More on K8L models and announced speeds

[Serge84]

http://badhardware.blogspot.com/2006/10/amds-k8l-revealed-in-cray-rainier.html

I would get the Antares core at 2ghz 35w and just oc it to whatever speed. I wonder when the 3.5ghz cores will come out at the max HT3 speed of 2.6ghz.

[FghtinIrshNvrDi]

I'll be sittin on my haunches till AM3 is debuted. I'll save up tons of cash next summer waitin for K8L to get closer.

Interesting thing they're doing with HT 3.0. Why are they all different clock speeds, unlike the A64 line? Seems like they're using a multiplier as they would for the core speeds.

Ryan

[erwinz]

i just read in some recent post that it uses.. 0.5x multis? I cant really be sure? hehehe :hehe:

[Serge84]

i just read in some recent post that it uses.. 0.5x multis? I cant really be sure? hehehe :hehe:

That is true even for rev G. Saw it from AMD slide shows from some people on here as well. :fact:

[Serge84]

I'll be sittin on my haunches till AM3 is debuted. I'll save up tons of cash next summer waitin for K8L to get closer.

Interesting thing they're doing with HT 3.0. Why are they all different clock speeds, unlike the A64 line? Seems like they're using a multiplier as they would for the core speeds.

Ryan

The HT 3 speed is prob linked with the core speed of the cpu so they can make models cheaper or more expencive depending on what HT3 speed and cpu speed you want per model. Now since clocks will be made between 100mhz speeds, 4 core vers, 3 cache vers, and HT3 speed vers. We will have more models to choose from witch means cheaper single and dual cores because of quad cores. And best part theres not just going to be afew models but a lot of new series of them. This is really good news.

[bestmancajun]

oooooo a 40% performance rise, almost sounds too good to be true.

[nn_step]

oooooo a 40% performance rise, almost sounds too good to be true.
meh sounds like the low end if you ask me

[Serge84]

As is the advancement of technology and progress every 2 years cpu's would show double the advancements. Just like Intel upgraded, now its AMD's turn. 40% over K8's sounds true because it has conroe like processes so I don't see whats so fake about it. Its on every AMD data sheet that shows the new arc.

But I'm just more opened minded to things. We'll see what happens when it comes out. Even if its 20% faster at the same clock speed, we will still have 3.5ghz stock speed models after all if we want to see full speed HT 3.0. It will atleast match conroe. But again it wasn't desined to go up agenst that cpu anyways. Its desined to go up agenst Duo3 arcs.

[ozzimark]

oooooo a 40% performance rise, almost sounds too good to be true.
heh, let's not forget the % increase in performance intel got from netburst -> core at the same clocks :eek: :eek: (then again, netburst was all about speed, not high performance per clock :p: )

[cky2k6]

umm... did everyone miss the fact that altair would be going against yorkfield at 3.4-3.7ghz? if it gains 15% clock for clock over k8, it basically matches conroe's performance clock for clock. yorkfield is definately not any slower than conroe... so amd either better release altair at 3.5ghz or we better hope that its more than 15% improvement clock for clock, which i find doubtfull since amd is not really adding anything that conroe doesnt already have. if amd's torrenza doesnt pan out, then amd mght loose the majority of the enthusiast market for good, whatever it means to mr. hector opteron ruiz.

[nn_step]

once again I have to come out and say Opteron is not about the enthusiast market, it is about Servers and the lucrative profit margins created by the sale of servers. AMD WILL put a smack down on Conroe and yorkfield with K8L, simply because it protects their Server market. And the reason they are SOO busy pimping HTT and Torrenza is because it protects them and the rest of the market from Intel. It gives them the performance edge in EVERY SINGLE benchmark (that matters), It provides them with chipset sales EVEN if their processor isn't used. It makes it a FAR safer and profitable reason for Chipset makers to make Chips that support AMD's sockets. Also being an Open socket, it makes it cheaper for other Processor manufacters to make processors, since now they don't even need to license a bus spec or develope their own. They can just use AMD's which is proven to be of high quality and VERY well designed. They are TOO protected by this useful TOOL. Because THEY KNOW they will have cheap high quality motherboards that COMPLETELY support their new processor design. Weather their design enhances AMD's product or someone else enhances their own. It is a WONDERFUL creation, as long as AMD/IBM put the developement of the Torrenza design OVER their short term profit. Because in the long term it will reap them MASSIVE benefits.

