• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

amd fightback 2007??

dafloppyone

dafloppyone

dafloppyone

Associate
Joined
15 Jun 2006
Posts
2,178
Location
Amsterdam
AMD shows off details of K8L

Spring Processor Forum Details, details, and pics

By Charlie Demerjian in San Jose: Tuesday 16 May 2006, 18:50
IN HIS KEYNOTE, AMD's Chuck Moore basically laid out the K8L in slightly more detail than we did. Either way, here are the highlights.

First, it has a shared expandable L3 cache, necessary because it is a native quad-core design. The one massive enhancement to the mix is that AMD finally has the ability to independently change core voltages for power savings. It now can also change the north bridge voltage independently of the cores. This is a huge win, we are told voltage differentials and problems with them were one of the main scaling headaches of the K8 core to this point.

AMD K8L die shot

Next is memory. The new core will support 48-bit addressing and 1GB pages. Cray and SGI will be very happy with this, until they hit that memory wall again. There is also official co-processor support, strongly hinted to be on a HTX card. The key here will be the platform is aware of them vs having to hack them in.

The other whopper Chuck dropped was that DDR2 is coming and DDR3 is in the wings when the spec 'settles down'. Old news, FB-DIMMs are the future, right? AMD has said they are supporting them, but the big news is that they are not forcing support. Unlike Intel's approach, Blackford supports only FBD, AMD will let you choose. This seems to strongly suggest that the controller on the later gens will be quite flexible indeed.

Next up is RAS, another area where AMD is sorely lacking. It is addressing the major sore points with support for memory mirroring, data poisoning support, and HT retry. It looks like it is following the IBM roadmap more than the Intel one here.

IPC is also going up in a big way. It is doing the obvious doubling of SSE/FP resources, old news now, but it goes a lot deeper than that. There are a bunch of added instructions, starting with the bit manipulation instructions LZCNT and POPCNT. It also added SSE extensions EXTRQ/INSERTQ and MOVNTSD/MOVNTSS. No word on SSE4 though.

The last bit is much more aggressive prefetch to 'feed the beast'. It has gone from 16B to 32B, an obvious step with the added SSE number crunching power. On top of this, it has out of order loads, and other tweaks to use the available bandwidth in a much more efficient manner.

For those who thought K8L was more or less a tweaked K8, you are wrong. It looks like no part of the core has been left unmolested by the elves working the CAD stations. It looks like AMD will have a credible response to the Intel MCW architecture after all. 2007 will be a fight after all. µ
http://www.theinquirer.net/?article=31761
 
I'm still going to buy a new Socket 939 chip before I even think about a new platform, So in 2009 I may go Intel lol.

Will there be any new releases for socket 939 or am I stuck getting an X2 4800+ when the price drops?
 
Time and again people forget that K8L is a high end server/workstation part. It won't see the light of day on desktops.

It's good to hear AMD have finally acknowledged their prefetching problems (hence their reliance on low memory timings). Hopefully one day those changes made to K8L will filter down to the lower consumer parts.
 
Last edited:
Reality|Bites said:
I'm still going to buy a new Socket 939 chip before I even think about a new platform, So in 2009 I may go Intel lol.

Will there be any new releases for socket 939 or am I stuck getting an X2 4800+ when the price drops?
Click to expand...

Don't think so. I think they'll be pushing on AM2 for quite a bit yet.
 
he might be right that K8L won't see the light of day, but i bet a desktop adaptation of it does, then conroe won't be picking on a rather old architecture anymore and will proberbly have no lead at all, conroe fanboys forget when there raving about how good conroe is, that there compairing it to an oldish architecture in K8, soon as AMD bring out a new design, im sure it'll be much more competative
 
Gashman said:
conroe fanboys forget when there raving about how good conroe is, that there compairing it to an oldish architecture in K8, soon as AMD bring out a new design, im sure it'll be much more competative
Click to expand...
I hesitate to say this, but "we" don't forget that at all. It's just that the current AMD roadmap shows no direct competitor to Conroe... so what are we meant to compare to? K8L and that 4x4 thing are totally different.

FrankJH said:
THats what people were saying about opty's a few years ago
Click to expand...
Yeah... I'm sure eventually it will get there. But at least for the first year of its life it will not be on any desktop. Dual core has barely even got started, let alone quad core. It'll be a specialty reserved for really expensive workstations and servers. I won't be at all surprised if K8L chips start from £1500. These will be pushed as super mega high end parts. AMD won't be selling off their gold dust to consumers. It'll be going to the corporate/SME markets where they have really done well for themselves these past couple years.

Any "desktop adapation" like what Gashman suggested is very likely just going to be the binned chips which have 1 or 2 failed cores. Because yields on K8L will be quite bad (massive wafer size, and yet another reason why these things aren't coming to the desktop any time soon)... so they'll disable the dead cores on those chips and sell 'em off to consumers. So yes it's "possible" a variation of K8L could reach the desktop, bringing the advantages like better prefetching as mentioned in the OP. However, again, this isn't on the roadmap so it is just speculation.
 
Last edited:
NathanE said:
I hesitate to say this, but "we" don't forget that at all. It's just that the current AMD roadmap shows no direct competitor to Conroe... so what are we meant to compare to? K8L and that 4x4 thing are totally different.

Yeah... I'm sure eventually it will get there. But at least for the first year of its life it will not be on any desktop. Dual core has barely even got started, let alone quad core. It'll be a specialty reserved for really expensive workstations and servers. I won't be at all surprised if K8L chips start from £1500. These will be pushed as super mega high end parts. AMD won't be selling off their gold dust to consumers. It'll be going to the corporate/SME markets where they have really done well for themselves these past couple years.

