• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

AMD Next Generation Processor Technology Slides

More details coming in the follow-up story, hopefully Thursday. Wanted to get the slides posted as there is a lot of information there that speaks for itself and does not require a lot of naration or input from me. :)
 
meh.. one or two of those slides has already been floating around for quite a while, but more info is always good. =) I'm looking forward to this new stuff coming out. Hopefully AM2 socket?

EDIT: Here is the new Opteron codes----
Up to 1P = 12xx. Socket AM2
Up to 2P = 22xx. Socket F (1207)
Up to 8P = 82xx. Socket F (1207)
Click to expand...
 
w1retap said:
meh.. one or two of those slides has already been floating around for quite a while, but more info is always good. =) I'm looking forward to this new stuff coming out. Hopefully AM2 socket?

EDIT: Here is the new Opteron codes----
Click to expand...

Well as we all know, sometimes the slides we see "floating around" do not always contain information that comes to pass. This information will come to pass. That is what differentiates HardOCP from a rumor page. At least when we post a rumor, we make fully sure that it is framed as such.
 
I *really* like how the CPU will scale according to load--shutting down cores (instead of putting them into a tight loop) when CPU usage is low.

One thing that struck me is the huge amount of cache that will be on the chip. A 64k L1 cache for each core, a 512KB L2 cache for each core, and a shared 2MB L3 cache. That's over 4MB cache total. Granted, we're talking about quad-core, but that's going to be a big die, even with a 65nm process.

It will also be interesting to see what the power dissipation profile will be. Four cores running flat out (Folding@Home, anyone?) with 4MB cache is gonna need a pretty good heatsink.

On a side note, I've seen the idea of integrating a GPU onto the CPU, now that AMD has acquired ATi. What if, say, you replaced two of those cores with a fairly powerful GPU?
 
Render said:
This looks VERY promising :)
Click to expand...

It does look very nice, the question is WHEN will we see these?

Hopefully it will be a fast implementation like going from A64 to X2. NOT like AthlonXP to A64.

(not to be a troll, but Kentsfield is already basically done from the benches I've seen, I assume they're just doing quality / platform verification before they release it)
 
Any word yet on the support of socket AM2 in their next line of chips? I went with that in my current build because I thought it would give me some ability to upgrade when AMD's next lineup hit. Is this correct?
 
Doesn't the math seem funny on those slides with the "power states"? I thought 100% + (33%*3) would equate to an overall 50% power state, not 60%. Likewise, with 100% + 50%...that's only 25%+12.5% for a total of 37.5%, not the 45% they listed.
 
WHOAguitarninja said:
Any word yet on the support of socket AM2 in their next line of chips? I went with that in my current build because I thought it would give me some ability to upgrade when AMD's next lineup hit. Is this correct?
Click to expand...


Yes, these will be AM2 to start with. They will be directly compatible with any AM2 motherboards out today. I asked the question of AMD last week, if I have a server farm of AM2 socketed rack servers, I will simply be able to drop in new quad core?

The answer was yes. Same power and thermal envelopes.
 
Mav451 said:
Doesn't the math seem funny on those slides with the "power states"? I thought 100% + (33%*3) would equate to an overall 50% power state, not 60%. Likewise, with 100% + 50%...that's only 25%+12.5% for a total of 37.5%, not the 45% they listed.
Click to expand...

There are issues here that are not apparent, the most likely being the fact that that each core's cache is not shut down when the core itself is powered down.
 
I love hearing about new cores and new tech coming out! But games, for example, are only just starting to take advantage of 2 cores, let alone quad and more cores. For me (I understand most people use their PC's for many, many other things), gaming is my thing on PC's and since today we are only starting to now see multithreaded games for dual core, isn't having quad core chips just overkill and more of a "B/c We Can" kind of thing for the chip manufacturers?
 
Why all the excitement of all this power when most programs out there cannot take full advantage of multi core? ALso i remember the day when Windows was (is) licensed per processor......does a core count as a processor?