[perkam]

once again I have to come out and say Opteron is not about the enthusiast market, it is about Servers and the lucrative profit margins created by the sale of servers. AMD WILL put a smack down on Conroe and yorkfield with K8L, simply because it protects their Server market. And the reason they are SOO busy pimping HTT and Torrenza is because it protects them and the rest of the market from Intel. It gives them the performance edge in EVERY SINGLE benchmark (that matters), It provides them with chipset sales EVEN if their processor isn't used. It makes it a FAR safer and profitable reason for Chipset makers to make Chips that support AMD's sockets. Also being an Open socket, it makes it cheaper for other Processor manufacters to make processors, since now they don't even need to license a bus spec or develope their own. They can just use AMD's which is proven to be of high quality and VERY well designed. They are TOO protected by this useful TOOL. Because THEY KNOW they will have cheap high quality motherboards that COMPLETELY support their new processor design. Weather their design enhances AMD's product or someone else enhances their own. It is a WONDERFUL creation, as long as AMD/IBM put the developement of the Torrenza design OVER their short term profit. Because in the long term it will reap them MASSIVE benefits.Man...who was guy the made the birthday wish that nn_step would start making more serious posts...I have to find him...and have a little talk... outside...

You don't even use smilies anymore :(

Perkam

[nn_step]

That is because I was being serious Perkam.
:D
But no worries. Lots of good news on the way. Plus for the first time We will get to see the performance difference between a tweaked VMX unit and a SSE4 processor :D

[cky2k6]

nn_step, your post made no sense, and honestly, when will you get the fact that you are as much of an amd authority as my dog... you dont have any insider info, this has been proven many times over as your ridiculous claims like ddr3 this year and amd unleashing some secret weapon or whatever, have been complete b.s., so please, you have no idea that k8l will do anything against conroe or especially yorkfield, just like anybody on this forum. put your e-penis away...

[nn_step]

nn_step, put your e-penis away... (its too large and it scares me)

After removing all the :banana::banana::banana::banana::banana::banana:ho le remarks I am left with... :rolleyes:

[cky2k6]

maybe i should take up your role? k8l will be 69% quicker and use xdr with a quad channel imc. proof? i dont need proof! wait and see. (funny how imitations of your type of posts look ridiculous from the side, huh?)

[FghtinIrshNvrDi]

wow, I actually followed nn_step's post just fine.

Toilet's are made for crapping. Use them, instead.

Ryan

[CPLB]

wow, I actually followed nn_step's post just fine.

Toilet's are made for crapping. Use them, instead.

Ryan

I followed his post, but I agree with CKY2K6 in that it is difficult to understand where nn_step gets his information, or why we should even believe him.

[GMX]

Just put him on the ignore list and be done with it. But sometimes i enjoy reading his bull:banana::banana::banana::banana:. Funny stuff, pulled out of his behind.

Speculators.... I HATE speculators. Weather it be in computer hardware, other industries and ESPECIALLY the economy. Speculators :banana::banana::banana::banana: up everything.

[The Ghost]

once again I have to come out and say Opteron is not about the enthusiast market, it is about Servers and the lucrative profit margins created by the sale of servers. AMD WILL put a smack down on Conroe and yorkfield with K8L, simply because it protects their Server market. And the reason they are SOO busy pimping HTT and Torrenza is because it protects them and the rest of the market from Intel. It gives them the performance edge in EVERY SINGLE benchmark (that matters), It provides them with chipset sales EVEN if their processor isn't used. It makes it a FAR safer and profitable reason for Chipset makers to make Chips that support AMD's sockets. Also being an Open socket, it makes it cheaper for other Processor manufacters to make processors, since now they don't even need to license a bus spec or develope their own. They can just use AMD's which is proven to be of high quality and VERY well designed. They are TOO protected by this useful TOOL. Because THEY KNOW they will have cheap high quality motherboards that COMPLETELY support their new processor design. Weather their design enhances AMD's product or someone else enhances their own. It is a WONDERFUL creation, as long as AMD/IBM put the developement of the Torrenza design OVER their short term profit. Because in the long term it will reap them MASSIVE benefits.
good post

this one of the reasons why amd is not losing any market share even against a better cpu , especially in the server market :)

[Lightman]

umm... did everyone miss the fact that altair would be going against yorkfield at 3.4-3.7ghz? if it gains 15% clock for clock over k8, it basically matches conroe's performance clock for clock. yorkfield is definately not any slower than conroe... so amd either better release altair at 3.5ghz or we better hope that its more than 15% improvement clock for clock, which i find doubtfull since amd is not really adding anything that conroe doesnt already have. if amd's torrenza doesnt pan out, then amd mght loose the majority of the enthusiast market for good, whatever it means to mr. hector opteron ruiz.

Small correction, K8L will have doubled FPU units, not only SSE part. This alone in FPU intensive task like 3D Rendering/CAD/HPC/ and to some degree games, will result in BIG speed bump (more than 40%, especially in 3D rendering we can expect even 80% improvement clock for clock from K8).

