Details Leak on AMD's Upcoming Hexa-Core Processor

Details Leak on AMD's Upcoming Hexa-Core Processor

There's been an invisible hexa-core processor lurking on AMD's desktop roadmap ever since the company launched Shanghai on 45nm, but the company has refused to confirm or deny that such a chip was in the works—until now. As of today, AMD's hexa-core desktop processor—codenamed Thuban—is officially on the board. Exact launch dates aren't yet available, but we've got a few details on the core architecture. For those of you who care about this sort of thing, Thuban is a star in the constellation Draco and was once a pole star roundabout 3000 BC.(Six inch clear platform heels are much older than anyone realizes -Ed).

Thuban, like Istanbul, will be a monolithic core and will utilize the same 45nm process as its cousin in the server space. In a surprising move, AMD has apparently confirmed that the upcoming processors will be drop-in compatible with AM3, AM2+, and AM2 motherboards. If that turns out to be accurate, Socket AM2/AM2+ owners that first bought boards when 90nm dual-core Athlon 64 X2's were all the rage will be able to upgrade to 45nm hexa-core processors that could well occupy the same power envelope. The Phenom II X6, as it's expected to be known, will reportedly carry 3MB of L2 (512K per core) and a 6MB unified L3 cache. While this has yet to be confirmed, it's probably accurate; the L2/L3 cache configuration on Thuban is identical to both Istanbul and Shanghai. AMD isn't spilling the beans on any of its launch frequencies at this point, but a glance at the company's current product line, plus a little educated guessing, can get us pretty close.

Istanbul, Constantinople and now Thuban

Based on the company's current lineup of Phenom II, quad-core Opteron, and hexa-core Istanbul processors, Thuban will likely debut at or around 3GHz, with a TDP between 95-125W, depending on how mature the 45nm production lines at GlobalFoundries are by then. Scaling Thuban will be tricky; there's no getting around the fact that more cores = more heat. Expect a relatively low frequency ceiling compared to the Phenom II X4; it'll be surprising if AMD pushes these chips above 3.2GHz; 3.4GHz would likely knock its head on the 140W TDP ceiling.

It's not immediately obvious which Intel processors Thuban will target. Based on what we've seen from our quad-core Nehalem vs. Shanghai match-ups, Intel's ultra-high end, hexa-core Gulftown (Core i9) is virtually certain to pummel Thuban in any head-to-head competition. Knowing this, AMD will likely target Thuban against Intel's Core i7 series, and possibly the upper-end Core i5s as well. In such comparisons, the extra two cores will definitely make a difference. HyperThreading is one of the main reasons the Core i7 takes Phenom II to the mat time after time, particularly in SMP-friendly tests. An extra two cores won't work magic, but it'll put the two processors on much more even ground. AMD could theoretically attempt to press Intel's Core i5/i3 processors with lower-clocked Thubans, but the size and complexity of the die (346mm2, 904 million transistors) make this unlikely. As with Shanghai, Thuban will have to balance itself on a relatively narrow price/performance tightrope.

We could launch into a discussion of how many people actually need six processors, but that horse done been beat before.
0
+ -

Encoders might like the extra cores.  For me, I'll stick with 4.  45nm isn't bad, but Intel should be moving onto 32nm very soon (they're showing it off to the press right now).

It'll probably have a high retail price since 6 cores will provide great performance when utilized, so could AMD once again actually sell a CPU above $250 after all of these years?

0
+ -

I like the idea that it might could be dropped into an existing motherboard. That's a selling point that Intel just can't match.

I just have to wonder why they went with the Greek rather than with the Latin prefix, as has been done with 4 (quad-) and 2 (duo-)? OK, they used the Greek also for 3 (tri-), but how about some consistency? I say we make it all Latin, and talk about sexacore processors. ...I'm very lonely.

0
+ -

Still, who is their target market? Hexacores is great for F@H, but for gaming, frequency matters more.

0
+ -

I think the big target for AMD right now is to try to stay within one generation of Intel and IBM in all markets.  AMD Hexacore target audience has to be businesses since most other non-server markets would not benefit.  More money is made there anyway.  Outside of the synthetic bench, I imagine AMD's and Intel's hexacores will be fairly comparable otherwise.  The big thing to watch for is getting over the "memory barrier" for CPU's is the next big hurdle.  CPU cache dealyed this issue but it is quickly looming now... I wonder who will solve the issue first...

0
+ -

ClemSnide:

I like the idea that it might could be dropped into an existing motherboard. That's a selling point that Intel just can't match.

I just have to wonder why they went with the Greek rather than with the Latin prefix, as has been done with 4 (quad-) and 2 (duo-)? OK, they used the Greek also for 3 (tri-), but how about some consistency? I say we make it all Latin, and talk about sexacore processors. ...I'm very lonely.

Well I think it might be because sexa- would make all the kids in school giggle.

And that using quinque- for 5 instead of penta- would be ridiculous. Quinquecore? I don't think so. lol (Quintuplet & quintet are the only words I know that use it.)

0
+ -

I'm liking my Phenom's four cores just fine right now. If they develop games that can USE 6 cores, then I'll join the fun.

The new Radeon HD 5870 excites me allot more than 6 cores at this point.

0
+ -

Gibbersome,

The idea, as far as I can tell, is that program developers of every sort will magically start creating software to use these cores, now that everyone has them.

Give it a year. Or probably a decade.

Login or Register to Comment
Post a Comment
Username:   Password: