AMD K8L ready to blood Intel
INQ reports more detail about K8L here. It is also reported that AMD's next generation will consume 43% less power than Intel's Conroe/Woodcrest.
A force multiplier for Opteron is co-processor: media processing, JVM/CLR acceleration, TOE, XML or SSL processing, you name it. I envisioned an industry food chain for these co-procs based on ccHT previously. I hope there will be a Parrot co-processor that can speed up PERL6, Python and other dynamic languages to 2x C++ speeds.
Intel, ready for cluster frags!
posted by Sharikou, Ph. D. @ 5/16/2006 11:09:00 PM
19 Comments:
3. 128b FPUs - see 14,15
Where are 14, 15?
For that matter, where's 12?
14, 15 are about SSE .
http://news.com.com//i/ne/p/2006/amdslide3_539x403.jpg
http://news.com.com//i/ne/p/2006/amdslide4_532x402.jpg
The link is not working ..
Do you have a copy somewhere
The missing points are:
11. 32B instead of 16B ifetch
12. Indirect branch predictors
13. OOO load execution - similar to memory disambiguation
14. 2x 128b SSE units
15. 2x 128b SSE LDs/cycle
16. Several new instructions
The question is, when?
The Opterons in 2007, and the desktop varaints in 2008?
Either way, not too good. Intel's 45nm quad cores slated for 2007; and Intel's next-gen microarchitecture Nehalem will come out in 2008.
And Nehalem might just have CSI and and IMC, just like the Tukwila...
If you look at die, it is ZRAM, not standard cache! If you compare dual-core Athlon64 you will notice that 1MB of L2 cache is the same as CPU. On this picture it is clear that L2 cache is 4x smaller (size).
So the size of quad-core chip should be 40% bigger than dual-core.
here is the correct link to presentation slide.
"The question is, when?
The Opterons in 2007, and the desktop varaints in 2008?
Either way, not too good. Intel's 45nm quad cores slated for 2007; and Intel's next-gen microarchitecture Nehalem will come out in 2008.
And Nehalem might just have CSI and and IMC, just like the Tukwila..."
They haven't even released conroe and now you are something few generation ahead. You know what, AMD will have a 10GHZ by 2008!
Intel:
65 nm = 2006
(Yonah, Presler, CedarMill -> Core 2 Duo)
45 nm = 2008 (FDSOI -> AMD ?)
32 nm = 2010
If Intel is planning to have 45nm by next year, than there must be a magic FAB that is hidden to us all, mass producing CPU's that haven't been thought up yet.
Intel has yet to produce a Conroe for Consumers at their supposed low prices, and they're already (according to you Shar) working on 45nm for next year? hmm...
Last time I checked, Nehalem was the original name for the Pentium 5, and since we're all good boys and girls who read every article we can, we would know Nehalem is now Conroe.
I've seen Tukwila slides and all the fancy information, it has IMC and CSI, but the problem is, is that be it a Quad-Core, not all 4 cores can be working 100% at once due to it having to stay within it's thermal design package (TDP).
So...as we see, currently, Intel cannot make a working Quad-Core, because for it to be called Quad-Core, I think all 4 cores have to work all the time....yea...just maybe....
So far, CSI in the labs (from what I read) is sketchy at best, and the performance is below that of HT 1.0, and I think we're at 3.0 now :O!
I hope Shar let's this message through, because Intel has nothing really to combat K8L or K10 coming, unless you think Conroe ES's are true performance, but adding 1 extra issue in the pipeline, and a few more instructions, do not give you 40% over a K8, that's just reedeeculous!!
I hope that K8L will finally bring back AMD to desktop performance game. Cause right now, Intel product that is 5 generations behind AMD is beating the sh!t out of K8. Scare to even imagine if Intel somehow managed to add some of the "generation" stuff like IMC.
Intel does not have a released product that "beats the hell out of AMD". Conroe most likely will, but it won't be released until later this summer.
Will brisbane due december 2006, sport the K8L already? Or will be just a K8 die shrink?
I want to re-think over why Intel showed "engineering" samples of the Conroe ..ofcourse for benchmarketing.
But my real question is ..if they have made engineering samples ..why is the delay in mass producing ? given Intels situation ..that is quite not a liberty thay can take..something doesnt seem to add up though..
Anonymous said...
I hope that K8L will finally bring back AMD to desktop performance game. Cause right now, Intel product that is 5 generations behind AMD is beating the sh!t out of K8. Scare to even imagine if Intel somehow managed to add some of the "generation" stuff like IMC.
Yeah.
Well, be sure and let us know when that is,... cuz I don't see any of these supposed chips in the stores, or online, or anywhere besides a few, well-connected enthusiasts.
When these chips are available to the general public, and have been reviewed by the mainstream hardware sites in a fully disclosed environment, then you can talk trash about Intel's Core "beating the sh!t out of K8." Until then, you sound like a sad, sad fanboy.
How long now has AMD had the performance lead? Oh yeah,... ever since K8 arrived? We shall see if Intel is able to do the same, once conroe actually arrives.
How much bandwith will the K8L have? HTT 5.2 Ghz?
Looks like AMD is following intel here..
intel already has the TCP/IP & SSL acceleration with Speed-step technlogy in their current processors & chipsets
Intel has the RAS features from long time...now AMD is trying to do some thing in that space...Intel had a lot of design wins beacuse of RAS features which are critical in high-end server/ HPC environments
intel showed off the XML engine chip in last year IDF.. they are going to integrate it into their next gen chipset...
as far as the AMD speculation of 43% power advantage goes.. it's pure speculation
Its just a presentation. Still too early to speculate its performance (and power consumption). Also when se will see it MATERIALIZE. AMD haven't even completed move to 65nm YET while Intel has already (the Core Duo and newer EEs). By the way, that die pix (chip) slide thingy doesn't even look like an actual silicon.. (finished or unfinished) More like something patched up using Photoshop. AMD should show some REAL silicon at least (even if it can't provide any chips or test rigs). Early 2007 sounds too soon, by the way.. we'll wait and see...
Post a Comment
<< Home