PS. 15% improvement for K8L is VERY conservative because AMD is adding to K8L 4th decoder (like in conroe), expanding SSE to 128bit at one clock (like in conroe), adding OOO Load buffers and other memory tricks (maybe OOO writes as well, again like in conroe), but what conroe didn't improved (from previous gen) and maintain 15-25% speed increase is cache hierarchy! AMD can win extra %% in this area thanks to doubled L1 bus width + 3rd level!

[ozzimark]

Just put him on the ignore list and be done with it. But sometimes i enjoy reading his bull:banana::banana::banana::banana:. Funny stuff, pulled out of his behind.

Speculators.... I HATE speculators. Weather it be in computer hardware, other industries and ESPECIALLY the economy. Speculators :banana::banana::banana::banana: up everything.
you are kidding right? more than 1/2 the stuff posted on this forum is speculation. "oh, how will my E6300 clock, it's blah stepping?!!!?!?"... "hmm, probably around 3.0-3.4ghz?"
SPECULATION FTW! :slapass:

[savantu]

you are kidding right? more than 1/2 the stuff posted on this forum is speculation. "oh, how will my E6300 clock, it's blah stepping?!!!?!?"... "hmm, probably around 3.0-3.4ghz?"
SPECULATION FTW! :slapass:

Huh ? The 6300 thing is verified while OTOH the K8L is thin air.
When out of 100 6300 users , 70 reached 3GHz on air , I think is very legitimate to have certain expectations. :slap:

[Vincentvega18]

Its when a possibility at BEST is passed off as certain knowledge that it gets on everyones nerves.

[GMX]

Ozzi, that "how will my E6300 clock, or how will my CABYE clock" is based on gathered DATA so an EDUCATED guess can be made. Not some bull**** pulled from his arse or thin air. Big difference mate.

[FghtinIrshNvrDi]

actually, it's not total bull:banana::banana::banana::banana:. It's market speculation derived from supply/demand laws, and what we know about the K8L arch. I'm sorry, y'all are really backin yourselves in a corner.

Ryan

[nn_step]

Ozzi, that "how will my E6300 clock, or how will my CABYE clock" is based on gathered DATA so an EDUCATED guess can be made. Not some bull**** pulled from his arse or thin air. Big difference mate.
:stick: This isn't from my ass, it is from the facts, plain and simple. Learn to read better :stick:

[cky2k6]

so really, you have facts about how k8l will perform? are they like your facts about ddr3? or socket m2?

[Serge84]

Since we know the arc just about inside and out of K8L we know how it will perform because we can understand how it will work and what advantages it has over conroe that conroe doesn't have. When K8L comes out Core3 will be out. So don't be like K8L can't woop conroe its not god its just a cpu that has limits like netburst did and K8 has just like K8L has.

Well it did, because its a more advanced arc and again I say this made to go up agenst Core3, witch is better then conroe by atleast 20%. Who knows if it will beat core 3 or not. All that matters is that its 40% better then K8's and the wait is worth it for me atleast. And we aready have proof that the new K8L's will be at 3.5ghz on stock.

Thats why there is progress. Core3 will prob match it or blow K8L out of the watter, but K10 will then blow core3 away in 2008. So what, it happens until the end of time. lol K8L will just be a killer monster in games. And with that much zram along with its 4 cores it will be a great multi tasker.

Spec for spec K8L is just superior to conroe in every way. We can make jugement from specs and tell how it will perform because its conroe like. But with vast improvments and many things conroe doesn't have K8L does. K8's can not be compared with K8L because they simply are nolonger the same arc's related any more. K8L is so different its just not even funny.

Like many have said about the specs, and of my past posts saying everything about the arcs improvments, we are right in saying how it will perform. Even AMD has said it themselfs. With us giving smart info on it and how it will perform. But non beleavers will just have to see reality for themselfs when it comes out. People can beleave what they wish, some won't beleave at all others will. Thats your oppinion and others have their own.

We aready know the facts based on pervious evedince on the specs when they where released along with die shots afew months ago. There is not guessing. What everybody said that I said before in some way is right. Why don't you google it for yourself and see the truth. Nobody made anything up so far.

[cky2k6]

Nobody made anything up so far.
really?

AMD WILL put a smack down on Conroe and yorkfield with K8L
also, i have no idea what core 3 you are talking about, yorkfield is simply a 45nm quad core version of core 2 with shared cache, and nehalem is only coming in 2008. altair will not debute at 3.5ghz, it will max out at 3.5ghz according to that article, and we dont even know what process that is, it could be in 2008 when 45nm will launch. feature wise, again, altair is not significantly better than conroe, and yorkfield will likely have some slight architectural improvements in addition to reall sse4 support, like altair.

[nn_step]

I'm an intel fanboy and will refuse to admit that AMD will ever make anything that will ever surpass conroe
I think that explains everything you have said in this thread :stick:

[CPLB]

I think that explains everything you have said in this thread :stick:

Yet, he's running an AMD Opteron in his computer... I'm all for AMD and have no doubts K8L will beat Conroe, but nn_step, in all honesty I think you're the last person that should be accusing others of being "fanboys."

