an AM2 4800+ (4800+ lists at ~$630 now) being anywhere close to a T6700)

I agree there's a low budget market out there, but that's the realm of Celeron/Core Solo for intel, not Conroe.


On topic
K8L is probably the first chance AMD will have to be competitive on the high performance line with Conroe. Let's hope it's enough to at least be competitive, and lets hope AMD pushes it out to the desktop early in 2007.

 

[freeloader1969]

K8L at 65nm will beat Conroe at performance and perfomance/watt.

 

[EnJoY120]

K8L at 65nm will beat Conroe at performance and perfomance/watt.

Haha, gee thanks Hector Ruiz. I'd love to know how you determined this.

 

[DemonDiablo]

Conroe ranges from $210 for the 1.833/2MB to $530 for the 2.67/4MB. From today, AMD would have to cut their X2 prices by up to 1/3rd to be as cheap as Conroe. (probably a lot more to be competitive in the price/performance vein of the argument, as I don't see an AM2 4800+ (4800+ lists at ~$630 now) being anywhere close to a T6700)

i just searced all over Intel's website looking for info on this but couldnt find anything. where did you find this info and did they have an updated questimate at when conroe will come out?

 

[FreiDOg]

i just searced all over Intel's website looking for info on this but couldnt find anything. where did you find this info and did they have an updated questimate at when conroe will come out?

Conroe pricing
The prices aren't 'official' yet, but the same prices (give or take a $1) have been sighted at several different sources.

Conroe is going to be out in July if the current roadmaps holds. (Woodcrest in June, Merom in August).

 

[DemonDiablo]

thanks. i was just over in the intel forum boards and found some stuff out over there. if conroe comes out in july that will be nice since as i was planning on upgrading in august. happy times.

 

[Asian Dub Foundation]

thanks. i was just over in the intel forum boards and found some stuff out over there. if conroe comes out in july that will be nice since as i was planning on upgrading in august. happy times.

indeed... nice to see intel (and moreover, competition) coming back to the fold

 

[freeloader1969]

Enjoy120...let's just say I know someone who's a "fly on the wall".

 

[xFlankerx]

Hard to believe that K8L, a refreshment of an architecture, can take on Conroe.

 

[BigTaf]

Enjoy120...let's just say I know someone who's a "fly on the wall".

is his name speculation?

I beleive AMD will make small improvements to the K8 core until theyre ready for a new arcitecture. They may lose the performance crown temporarily, but I think theyll bring some interesting things to the table with the new technology. I also disagree that 65nm will bring AMD the performance crown again. I beleive Intel saw a much larger benefit going to 65nm then AMD will see. Due to the high effeciency of AMD processors, and theyre ability to run relativley cool (except a few optys apparently). Lets not forget that in the move to 65nm Intel increased theyre maximum retail clockspeed by a whopping 0%.

I also dont beleive AMD has much clockspeed headroom in the X2 from the looks of the average overclock(and I repeate average oc not the 1.2ghz ocs we love to brag about) at stock voltage. So I imagine theyll do what Intel has been doing for two years prior to dual core and Conroe, adding features and trying to make small effeciency improvements.

I for one am actually dreading the Quad core cpus. I dont beleive software in its current state(atleast for home pcs) has reached the level where even two cores are taken advantage of properly. When these quad cores are released, I can forsee them being clocked at levels compareable to todays cpus, given the increased heat and power requirements(yes even at 65nm). Im theorizing that quad core will be overhyped and under utilized for years after its release. This means in software lacking SMP support or software with poor SMP support, were at the same level of performance as today. I am not an expert programmer by any means but have always had great intrest in SMP software, and software engineer's I have spoken with shares my skepticism about how well alot of software will translate to SMP operations. SMP programming is an entirly new way of thinking for many developers and I beleive it will be a rocky start. It doesnt mean it won't happen but I think its going to take some time.

Now I state this is all speculation at this point and I'm sure some will not agree. To those who don't agree, honestly I hope your right.

 

[freeloader1969]

I for one am actually dreading the Quad core cpus. I dont beleive software in its current state(atleast for home pcs) has reached the level where even two cores are taken advantage of properly. When these quad cores are released, I can forsee them being clocked at levels compareable to todays cpus, given the increased heat and power requirements(yes even at 65nm). Im theorizing that quad core will be overhyped and under utilized for years after its release. This means in software lacking SMP support or software with poor SMP support, were at the same level of performance as today. I am not an expert programmer by any means but have always had great intrest in SMP software, and software engineer's I have spoken with shares my skepticism about how well alot of software will translate to SMP operations. SMP programming is an entirly new way of thinking for many developers and I beleive it will be a rocky start. It doesnt mean it won't happen but I think its going to take some time.

