Phenom Models/Clocks/Cache and TDP Announced - Launch Q3 07

[Kamakazie!]

Dailytech just posted an AMD slide giving details of the launch models for the next generation Phenom and Athlon models.

Article

I just hope i can drop that Phenom X4 in to my AM2 motherboard like they said i could. 89W TDP!!! So thats more like 65W under normal usage. These may just be worth the wait and overclock nicely to boot.

 

[Emlyn_Dewar]

Yeah, I'm hoping that Asus will provide a bios update for my board, would be awesome to be able to chuck in a new cpu.

 

[NathanE]

Still no benchies

 

[riddlermarc]

AMD’s dual-core Kuma processors will carry the Phenom X2 name and drop into Socket AM2+ and AM2 motherboards.

Definite bonus if just a bios update will be all that's needed for most/all AM2 motherboards, but we're looking at Q4'07 for that according to the graphic

Some further interesting info:

quote:
but are these older AM2 boards going to severly cripple the performance of Stars processors?

I highly doubt it. The only time this should have any effect at all is if the current Hypertransport bandwidth (4.0GB/s in either direction) is insufficient for your I/O. The one time that it's likely to have a noticeable (easily 10%) effect will be on system boards with integrated video when playing 3D games. Such a setup sends a LOT of data between memory and the motherboard chipset and right now Hypertransport is a bottleneck (4.0GB/s of bandwidth vs. up to 12.8GB/s from main memory). There may also be a TINY difference in really high-end 3D video setups with add-in cards, particularly if you're talking about Crossfire or SLI setups. However considering that a single PCIe 16x video card is limited to 4.0GB/s in each direction as well, HT 1.0 isn't much of a limitation here, we're probably looking at less than a 5% difference in performance.

The real advantages for HT 3.0 will only be realized in server and I/O intensive workstation applications in the near future. If you've got a 16+ 15Krpm SCSI disks in multiple RAID arrays and dual 10-gigabit Ethernet adapters, then HT 3.0 will be a godsend. For desktops though there just isn't that much I/O and Hypertransport 1.0 already provided a LOT of bandwidth.

 

[mosfet]

I can't helping thinking that the momentum will be all but lost by the time these are released.

Seriously, a proper Conroe competitor more than a year later? Still on 65nm? Doesn't really help AMD in their less-than-admirable situation in competing with the 8000 series, which has been out for what, 8 months already?

I hope that this is the chance AMD have been waiting for to pull their finger out, I hope that the R600 and the new X2 and X4 cores can provide some meaningful competition. Not just for the AMD buyers, but for the whole market, because without a credible competitor the market stagnates, which only damages us - the people that want fast, cheap and available CPUs.

 

[snazzy_dazzy]

AMD will be hoping that intel will get complacent again.

However I think the Pentium/ Athlon fiasco was the kick up the butt that Intel needed, and that for the next couple of years they will be owning due to vastly superior staff and resources

Hell they cut 11,000 jobs. That was the clutter on their hard drives that was slowing them down. I bet the rest have been told to get into shape and its running a lot more efficiently now

 

[schumi84]

It seems that AMD and ATi have lost their way a bit, releasing new product lines that are at least a year behind their closest competitor.

It seems they both have got a bit complacent with being ahead so need this kick up the rear to get them back on track.

 

[NathanE]

It seems that AMD and ATi have lost their way a bit, releasing new product lines that are at least a year behind their closest competitor.

It seems they both have got a bit complacent with being ahead so need this kick up the rear to get them back on track.

To be honest though it did take Intel a good 2, almost 3 years, to properly respond to K8 with the Conroe

 

[Jaffa_Cake]

To be honest though it did take Intel a good 2, almost 3 years, to properly respond to K8 with the Conroe

The conroe smokes the K8, to keep up tradtion the Phenom will have to smoke the conroe.. I don't see that happening easily.

 

[mrthingyx]

AMD managed to do it in practically every market sector with one processor and its derivatives.

As NathanE said, they kept it for 2/3 years while Intel was wondering how to kill off Netburst.

 

[simonnance]

hopefully the time spent getting the new architecture together will have been well spent.

The Phenom Vs Penryn/Nehaem battle will be interesting.... Penryn is a die shrink with some minor tweaks, and Nehaem seems to be Penryn with an on die memory controler....... Whereas Phenom looks a damn impressive step up from A64.

89W TDP on quad cores? 2MB L3 cache? All 4 cores on the same die?

Could be in for some interesting times......

 

[Kamakazie!]

89W TDP on quad cores? 2MB L3 cache? All 4 cores on the same die?

Could be in for some interesting times......

Aye, and as i said, that is 89W MAX TDP. In Intel terms it is something like 65W or less. Imagine the low power versions of these.

The conroe smokes the K8, to keep up tradtion the Phenom will have to smoke the conroe.. I don't see that happening easily.

Lets not get carried away here. Clock for clock is it 10-20% better, there or there abouts, than a 3-4 year old architecture. It is the fact that is overclocks so well where the big benefit comes.
AMD have been touting an 80% improvement for the K10 in some areas. Obviously this will be mainly in the encoding side with the new SSE instructions and better SSE execution but still.
Lets hope though, like you say, that the Phenom is a belter.

 

[NathanE]

I think the Phenom will probably "smoke" the Conroe by a large margin. But unlike last time Intel don't have to completely change gears... they just have to make a few tweaks to overtake again and those tweaks are already in the pipeline with the Nehalem core.

 

[Corasik]

AMD have been touting an 80% improvement for the K10 in some areas. Obviously this will be mainly in the encoding side with the new SSE instructions and better SSE execution but still.
Lets hope though, like you say, that the Phenom is a belter.

Wasnt that 80% quote a reference to the performance of a Dual Core Opteron V a Quadcore K10, running floating point instructions (K10's reputed to have a pair of FPU's per core instead of 1).

Pretty sure the AMD will have the best performance in floating point, but for Integer, and SSE they will have a tougher time really beating down on intel.

+ as NathanE says, Intel are ready this time, with many new ideas in the pipeline, including stuff like CSI (Same as Hypertransport more or less), Onboard memory controllers, 45nm process, and the return of Hyperthreading!

Intels NGMA has plenty of life left in it yet.

 

[simonnance]

shouldnt the SSE128 help with catching up with Intel on that front?

 

[drunkenmonkey]

i reckon these chips are going to fly and be a massive jump over k8 AND conroe + they are absolutley gonna smoke intels tacked together with duct tape quad cores

plus i think amd are gonna pull one of intels tricks and stick 2 quads together for the first octo - core chip!

looks like its starting to get interesting again!

think il be waiting for a native quad core if the benchies show some good results

 

[deadsquirrel]

too much smoking going on in this thread for my liking... cough...

 

[Kamakazie!]

plus i think amd are gonna pull one of intels tricks and stick 2 quads together for the first octo - core chip!

looks like its starting to get interesting again!

think il be waiting for a native quad core if the benchies show some good results

At an 89W TDP for the quad cores, i highly doubt it. Maybe if they can get the TDP down to 65W in the low power models. Still, i don't think we will see it till 2H 08 when 43nm comes about.

Intel will no doubt do it before that with Nehalem, being that it is a true quadcore.