[cky2k6]

lol, im being called an intel fanboy when my last intel was before i ever built a computer... difference is, intel either sucked or was overpriced for the longest time, but now, it isnt... if you notice, i never even said amd wont beat conroe. im saying that its unlikely from what information is known that amd will beat yorkfield clocked at 3+ghz. for as much as i dont like intel, i will never be brand loyal to the point of losing out, and i wont cut amd slack for being behind intel for whatever reason it may be.

[ozzimark]

Ozzi, that "how will my E6300 clock, or how will my CABYE clock" is based on gathered DATA so an EDUCATED guess can be made. Not some bull**** pulled from his arse or thin air. Big difference mate.
well i see it as the same thing. maybe i'm just weird. i'm just one of those ones who thinks that people should try to avoid randomly posting negative comments about others, like you did. the forum would be a MUCH better place to randomly browse.. :fact: ;) ;)


so really, you have facts about how k8l will perform? are they like your facts about ddr3? or socket m2?
as gmx said, educated guesses on what changes are made to the cpu my friend :toast:

[Kingcarcas]

Since we know the arc just about inside and out of K8L we know how it will perform because we can understand how it will work and what advantages it has over conroe that conroe doesn't have. When K8L comes out Core3 will be out. So don't be like K8L can't woop conroe its not god its just a cpu that has limits like netburst did and K8 has just like K8L has.

Well it did, because its a more advanced arc and again I say this made to go up agenst Core3, witch is better then conroe by atleast 20%. Who knows if it will beat core 3 or not. All that matters is that its 40% better then K8's and the wait is worth it for me atleast. And we aready have proof that the new K8L's will be at 3.5ghz on stock.

Thats why there is progress. Core3 will prob match it or blow K8L out of the watter, but K10 will then blow core3 away in 2008. So what, it happens until the end of time. lol K8L will just be a killer monster in games. And with that much zram along with its 4 cores it will be a great multi tasker.

Spec for spec K8L is just superior to conroe in every way. We can make jugement from specs and tell how it will perform because its conroe like. But with vast improvments and many things conroe doesn't have K8L does. K8's can not be compared with K8L because they simply are nolonger the same arc's related any more. K8L is so different its just not even funny.

Like many have said about the specs, and of my past posts saying everything about the arcs improvments, we are right in saying how it will perform. Even AMD has said it themselfs. With us giving smart info on it and how it will perform. But non beleavers will just have to see reality for themselfs when it comes out. People can beleave what they wish, some won't beleave at all others will. Thats your oppinion and others have their own.

We aready know the facts based on pervious evedince on the specs when they where released along with die shots afew months ago. There is not guessing. What everybody said that I said before in some way is right. Why don't you google it for yourself and see the truth. Nobody made anything up so far.
This man know what he's talking about, K8L will be my next upgrade :banana:

[Theli]

we know the arc... we can understand how it will work and what advantages it has over conroe that conroe doesn't have... don't be like K8L can't woop conroe... more advanced arc... for me atleast... we aready have proof that the new K8L's will be at 3.5ghz on stock... blow K8L out of the watter... blow core3 away... in 2008... until the end of time... lol... killer monster in games... We can make jugement from specs and tell how it will perform because its conroe like... vast improvments... nolonger the same arc's related any more... its just not even funny... Like many have said about the specs, and of my past posts saying everything about the arcs improvments, we are right in saying how it will perform... With us giving smart info... non beleavers... some won't beleave at all others will... Thats your oppinion... facts based on pervious evedince on the specs... There is not guessing... Nobody made anything up so far.
What?

[KeZzZu]

stop talking :banana: :banana: :banana:

amd k8 old arc-->am2 only ddr2 support and little improvements

and then conroe new fresh arc new desing, you cant even compare those cpus!

and then comes intelfan boy to say "my conroe will beat your k8, its so pwn cpu... etc " <--- that is so :banana: :banana: :banana: new tech and old tech you just can't compare k8 and conroe! thats is my opinion and its make me very angry when somebody are comparing k8 and conroe.

but i wanna k8l cpu! sorry my bad english skills

[FghtinIrshNvrDi]

sorry my bad english skills

Lol! You did fine!

Ryan

[perkam]

AMD putting the smackdown on conroe with the K8L is pretty much expected, considering Intel would have not announced a brand new architecture one year ahead of its release had intel not thought of K8L has a threat to Conroe...Think about it...Intel has superior R&D and manufacturing capacity, hence the reason how it got to 65nm so much faster than AMD...why on earth would it make a product with similar performance to K8L and still only be able to release it AT THE SAME TIME AS K8L ??? (Both coming Q3 2k7)

Had Intel been planning to debut yorkfield prior to K8L, the roadmaps a year ago with penryn and Nehalem would have mentioned Yorkfield.