Software has yet to even utilize 64bit chips to their full potential, let alone SMP. That may change with the release of Vista though (crosses fingers!). Hopefully in 2007, we'll see much more SMP aware software and 64bit applications.

 

[xonik]

iDue to the high effeciency of AMD processors, and theyre ability to run relativley cool (except a few optys apparently).

Efficiency in terms of what factor, and relative to which processor architectures?

 

[BigTaf]

Efficiency in terms of what factor, and relative to which processor architectures?

k8 and more effecient in terms of the ever famous performance/watt ratio being touted everywhere. Intel had significantly lower performance per watt when comparing 90nm cpus.
with 65nm Intel was able to decrease wattage as opposed to increasing performance to improve their performance/watt rating. AMD is already very effecient in terms of power, so I don't beleive theyll see the same benefit.

 

[visaris]

Hard to believe that K8L, a refreshment of an architecture, can take on Conroe.

Bunch of n00bs. So many people like to latch on to this "all new" thing.

You obviously don't know anything about P6 or the Conroe architecture. They just say is it "all new", when in fact, it is based on the P6 core. If you went down the list of features from P6 and Conroe, you would see that many things are amazingly similar, from the cache structure, the execution units, all the way down to the FSB.

People do not say that the Conroe is an "all new architecture" because it actually is, but because Intel told them so. So, xFlankerx, do you understand the differences enough to try and educate us on how "new" the Conroe actually is? I didn't think so. You're just another drone saying the same things everyone else is. And it originated at Intel.

K8L is based on an existing architecture the same way Core and Cornoe are based on existing designs. K8L has many new features, from double the number of floating point units to wider cache ports and larger internal buffers. Just as Conroe improves upon Core which improves upon Dothan, K8L improves upon K8 which improves upon K7.

What, is the "K" part confusing you? Just because you see the "K" you think that the K8L is really just a K7 with a mask on? What, Intel tricked you with the totally different names: Dothan, Core, Conroe? Is that it? You just can't look past the names?

You make me sick.

 

[visaris]

with 65nm Intel was able to decrease wattage as opposed to increasing performance to improve their performance/watt rating. AMD is already very effecient in terms of power, so I don't beleive theyll see the same benefit.

No, not the same benifit. Here is the way things typically work. One can always trade performace for lower power and vice versa. Intel decided to use their 65nm advantage to address their horrible power consumption issues. Because AMD doesn't have those issues, AMD will be able to spend more of the 65nm advantage on performance rather than lower power.

Further, AMD and Intel have totally different CPU manufacturing processes. Intel uses strained silicon on a bulk CMOS process. This is a bit cheaper, but produces slower, hotter chips. AMD uses strained silicon on a very advanced SOI process. This costs a little more, but produces cooler, faster chips.

One cannot look at Intel's transition to a new manufacturing node and try to extrapolate AMD's transition because the two companies use totally different processes.

Intel may have the smaller node, but in all other aspects, AMD's process it miles ahead.

 

[visaris]

Software has yet to even utilize 64bit chips to their full potential, let alone SMP. That may change with the release of Vista though (crosses fingers!). Hopefully in 2007, we'll see much more SMP aware software and 64bit applications.

As usual, a short-sighted post from an individual with a small world view.

64-bit has been around long before AMD broke x86 into the 64-bit world. SMP has been around much longer than that. There are millions of 64-bit apps all over the place. The only reason you don't see them is because you are stuck on MS Windows, which is typically miles behind the rest of the industry. Hell, even in the MS world, there are a number of apps ( from games to databases ) which take advantage of 64-bit software.

My whole, enitre operating system and all of my software runs in 64-bit mode. Every last app, from browser to media player to burning software. Also, I typically get great CPU utilization across my four cores (two Opty 265s).

If you want to claim that none of the software you use is 64-bit or SMP enabled, fine. That's not what you said though. You said, "Software has yet to even utilize 64bit chips to their full potential, let alone SMP.", which is totally wrong in almost every way. You obviously don't know what you are talking about and just want to repeat garbage you read elsewhere on the net. :shidders: You people make me sick.