And yes to all of you windows haters out there i know linux and unix can do what ever with them for what ever money....my point is the other 99% of us out there running windows boxes may see a licensing issue
 
Interesting read, but I have a question: What does AMD 'taping out' their CPUs mean? I'm lost on that.

And just a general comment...Quad Core is tasty, but will it be very practical for gaming. I mean, I don't know of any games today that are even taking full advantage of dual core CPUs.
 
Doomrulz and Lamont Cranston: That was weird. Looks like we all had the same general thoughts here about going Quad Core...
 
Mohonri said:
I *really* like how the CPU will scale according to load--shutting down cores (instead of putting them into a tight loop) when CPU usage is low.

One thing that struck me is the huge amount of cache that will be on the chip. A 64k L1 cache for each core, a 512KB L2 cache for each core, and a shared 2MB L3 cache. That's over 4MB cache total. Granted, we're talking about quad-core, but that's going to be a big die, even with a 65nm process.

It will also be interesting to see what the power dissipation profile will be. Four cores running flat out (Folding@Home, anyone?) with 4MB cache is gonna need a pretty good heatsink.

On a side note, I've seen the idea of integrating a GPU onto the CPU, now that AMD has acquired ATi. What if, say, you replaced two of those cores with a fairly powerful GPU?
Click to expand...

CPU idea is great but what feeds the GPU? GDDR3 or now GDDR4 via 256mbit bus.

How exactly will you feed the CPU that unless the mobo supports a special slot that interfaces with the Hypertransport bus that has all those ram chips.
 
Lamont Cranston said:
Why all the excitement of all this power when most programs out there cannot take full advantage of multi core?
Click to expand...
Because it seems that both AMD and Intel are having a hard time making their processors fasters in terms of just adding more MHz. The parallel route seems to be the preferred way to increase power. Yes, there's a "Chicken + Egg" problem with this approach, but it is not too far fetched to see that future software may be developed with multi-core architectures in mind.
Lamont Cranston said:
ALso i remember the day when Windows was (is) licensed per processor......does a core count as a processor?
Click to expand...

Microsoft says: "A processor is defined as a socket".
 
Any rumors on price point? a $999 chip does me no good, unfortunately.. nor does it do most people any good.

Performance will be interesting too, to see if any of its enhancements are enough at the end of the day to take on Kentsfield in real-world applications.. if not, thats okay too, AMD got used to this game plan before: slash prices for the price/performance win.

Edit: Saw no other mention beyond it being an FX part first; I guess that by definition means its $1k. :(
 
I can't imagine a quad core making much of a splash in the consumer market, the massive majority of programs don't even handle dual core properly.

Servers and workstations could use the capacity, though it makes me wonder how much memory bandwidth would be needed to advert a bottleneck.
 
With all of these independant clock-generators, will overclocking become vastly more complicated?

Nitpick: the slides are dated August 21, 2005.
 
Hmm...so let me get this straight.

Kentsfield is connected via the L2 cache, through which the processors can talk directly to each other.
These new AMDs are connected similarly, only through the L3 cache, not L2 cache.

Correct? I wonder how this will work, doesn't L3 cache have a huge latency penalty? Is the L3 cache the only way these CPUs are connected?
 
Since its all ON CHIP, I doubt you should be worried about latencies. If you can notice a difference that minimal, you are a superhuman.
 
But.. since everyone is all worried about dual cores, quad cores, and so on... I'm not so concerned over how many cores. I just want everything to be optimized for that many cores, and 64-bit! Hopefully Vista will help impliment everything toward 64-bit so we can process the data more efficiently, then offer better integration toward multi-core equiped PC's. Its kinda funny with the Athlon "64" is still using 32-bit everything. The 64-bit environment on the consumer end is totally lacking.
 
brentpresley said:
Not to burst anyone's bubble, but all the slides there except the PLL slide are over 6 weeks old.

http://www.amd.com/us-en/assets/con...df#search="AMD Technology Analyst day slides"

This info was out on the AMD Technology Analyst Day June 1, 2006.
Click to expand...