However, K8L laying the smackdown on Yorkfield is unlikely as if it IS supposed to be a response to K8L, its going to be something special.

Perkam

[cky2k6]

wait, so yorkfield is more than just a core 2 quadro with shared cache? what improvements will it have? that seems hard to believe, im so damn confused with intel's roadmap...

[savantu]

AMD putting the smackdown on conroe with the K8L is pretty much expected, considering Intel would have not announced a brand new architecture one year ahead of its release had intel not thought of K8L has a threat to Conroe...Think about it...Intel has superior R&D and manufacturing capacity, hence the reason how it got to 65nm so much faster than AMD...why on earth would it make a product with similar performance to K8L and still only be able to release it AT THE SAME TIME AS K8L ??? (Both coming Q3 2k7)

Had Intel been planning to debut yorkfield prior to K8L, the roadmaps a year ago with penryn and Nehalem would have mentioned Yorkfield.

However, K8L laying the smackdown on Yorkfield is unlikely as if it IS supposed to be a response to K8L, its going to be something special.

Perkam

Why not look at the matter in a slightly different way ?

AMD has no response to the summer 2006 Conroe but the 1 year later K8L which will equal Conroe and get it's butt kicked by the 2008 Nehalem generation ?

40% faster than K8 is probably for FP intensive codes like Linpack ( extra FPU helps there most ) which translates into roughly 10% in general tasks.

[The Ghost]

Why not look at the matter in a slightly different way ?

AMD has no response to the summer 2006 Conroe but the 1 year later K8L which will equal Conroe and get it's butt kicked by the 2008 Nehalem generation ?

40% faster than K8 is probably for FP intensive codes like Linpack ( extra FPU helps there most ) which translates into roughly 10% in general tasks.
what makes you think that the K8L will only equal conroe core ?

what do you think that the nehalem is going to have that core 2 doesn't have ?

[Serge84]

I'm going to ignore the BS and try something intresting. Lets see if I can't figure out prices. These are just guesses from now and the future when the higher models come out. Who knows what they will really be but its cool to hope.

Spica Core Sempron/Athlon
1.6Ghz 512Kb L2 $50
1.7Ghz 512Kb L2 $55
1.8Ghz 512Kb L2 $60
1.9Ghz 512Kb L2 $65

2.0Ghz 512Kb L2 $70

2.1Ghz 512Kb L2 $75
2.2Ghz 512Kb L2 $80
2.3Ghz 512Kb L2 $85
2.4Ghz 512Kb L2 $90

Arcturus Core X2 W65
2.1Ghz 1Mb L2 $100
2.2Ghz 1Mb L2 $115
2.3Ghz 1Mb L2 $130

Antares Core X2 W89
2.0Ghz 2Mb L3 $120
2.1Ghz 2Mb L3 $140
2.2Ghz 2Mb L3 $160
2.3Ghz 2Mb L3 $180
2.4Ghz 2Mb L3 $200
2.5Ghz 2Mb L3 $220
2.6Ghz 2Mb L3 $240
2.7Ghz 2Mb L3 $260
2.8Ghz 2Mb L3 $280
2.9Ghz 2Mb L3 $300

Remember the quad cores will use the same power draw as todays X2's so K8L X2's will beable to use half of the power at much higher speeds.

Antares Core X2 W65
2.0Ghz $130
2.1Ghz $150
2.2Ghz $170
2.3Ghz $190
2.4Ghz $210
2.5Ghz $230
2.6Ghz $250
2.7Ghz $270
2.8Ghz $290
2.9Ghz $310

Antares Core X2 W35
2.0Ghz $140
2.1Ghz $160
2.2Ghz $180
2.3Ghz $200
2.4Ghz $220
2.5Ghz $240
2.6Ghz $260
2.7Ghz $280
2.8Ghz $300
2.9Ghz $320

Altair Core X4 At Launch & Price Drop
2.7Ghz 4Mb L2/L3 $500/$400
2.8Ghz 4Mb L2/L3 $650/$500
2.9Ghz 4Mb L2/L3 $800/$600

Altair Core FX At Launch 07
2.7Ghz 4Mb L2/L3 $800/$600
2.8Ghz 4Mb L2/L3 $900/$650
2.9Ghz 4Mb L2/L3 $999/$700

Altair Core Fx After Launch 08 Prob at 45nm
3.0Ghz Unknown $750
3.1Ghz Unknown $800
3.2Ghz Unknown $850
3.3Ghz Unknown $900
3.4Ghz Unknown $950
3.5Ghz Unknown $999

[vitaminc]

what do you think that the nehalem is going to have that core 2 doesn't have ?