 

[Jason711]

You obviously don't know anything about P6 or the Conroe architecture. They just say is it "all new", when in fact, it is based on the P6 core. If you went down the list of features from P6 and Conroe, you would see that many things are amazingly similar, from the cache structure, the execution units, all the way down to the FSB.

actually they took the best parts of the p6 and netburst and put them together (simply put). you wont see me knocking it tho.. it works.

 

[visaris]

actually they took the best parts of the p6 and netburst and put them together (simply put). you wont see me knocking it tho.. it works.

Very true. The point is not where they took the various parts from, but that it is not "all new". I really hate it when people think that since it's "new" it must be the next best thing and will take an even newer design to beat it.

Just like all the other commercials in american media. NEW hotpockets! NEW nike shoes! NEW Ford truck! NEW! NEW! NEW!

I just get so irritated when people actually fall for that BS.

 

[Ockie]

Can you guys point me in the direction as to where I can purchase this Crystal Ball that you guys are using?

 

[(cf)Eclipse]

As usual, a short-sighted post from an individual with a small world view.

64-bit has been around long before AMD broke x86 into the 64-bit world.

semantics. he meant for the desktop/consumer world and you know it

 

[FreiDOg]

You obviously don't know anything about P6 or the Conroe architecture. They just say is it "all new", when in fact, it is based on the P6 core. If you went down the list of features from P6 and Conroe, you would see that many things are amazingly similar, from the cache structure, the execution units, all the way down to the FSB.

*yawn*
Conroe is a radically different chip than even the current Pentium M Yonah chips. And it has about as much in common with Pentium Pro as K8 does.

Not that any of this matters, it could be a rip off of the 68HC11 and it would still be posting the same record breaking benchmarks.

 

[Mackintire]

First Conroe is for-real!

It is very a competitive product. In my opinion it is the first win win processor Intel has released since the PIII-S or the Celeron A. It is the proverbial cake and eat it too chip.

AMD is not sitting on their butts tho...

There is a process change coming that should allow them to increase clock speed and performance without increasing the wattage over 100 watts for the desktop if they want.

They also have a ton of momentum in the server market that they will not give up easily.


The K8L is not just a small modification is a pretty big change.

Some of the more interestingf details have already been pulled from the web.
3 additional FPU units
2 additional complex 128bit SSE2 type units
Additional prefetch and buffer architecture changes to impliment the above changes

Its kinda like going from a V8 multiport fuel injection system to a V10 Sequential port injection system.

There will be more engine and it will be more refined.

Even with all those changes the K8L should have a IPC about 20% higher then the current K8 which should (AMD hopes) put its IPC slightly higher then Conroe(s) successor.

If I remember correctly the K8L will arrive dual core and transition to quad core 6-9 months after.

It will also be able to run in the 3.4-3.8Ghz range max. Which makes is quite competitive to what Intel will be releasing.

And then there is AMD's project of "reverse SMP" Or taking a single thread and parrallizing it over multiple cores.

Good Stuff,

Flame on,

Mackintire

 

[visaris]

semantics. he meant for the desktop/consumer world and you know it

I'm a desktop user / consumer and you know it.

But, yes, you are correct. I'm a grumpy bitch much of the time. I have this motto for posting: "If one can't bother to take the time to to say things correctly, one shouldn't bother saying them at all."

If freeloader1969 wanted to say something about the desktop/consumer market, he should have said so. If he can't take the time to say what he wants to, he shouldn't post anything at all.

 

[visaris]

*yawn*
Conroe is a radically different chip than even the current Pentium M Yonah chips. And it has about as much in common with Pentium Pro as K8 does.

That is fair. Conroe is a derivitive of lots of pervious chips. It does, however, differ in many ways from all of them. My point is that Conroe is based on other chips the same way K8L will be. Just because "K8L" contains "K8", doesn't meant that the chips are all that similar. Anyone saying that K8L cannot beat the Conroe because the Conroe is "new" is spouting nothing but total BS.

Not that any of this matters, it could be a rip off of the 68HC11 and it would still be posting the same record breaking benchmarks.

Records I don't care about. SuperPI 1M Wake me up when they post some official SPEC scores.

 

[freeloader1969]

f freeloader1969 wanted to say something about the desktop/consumer market, he should have said so. If he can't take the time to say what he wants to, he shouldn't post anything at all.

Who the hell pissed in your corn flakes this morning?

