A bunch of brand new info on K8L for you guys. With info on a 3.5Ghz K8L

[RichUK]

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

 

[BrownTown]

don't see anything new there, except of course the random bold green type declaring that K8L will run ad 3.5G even though the charts don't support that fact...

 

[Regs]

Well, at least the bottom chart shows truth. A lot of blank white segments to show what they plan to compete with....nothing lol.

So it will be 65nm vs 45nm in 3rd quarter of 07. Nothing suprising of AMD. A die shrink is just a die shrink. Though it would be interesting to see a 65nm quad vs a 45nm quad CPU.

 

[RichUK]

Yeah it does seem like you have sieve through a lot of sh!t on that page to get the little tid bits of info.

 

[easy123]

Originally posted by: Regs
Well, at least the bottom chart shows truth. A lot of blank white segments to show what they plan to compete with....nothing lol.

LOL that's pretty funny. Shouldn't the current X2/FXs be there? I guess they aren't deemed 'competitive' enough to be in the shame chart. :/

 

[lopri]

What I am not getting is that the page claims K8L's performance 40% performance increase will come from clockspeed (25%) and architectural enhancement (15%). I'd think 'performance increase' is measured on per-clock base.

 

[easy123]

Yeah that's pure speculation at this stage.

Either way, whether it be clockspeed or IPC or both, I hope AMD can deliver on it's promise of a 40% performance improvement over K8.

 

[BrownTown]

They gonna have to if they want to take the performance crown back anytime soon.

 

[jfkaiser]

They took the information from this url

http://www.hkepc.com/bbs/news.php?tid=679532&starttime=0&endtime=0

And mis-read the data presented ..

 

[lyden]

we are about 1 year away from actualy being able to go to a store and buy it ... also i hope they do speed up the development as when tehy say Q3 its always near October b4 that product actually comes ...

couple hopes:

1. speed things up as intel is slowly eating away ots market...

2. k8l better havge fast clock for clock performance compare to Conroe. ie singe core performance, best way to measure is to run some older games at lower resolution.

3. not require tight timing memory to get the best performance. sometimes i go to AMD forums .. and seeing people with 2.5-3-3-4 RAM at DDR 400 ... tight timing means slower suer pi tiome for intel and thats pretty much it, not much real world performance lost.

 

[dmens]

god, what a load of garbage in that "article".

25% cycle time decrease while gaining 15% on uarch perf on the same process? good luck on that. impressive if it works though, ill admit.

 

[myocardia]

Originally posted by: dmens
god, what a load of garbage in that "article".

25% cycle time decrease while gaining 15% on uarch perf on the same process? good luck on that. impressive if it works though, ill admit.

K8L is far from being the same process as K8. K8L is actually K9, though they've decided not to give Intel and it's fanboys the K9 moniker to toy with. So, they decided to name it K8.5, or K8L, since L=50 in Roman numerals.

 

[lopri]

On a second thought, I guess clock speed increase can actually be acclaimed if the whole line-up of K8L shift by like .5GHz. (Starting-cheapest model 3.0GHz, for example)

 

[Regs]

The K8L will first need to feed 3.5GHz without pipeline stalls. Example: Prescott.

 

[myocardia]

Originally posted by: Regs
The K8L will first need to feed 3.5GHz without pipeline stalls. Example: Prescott.

Agreed, but then, the Preshott architecture wasn't efficient at any speed.

 

[zephyrprime]

Originally posted by: myocardia

Originally posted by: dmens
god, what a load of garbage in that "article".

25% cycle time decrease while gaining 15% on uarch perf on the same process? good luck on that. impressive if it works though, ill admit.

K8L is far from being the same process as K8. K8L is actually K9, though they've decided not to give Intel and it's fanboys the K9 moniker to toy with. So, they decided to name it K8.5, or K8L, since L=50 in Roman numerals.

Everything I've makes me doubt that the K8l is really a different enough to be truly the K9. Besides, wasn't the K9 design effort canned?

dmens: The K8L is on a different process so I don't get where you get the k8l is on the same process. A 25% from going from to smaller process isn't exceptional.

 

[dexvx]

Originally posted by: myocardia

Originally posted by: dmens
god, what a load of garbage in that "article".

25% cycle time decrease while gaining 15% on uarch perf on the same process? good luck on that. impressive if it works though, ill admit.

K8L is far from being the same process as K8. K8L is actually K9, though they've decided not to give Intel and it's fanboys the K9 moniker to toy with. So, they decided to name it K8.5, or K8L, since L=50 in Roman numerals.

Based on the architectural specs, it seems K8L is just 4x K8 dies with a shared L3 and some tweaks to the IMC/HT on fabbed on a 65nm process. Btw, K9 was another project and it was canned.

 

[HopJokey]

Originally posted by: myocardia

Originally posted by: dmens
god, what a load of garbage in that "article".

25% cycle time decrease while gaining 15% on uarch perf on the same process? good luck on that. impressive if it works though, ill admit.

K8L is far from being the same process as K8. K8L is actually K9, though they've decided not to give Intel and it's fanboys the K9 moniker to toy with. So, they decided to name it K8.5, or K8L, since L=50 in Roman numerals.