If AMD could improve the current K8 core for higher IPC, Intel could done exactly the same thing, i.e., wider ID/IF stage, more EX stage ALU/SSE units, etc.

[savantu]

what makes you think that the K8L will only equal conroe core ?

AMD disclosed a few perf expectations of K8L and it was 100% faster than K8 for server tasks not FP oriented => equal perf core/core.
Remember that the L2 is cut in half and the L3 has to supply 4 cores.AMD was never a reference point in cache design so it might happen for the cores to be starved for data.


what do you think that the nehalem is going to have that core 2 doesn't have ?

IMC and CSI for first.Double L2 cache.Tweaked core.Improved FPU units.On-die accelerators.Enough ?

[Sparky]

After holding the crown for 3 years I don't think AMD is really going to be ready to sit back and release something that only just matches core 2. That just isn't how they operate. They design and build a CPU to overtake intel, not match. I'm sure AMD is fully aware of kentsfield and will surely have something in mind to overpower it as well.

I'm waiting for K8L personally, I think I'm going to totally skip the X2... I don't have the money to upgrade even if I wanted to anyway :p:

[perkam]

Why not look at the matter in a slightly different way ?

AMD has no response to the summer 2006 Conroe but the 1 year later K8L which will equal Conroe and get it's butt kicked by the 2008 Nehalem generation ?

40% faster than K8 is probably for FP intensive codes like Linpack ( extra FPU helps there most ) which translates into roughly 10% in general tasks.I'm going to make this VERY clear:

You must have some pretty good friends at Intel to feel confident saying that Hector Ruiz is going to debut a brand new architecture one year from now that is going to be one year behind its time...

This is amd...remember...the same one that the P4 was 3 years behind ???

What kind of naive butter do ppl eat in the morning to start assuming K8L will equal conroe performance ?

Must be some think lard in that stuff cos its blocking the blood to the brain right there.

Perkam

[Death^Dread]

stop talking :banana: :banana: :banana:

amd k8 old arc-->am2 only ddr2 support and little improvements

and then conroe new fresh arc new desing, you cant even compare those cpus!

and then comes intelfan boy to say "my conroe will beat your k8, its so pwn cpu... etc " <--- that is so :banana: :banana: :banana: new tech and old tech you just can't compare k8 and conroe! thats is my opinion and its make me very angry when somebody are comparing k8 and conroe.

but i wanna k8l cpu! sorry my bad english skills


What's wrong with comparing two competing technologies that are their respective company's top of the line?

Do you think when the G80 comes out, people won't talk about how bad it owns the x1950xt? Do you think when the first jet engined aircraft came out, people didn't compare them to turboprop aircraft? Whatever currently available technologies exist will be compared. It's what's out there now, not "yeah well just wait 8 months and see how your Core 2 Duo looks then." Personally, I'm really looking forward to AMD and Intel warring with each other. In fact, I desperately hope that NEITHER has significantly better performing technologies, that way price wars will become more aggressive. AMD WILL come out with a technology better than Conroe. Intel will come out with technologies better than Conroe. Question is, which will be better, by how much, and when?

[The Ghost]

If AMD could improve the current K8 core for higher IPC, Intel could done exactly the same thing, i.e., wider ID/IF stage, more EX stage ALU/SSE units, etc.
lets see , intel has already done that , now amd is going to do it it's way


AMD disclosed a few perf expectations of K8L and it was 100% faster than K8 for server tasks not FP oriented => equal perf core/core.
Remember that the L2 is cut in half and the L3 has to supply 4 cores.AMD was never a reference point in cache design so it might happen for the cores to be starved for data.
correct me if i am wrong , but doesn't the L2 cache get filled up with what the L1 can not use right away ? and the same thing would be true for L2 and L3 ?

L1 cache is the fastest memory on the cpu , L2 is half the speed of L1 cache and L3 is slower L2cache , by the way , did cutting the L2 in half slow down the overall performance of the cpu ?


IMC and CSI for first.Double L2 cache.Tweaked core.Improved FPU units.On-die accelerators.Enough ?
i doubt that intel will have csi ready in 2008 , and they won't put imc in core 2 , so they would wait for their next all new core

how is intel going to improve it's fpu ? it has already done that , intel will have a all new core for imc and csi , it is possible , but from what intel has said in the past that it would be 2009 before they had csi ready

[nn_step]

correct me if i am wrong , but doesn't the L2 cache get filled up with what the L1 can not use right away ? and the same thing would be true for L2 and L3 ?
Depends if the memory is inclusive or exclusive

[The Ghost]

Depends if the memory is inclusive or exclusive
Exclusive

[nn_step]

basically the Information most expected to be used is stored in the L1, then the next most likely needed material is stored in the L2, and the rest of the Data expected to be needed are stored in the L3. Everything after that is evicted to the Ram and after that evicted to the Hard drive.
inclusive Cache on the other hand, can keep multiple copies of the same data in different cache levels

[FghtinIrshNvrDi]

If it's anything like this guy, you can find him exclusively at Rockhurst University. And ladies, I aint been evicted, so come on over to my place. ;)

Ryan

[IvanAndreevich]

basically the Information most expected to be used is stored in the L1
Actually in registers, unless it's an array or there are some other restrictions.

inclusive Cache on the other hand, can keep multiple copies of the same data in different cache levels
Can? It's inclusive, so it MUST.