Yes they are. I did figure it was a good thing to pull it all together to be a bit better placed for enthusiasts. The slides were pulled out of AMD's new northbridge presentation as noted.

I think the PLL/Clock/Power is THE big deal here now. Followup article coming that will answer some questions. :)
 
I tought that the Original Athlon and Opteron X2 already communicated through a direct connection? where as the Pentium D had to go through the Northbridge
 
x86LegacyLayout.gif
This should layout what is going one a little better. The cores in the legacy arch are connected via the FSB.
 
There we go. That picture perfectly illustrates what is wrong with x86. Thank you. Finally the computer scene is moving onto new architecture after 20 years. In my computer engineering classes we were trying to explain why companies stick with the same architecture for so long and don't move onto better stuff even if they already know how to do it. We came to the conclusion that most of it is a "game" for them so they can keep squeezing more money out of consumers without radically changing their lineup. They can just keep improving off of the same old crap, but basically you get nowhere in the end. You have to move on once you exhaust your resources and target audiance. Althought it seems like computers went really far in the last 10-20 years.. everything has remained the same but just got a little faster. (in some people's eyes, a lot faster --in reality, not really) I'm glad I am born when I was though.. its going to be fun to watch computers improve over the next 50+ years. Assuming I don't die sooner.. lol. Just in the last 10 years, look what they did. The industry kind of got off to a slow start, but in recent years, everything is being done with computers. Now we can't live without them.. or it would be really hard not to. The industry absolutely needs them. Case in point, this is a good start. :p
 
I cannot wait for a Quad Core FX running on a 4X4 system for a total of 8 Cores....


Oh man, I'm droolin now :)
 
InorganicMatter said:
Hmm...so let me get this straight.

Kentsfield is connected via the L2 cache, through which the processors can talk directly to each other.
These new AMDs are connected similarly, only through the L3 cache, not L2 cache.

Correct? I wonder how this will work, doesn't L3 cache have a huge latency penalty? Is the L3 cache the only way these CPUs are connected?
Click to expand...

No they communicate through the crossbar switch. But they also now have the ability to share cache in those rare occasions when it may be required.
 
Would it be possible for AMD to sell a three core chip? I was just thinking what they would do with the four core chips that were found to have one bad core, but the rest were fine. Maybe these would be their mid level chips ($300 to $600 range) with the quad core chips for the ultra high end and duo core chips on the low end.

I think the CPU and GPU and the same chip designs will have to wait for 45nm. I will be interesting to see how they cram all that onto one chip.

It will also be interesting when "Recommened Number of Cores" starts showing up on system requirement labels for games.
 
Ockie said:
I cannot wait for a Quad Core FX running on a 4X4 system for a total of 8 Cores....
Oh man, I'm droolin now :)
Click to expand...

And so are Power Supply makers and your local power company :p
 
Computers are ran like this.. ever notice a pattern?
1-2-4-8-16-32-64-128-256-512-1024 etc.
;)
 
I'm interested to see what impact the ATI acquisition has on the upcoming quad core implementation and eventual move to the AM3 socket, if any. :D



 
jarman said:
I'm interested to see what impact the ATI acquisition has on the upcoming quad core implementation and eventual move to the AM3 socket, if any. :D
Click to expand...

ATI will be utilizing AMD's enormous inventory of patents and CPU technology to develop a CPU socket on their graphic cards completely replacing the GPU.

ATI's R610 chip will be using Kentsfield quad core running at over 4x industry standard GPU core speed of 2.8ghz per core. AMD is coming out with a 16X16 mother board which will have 4 cpu sockets running 4 cores each and 4 GPUs running in crossfire using 4 cores each. It will be a paper launch and will most likely end up in store shelves by next year when intel releases a 16 core cpu.

Haha, but seriously, AMD's new cpu looks awesome.
 
You must log in or register to reply here.
Back
Top