My point still stands and you know it. Sure 64 bit chips have been around for years. Heck I remember using SGI workstations that had 64 bit MIPS chips in them and that was back in 1991/1992 or thereabouts. How many people and corporations use operating systems other than some variation from Microsoft? As a percentage, not many. There still isn't wide spread support for Windows XP 64. Most companies won't even port existing drivers from 32 bit to 64 bit. Doesn't that say something? I can go on and on, but I'll wait for you to tear apart this post and I'll have a good laugh. Have a good day!

 

[DemonDiablo]

Very true. The point is not where they took the various parts from, but that it is not "all new". I really hate it when people think that since it's "new" it must be the next best thing and will take an even newer design to beat it.

Just like all the other commercials in american media. NEW hotpockets! NEW nike shoes! NEW Ford truck! NEW! NEW! NEW!

I just get so irritated when people actually fall for that BS.

answer this question for me. is conroe not a new core design? does intel already have a core design EXACTLY like it already on the market?

 

[duby229]

answer this question for me. is conroe not a new core design? does intel already have a core design EXACTLY like it already on the market?

I think a better term would be "Improved" Or maybe "Refurbished" Possibly even "Redesigned"

New? No way. It is more or less a souped up PPro.

 

[DemonDiablo]

I think a better term would be "Improved" Or maybe "Refurbished" Possibly even "Redesigned"

New? No way. It is more or less a souped up PPro.

and all that would be is a play on words. what new arc hasnt borrowed a few ideas from previous core designs? yes the conroe is borrowing ideas from other core designs but that doesnt change the fact that its a new core.

 

[AaronP]

conroe has K8 beat. its done and done. AMD will need a new achitecture to take back the crown. we as enthusiats hope it dosent take them as long as it took intel

Clearly you haven't seen the Conroe Debunked threads. Conroe sucks in anything but absolute hardcore gaming, even in Encoding it barely beats a FX60. It's FPU is just plain shit, and if you happen to use more than the 4MB Cache you're performance is absolutely raped. Espcially in the Server realm, the Opterons will still own the land in the server and workstation realm, where FPU's really do matter.

 

[FreiDOg]

I think a better term would be "Improved" Or maybe "Refurbished" Possibly even "Redesigned"

New? No way. It is more or less a souped up PPro.

if you consider a PPro a souped up 486, sure. The over-arching swaths might be the same, but the implementations differ dramatically.

The appropriate (as far as I'm concerned) description is it's a compilation of what worked for Pentium M, Netburst, and the new idea's they've come up with in the last few years.


PPro had no SSE *, no data prefetch, no macro/micro op fusion, no 64 bit, no early retirment, no dual core, no L1-D sharing between cores, no dynamic memory alias prediction, minimal (if any) clock gating, I can go on for a while and not get to the parts that are just improved versions of what was on PPro.

 

[duby229]

PPro was a completely new design, based on RISC. It has nothing to do with 486. Just a front end that emulated legacy instructions. That is about it..

Conroe on the other hand is a direct decendant of PPro. Sure there are some cool things done. It is still going to be a great chip. But it certainly is not a new architecture.

I guess you can say it is a new spin on an old idea. You can also say that it has some new technology. You can even say that it is a new chip generation.

But you can't say that it is a new architecture, becouse it is not.

 

[BigTaf]

No, not the same benifit. Here is the way things typically work. One can always trade performace for lower power and vice versa. Intel decided to use their 65nm advantage to address their horrible power consumption issues. Because AMD doesn't have those issues, AMD will be able to spend more of the 65nm advantage on performance rather than lower power.

Further, AMD and Intel have totally different CPU manufacturing processes. Intel uses strained silicon on a bulk CMOS process. This is a bit cheaper, but produces slower, hotter chips. AMD uses strained silicon on a very advanced SOI process. This costs a little more, but produces cooler, faster chips.

One cannot look at Intel's transition to a new manufacturing node and try to extrapolate AMD's transition because the two companies use totally different processes.

Intel may have the smaller node, but in all other aspects, AMD's process it miles ahead.

but one could stipulate that an extremly effecient 90nm design would not see the same benefit as going to 65nm as an ineffecient 90nm core would. Intel should see a bigger difference because theyre 90nm process didnt turn out all that well and certainly didnt offer them the performance levels they were expecting (cancelled 4ghz).

I mean seriously when was the last time a die shrink provided huge gains?. Even going to 90nm in the first place maybe got the average oc up a few hundred mhz. This may seem like alot but its simply going give AMD one more high end product. If they plan to do new arcitecture mid next year, it means they are only going to be able to release maybe one or two new high end products between now and then. With Intels new arcitecture I think they may take the lead for a little while.