[savantu]

lets see , intel has already done that , now amd is going to do it it's way


correct me if i am wrong , but doesn't the L2 cache get filled up with what the L1 can not use right away ? and the same thing would be true for L2 and L3 ?

AMD used inclusive until now , but I expect K8L to have inclusive for L1-L2 and exclusive for L2-L3. ( or exclusive both like Intel )



L1 cache is the fastest memory on the cpu , L2 is half the speed of L1 cache and L3 is slower L2cache , by the way , did cutting the L2 in half slow down the overall performance of the cpu ?

I don't think the ratios are valid but anyway cutting the L2 in half will obviously affect performance ( especially for server tasks ) and there is where the shared L3 should help a lot.

The shared L3 also means less coherency problems but OTOH a shared cache able to supply 4 cores is a very very difficult task.



i doubt that intel will have csi ready in 2008 , and they won't put imc in core 2 , so they would wait for their next all new core

how is intel going to improve it's fpu ? it has already done that , intel will have a all new core for imc and csi , it is possible , but from what intel has said in the past that it would be 2009 before they had csi ready

That is Nehalem which has a target date of Q2-Q3 2008.Pat Gelsinger said they have improved FP power with Conroe but it wasn't enough.

[The Ghost]

AMD used inclusive until now , but I expect K8L to have inclusive for L1-L2 and exclusive for L2-L3. ( or exclusive both like Intel )
am i mixed up ?

http://forums.amd.com/lofiversion/index.php/t23896.html
The K7 and K8 use a more common trick aimed at a similar purpose. As I mentioned above, maintaining exclusivity between the caches often takes a few extra operations, and that has consequences in a number of places. If the requested data is not in the L1 but is in the L2, a line from the L1 would first have to be copied into L2, and then the requested data moved from L2 into L1.

http://forums.amd.com/index.php?act=Print&client=printer&f=28&t=23896
The P4 uses inclusive caching, while the K7 and K8 use exclusive caching.

http://www.cpuid.com/reviews/K8/index.php
K7 and K8 use an exclusive relationship between L1 and L2, in opposition to Intel that uses an inclusive relationship.

this last link has a chart that shows amd being exclusive and intel inclusive


I don't think the ratios are valid but anyway cutting the L2 in half will obviously affect performance ( especially for server tasks ) and there is where the shared L3 should help a lot.

The shared L3 also means less coherency problems but OTOH a shared cache able to supply 4 cores is a very very difficult task.
on the same page as the link above shows the difference in speeds of L1 and L2 cache , as far as i know amd might be using z-ram which is slower then L2 cache speed

quote from the page

As mentionned before, the theorical maximum bandwidth of the L1 is 28800MB/s when using 64 bits reads. The Athlon XP offers 65% of this bandwidth with a raw value of 18800MB/s, and the Opteron reaches 79% of this theorical value with 22700MB/s. These results confirms what we obtained for 32 bits reads.

The L2 of the Opteron provides a 8260MB/s bandwidth, and the Athlon XP 5700MB/s. In both cases, the results are far from the maximum theorical value of 1 op x 8 bytes x 1800MHz = 14400MB/s. With 64 bits reads, the L2 is saturated, and offers the maximum bandwidth it can provide. In comparison to 32 bits, the Opteron increased its L2 bandwidth from 27%, whereas the Athlon XP increased from 14% only. This difference is due to the Opteron largest L2 bus width, that allows the L2 to increase its bandwidth.

latencies are listed below on the same page

[Turtle 1]

Another interesting read . It seems that K8L is going to get 25% of its performance increase do to higher clocks. and only 15% from arch. improvements . According to the link posted at the top. Not impressive at all.

http://badhardware.blogspot.com/2006/10/amds-k8l-revealed-in-cray-rainier.html

[informal]

Another interesting read . It seems that K8L is going to get 25% of its performance increase do to higher clocks. and only 15% from arch. improvements . According to the link posted at the top. Not impressive at all.

http://badhardware.blogspot.com/2006/10/amds-k8l-revealed-in-cray-rainier.html

WOW man,that sure is the most reliable source of information i have ever seen.Wow again.
Not.
I think that people like you are in some nasty surprise Q2 come.

[informal]

Quit flamebating everyone and keep an open mind.