Any "desktop adapation" like what Gashman suggested is very likely just going to be the binned chips which have 1 or 2 failed cores. Because yields on K8L will be quite bad (massive wafer size)... so they'll disable the dead cores on those chips and sell 'em off to consumers. So yes it's "possible" a variation of K8L could reach the desktop, bringing the advantages like better prefetching as mentioned in the OP. However, again, this isn't on the roadmap so it is just speculation.
Click to expand...



You may be right but I personally think not, they will be used in workstations yes (and possibly even in dual cpu / 8 core workstations - however these chips will be relatively mainstream ( in comparison to Intel's EE chips) Once multithreaded apps become more mainstream - which is already starting to happen, the next wave of dual / quad core chips will be upon us and remember fabrication costs, it will be sooooooooo much cheaper for AMD to devote one line of a fab plant (or even one whole plant) to one type of chip - even if they then say 3/4 is for high end multi core (8 cores and over) cpu's and the other 1/4 of the fab plant is for desktop / small workstations
 
Conroe isn't that much faster compared to a similarly clocked K8 so there is little need for a spectacular response from AMD. Main thing is the K8 is still on 90nm and Conroe is 65nm which gives Intel the edge as far as price and performance per watt is concerned and obviously Intel has done what it can to ensure Conroe scales to 4 GHz and beyond.

AMD will get there eventually with their 65nm refresh but they'll still probably be 5% to 10% slower compared to an eqivelent Conroe but come on is that small difference really supposed to be a big deal? :)
 
str said:
Conroe isn't that much faster compared to a similarly clocked K8 so there is little need for a spectacular response from AMD. Main thing is the K8 is still on 90nm and Conroe is 65nm which gives Intel the edge as far as price and performance per watt is concerned and obviously Intel has done what it can to ensure Conroe scales to 4 GHz and beyond.

AMD will get there eventually with their 65nm refresh but they'll still probably be 5% to 10% slower compared to an eqivelent Conroe but come on is that small difference really supposed to be a big deal? :)
Click to expand...

that would be fine if they really were only 5-10% faster. they are not, they are a hell of a lot faster.

http://forums.overclockers.co.uk/showthread.php?t=17590588
 
Well it varies doesn't it. Some benchies it's like 5-10%, but others it's like 40%, and in the extremes even 100%... I think the average is something like 25% though.
 
james.miller said:
that would be fine if they really were only 5-10% faster. they are not, they are a hell of a lot faster.
Click to expand...
Sure it's been shown that Conroe is currently around 10% to 20% faster than the current 90nm K8 in most things but so what that's still not that big a deal. As for K8 on 65nm, it'll likely be tweaked a little for extra speed, hence why I said probably will be 5% to 10% slower than the equivelent Conroe.

The big deal is the fact Conroe is showing itself to scale way beyond 3 GHz and that's what makes it appealing compared to the current AMD K8 90nm CPUs so yeah Conroe is great because of that but the same will likely be so for AMD's 65nm refresh so AMD won't exactly be struggling to keep up. :D
 
str said:
Sure it's been shown that Conroe is currently around 10% to 20% faster than the current 90nm K8 in most things but so what that's still not that big a deal. As for K8 on 65nm, it'll likely be tweaked a little for extra speed, hence why I said probably will be 5% to 10% slower than the equivelent Conroe
Click to expand...

its not 10-20% either. in some situations its 50%, in the extremes its 100%. like those fear benchmarks. This isnt something AMD can fix with moving to 65nm im sure. they NEED a new core design. if a 3ghz conroe is 90% faster than a 3ghz opteron when running f.e.a.r @ 1600x1200, how far will the a64's have to scale on 65nm to remain competitive?

if 20% was all AMD needed, then i guess that means they'd need to be runinng around 3.7ghz-3.8ghz just to run alongside a 3ghz conroe (assuming 20% performance = 20% clockspeed). That's not an easy feat, but what makes it worse is that we are already seing conroes reach those speeds. Again, that's if the differences were only 20% across the board.

I own an opteron, and have only owned AMD's for a long while. the last intel i had was a 1.2ghz celeron-T. Dont get me wrong i repect AMD a great deal, but they need either to pull something out of their sleeves, or get their heads down and start working hard.
 
Last edited:
I didn't know Conroe was 90% faster in fear. Got a link to the benchies? Mostly I've seen around 10% to 20% in the previews although in some cases it can be higher.
 
Yup scalability is the key. Even if AMD launched something to match Conroe... It would take Intel about a month to react by increasing the clocks and/or FSB. However Conroe scalability on the multi-core front is still very very very poor. All of Intel's upcoming quad core chips require a dual FSB. Ludicrous. AMD's Hypertransport literally takes an engineer about 5 mouse clicks to add another core or interconnect to another CPU socket. With Intel's current FSB design it takes months of work by a whole team of engineers.

Personally I feel Conroe is _still_ a bit of a stop-gap solution (though obviously no where near on the scale that Smithfield was) for Intel. You get the odd person saying how they admire the Conroe architecture so much. But really it is nothing special. All they've done is taken the Pentium-M, put it onto a next-gen fabrication process, added a ton of new execution units, increased cache sizes, improved prefetching, improved ILP (macro/micro-ops fusion), and a few other logic tweaks. Nothing special there really. It's basically what they use to do back in the good old Pentium I to III days.

I don't believe Intel will be fully back in the game until they have a Conroe-derived chip running on a CSI interconnect. None of this ancient FSB crap.
 
You must log in or register to reply here.
Back
Top Bottom