Kind of off topic here, but AMD did not name it's next major uarch revision "K8L". That is something Intel came up with to describe AMD's next uarch change.

AMD's offical name of "K8L" is simply K8, Rev. H.

Just thought I clear that up

 

[RichUK]

Originally posted by: HopJokey

Originally posted by: myocardia

Originally posted by: dmens
god, what a load of garbage in that "article".

25% cycle time decrease while gaining 15% on uarch perf on the same process? good luck on that. impressive if it works though, ill admit.

K8L is far from being the same process as K8. K8L is actually K9, though they've decided not to give Intel and it's fanboys the K9 moniker to toy with. So, they decided to name it K8.5, or K8L, since L=50 in Roman numerals.

Kind of off topic here, but AMD did not name it's next major uarch revision "K8L". That is something Intel came up with to describe AMD's next uarch change.

AMD's offical name of "K8L" is simply K8, Rev. H.

Just thought I clear that up

LOL, Even the top people at AMD (Forget their names now), refer to its code name as K8L. Where the hell did you get that info from???

 

[HopJokey]

Originally posted by: RichUK

Originally posted by: HopJokey

Originally posted by: myocardia

Originally posted by: dmens
god, what a load of garbage in that "article".

25% cycle time decrease while gaining 15% on uarch perf on the same process? good luck on that. impressive if it works though, ill admit.

K8L is far from being the same process as K8. K8L is actually K9, though they've decided not to give Intel and it's fanboys the K9 moniker to toy with. So, they decided to name it K8.5, or K8L, since L=50 in Roman numerals.

Kind of off topic here, but AMD did not name it's next major uarch revision "K8L". That is something Intel came up with to describe AMD's next uarch change.

AMD's offical name of "K8L" is simply K8, Rev. H.

Just thought I clear that up

LOL, Even the top people at AMD (Forget their names now), refer to its code name as K8L. Where the hell did you get that info from???

http://www.theinquirer.net/default.aspx?article=33906

Also that is what a big wig at work told us during a forum.

Everyone uses the "K8L" term now to describe AMD's next uarch upgrade. It's pretty much what everyone calls it now, even Henri Richard at AMD used the term (I think that is the person you were eluding to).

BTW, nice sig Rich

 

[RichUK]

Originally posted by: HopJokey

Originally posted by: RichUK

Originally posted by: HopJokey

Originally posted by: myocardia

Originally posted by: dmens
god, what a load of garbage in that "article".

25% cycle time decrease while gaining 15% on uarch perf on the same process? good luck on that. impressive if it works though, ill admit.

K8L is far from being the same process as K8. K8L is actually K9, though they've decided not to give Intel and it's fanboys the K9 moniker to toy with. So, they decided to name it K8.5, or K8L, since L=50 in Roman numerals.

Kind of off topic here, but AMD did not name it's next major uarch revision "K8L". That is something Intel came up with to describe AMD's next uarch change.

AMD's offical name of "K8L" is simply K8, Rev. H.

Just thought I clear that up

LOL, Even the top people at AMD (Forget their names now), refer to its code name as K8L. Where the hell did you get that info from???

http://www.theinquirer.net/default.aspx?article=33906

Also that is what a big wig at work told us during a forum.

Everyone uses the "K8L" term now to describe AMD's next uarch upgrade. It's pretty much what everyone calls it now, even Henri Richard at AMD used the term (I think that is the person you were eluding to).

BTW, nice sig Rich

Thanks

Thats crazy about the K8L story, if thats the case 'll just call it next gen.

Thanks for the heads up.

 

[Munky]

Originally posted by: lyden
3. not require tight timing memory to get the best performance. sometimes i go to AMD forums .. and seeing people with 2.5-3-3-4 RAM at DDR 400 ... tight timing means slower suer pi tiome for intel and thats pretty much it, not much real world performance lost.

The K8 archtecture prefers low latency mem, as opposed to high latency, high bandwidth mem. Nothing wrong with that, and certainly no reason to switch just because intel goes the other way. I think the trend will continue with the K8L, unless AMD all of a sudden decides to pull it's own Presscott by lengthening the pipeline 2x and getting rid of the onboard mem controller.

 

[Regs]

Originally posted by: RichUK

Thanks

Thats crazy about the K8L story, if thats the case 'll just call it next gen.

Thanks for the heads up.

When in rome. Might as well go with the flow, right?

 

[Regs]

Originally posted by: dexvx
Based on the architectural specs, it seems K8L is just 4x K8 dies with a shared L3 and some tweaks to the IMC/HT on fabbed on a 65nm process. Btw, K9 was another project and it was canned.

How much did we know about the clawhammer before it landed in Anand's hand for preview? Maybe the HT bus and IMC very little of anything else.

What architectural Specs? Everything AMD released was too vague to even consider other wise. Everything else is speculation. At first impression it seems just a revamped K8 with only minor adjustments, you may think. It's not surprising for people to think so negatively when Conroe is cleaning house on performance and performance per watt. It also doesn't help when AMD is throwing out a new revision every other month.