Ok,tell me how that guy from that site have any idea how well new core performs if he hasn't tested one??
Maybe i don't get it,so you can enlighten me.:)

[informal]

Looks like it MIGHT BE a DOE engineer (I know one, in fact). They are the guys that run a lot of these Cray's. And they DO GET AMD ES chips about 2 months ahead of everyone else (US gov actually has a contract w/ AMD for this).

Anyway, I would say this site has about as much credibility as the Inq (take it FWIW).

But that is a SIDE POINT:

You could have said the EXACT same thing w/o being a complete p rick. You were flame-baiting him.

Well excuse me for not poluting intel topics with similar info.
I just don't like when someone continually posts negative comments in AMD topics(which he does,look at his post history).
I won't argue anymore and will not try to point out anymore what i've just said a couple posts ago.

[mesyn191]

Quit flamebating everyone and keep an open mind.

Dude, Turtle there is the one who is flame baiting and does so regularly.

[Turtle 1]

Dude, Turtle there is the one who is flame baiting and does so regularly.


I beg your pardon . I read a thread in forum that had a link to it. The link was supplied as a source for information.

After reading the info from that source (link) I was very uninpressed.

Further you have same people that said Conroe wouldn't be around until oct 06. Are now saying K8L will be here in qt1 qt2 of 07 . When all indications are a 3rd qt release. I still say 4th qt release but will see. :nono:

[nn_step]

If this IS THE CASE (and we have no evidence it is or is not, except this blog), then that just brings K8L up to Conroe speeds. That would be VERY disappointing.

Hopefully they are wrong.
Dude we have already covered this is incorrect, and an understatement of K8L performance

[Sparky]

speculation speculation speculation..... I hate it. I hate waiting too! I can hardly wait until we have the actual chips in our hands :D

Seems (to me) that AM3 has kind of disappeared from discussion - is that in the far future so it would make sense to buy AM2 soon or would it be better to wait for AM3 to get K8L?

Ehhhh I'm probably missing something someplace but hey what else is new.

[ozzimark]

i think most of us are looking forward to a new core MUCH more than a simple move to a new socket to support ddr3 and some other things

[Serge84]

Some people just can't wait to say "I told you so" to some of the unbeleavers here. Let K8L show you the path to enlightenment and rid your mind of all evil delusions... I think I been watching too much SG-1. =P

[[XC] Teroedni]

speculation speculation speculation..... I hate it. I hate waiting too! I can hardly wait until we have the actual chips in our hands :D

Seems (to me) that AM3 has kind of disappeared from discussion - is that in the far future so it would make sense to buy AM2 soon or would it be better to wait for AM3 to get K8L?

Ehhhh I'm probably missing something someplace but hey what else is new.

I think am3 definitively will be better:D

But you could also buy 2x 2xxx opteron system and upgrade to k8l already in 2007 h2, althought the price probably would be at premium then.
But as far as i know it will be possible to plug k8l server chips in socket f


Im seriosly considering wheher to go for a dual socket f system now or a bigger lcd.

[nn_step]

Some people just can't wait to say "I told you so" to some of the unbeleavers here. Let K8L show you the path to enlightenment and rid your mind of all evil delusions... I think I been watching too much SG-1. =P
You know there is no such thing as watching too much SG-1 :rolleyes:
K8L is a processor not a religion, the same goes for Conroe but alot of people seem to have forgotten that.

[freeloader]

Why argue over something that's not even here yet! :D

It's only a processor...

[ozzimark]

Why argue over something that's not even here yet! :D

It's only a processor...
because it's fun! :fact:


:D

[Sparky]

Why argue over something that's not even here yet! :D

It's only a processor...
cuz we wanna :p:

I suppose when K8L is released I'll have to decide whether to hang onto my aging s939 system until AM3 or upgrade right then. Money will probably play a big part in it though so we'll have to see. I hope K8L AM3 isn't that far away, it'll make deciding easier!

[vitaminc]

cuz we wanna :p:

I suppose when K8L is released I'll have to decide whether to hang onto my aging s939 system until AM3 or upgrade right then. Money will probably play a big part in it though so we'll have to see. I hope K8L AM3 isn't that far away, it'll make deciding easier!

sell ur s939 system when the resale value is still high, get am2 and u be k8l proof. :p

[Serge84]

You know there is no such thing as watching too much SG-1 :rolleyes:
K8L is a processor not a religion, the same goes for Conroe but alot of people seem to have forgotten that.

I know what you mean. But I said it like that cus alot of conroe lovers do love the thing that much and most don't even have a K8 to compare it with they just automaticly think conroe is a big leap when its only a small step. I was just trying to be funny with the consept they have with conroe. They think it has supernatual powers when oviously its just a cpu nothing to get super excited about like a kid at a candy store.