AMD Denab

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar


*plays drums on desktop* Yaaah toast...... French toast!!

Depends on where you look, the digitimes things says the AM2+ denebs will be out Q4 with an embargo on them till Jan 9th. If they already have the ability to get them out, then they need to just release the AM2+ chips and try and see how well they do against current chips. But the AM2+ chips are only suppose to be a temporary stop gap till the AM3 chips release in March or april, can't remember the month, when the AM3 platform chipsets come out. What it is is the am2+ chips could technically work in the AM3 boards, since they're essentially the same chip IMC and all. But, socket AM3 has 2 less pins than socket AM2/AM2+. Which is kinda ironic since they said they needed the extra pin for DDR2 support back when AM2 came out. So Am2+ chips won't physically be able to fit in the AM3 socket, but, the AM3 chips will be compatible with Socket Am2+ dual power plane boards, but not the older AM2 single power plane boards.

And if it's ready to come out and being shipped to OEM and board partners I want to see at least some more recent ES speeds and benchies dangit.

 


Wasn't there a general consensus when this article came out two months ago that it was fake? I'm hesitant to believe anything from that Hardspell site given how many completely fabricated rumors they have started. (Nvidia GTX350 anyone?)
 


But isn't this the same reason you say Nehalem isnt worth it? Because in ST it shows 5-10% performance gains? Kinda strange logic.

Ocing well is yet to be seen. Last OC for Deneb we saw at 4GHz was using 1.6v. 1.6 very hot volts. I personally think its the arch itself that doesn't let it OC or clock higher stock. But thats still to be seen.
 


Indeed, people are hypocritical that is just nature and we're all included. Anyway, I thought the Denab was supposed to be a different fab style. We can't know about the arch because this is supposedly a new thing.

Atleast that was what I got from the AMD presskit.

http://www.amd.com/us-en/0,,3715_15503,00.html
 

hector_toast-1.jpg
 

If AMDs new arch, not just die shrink only gets 5-7% in gaming, and even is slower in a few games, it wont be worth it either. This is just a die shrink, tho it may not be a simple dumb one, a die shrink to a different node nonetheless. Going to a whole new arch, that requires new memory, new mobo and all more expensive than what you had, plus the cpu itself costing more, and it only gives 5-7% gains in gaming, and is even slower.... I dont care who makes that cpu, it isnt worth it to me
 


Mine is multi grain its more future proof... its that and some overdose of preservatives...
 


Nope Deneb is the same arch. Its like Conroe to Penryn. Same arch just enhanced and a shrink to 45nm.

Deneb is only a shrink to 45nm, more L3 cache and a bit of improvement. Other than that it is essentially the same chip as a Agena Phenom.
 


Problem is those gains of more than 7% for even Deneb that you are talking about are also in MT, not ST. I haven't seen much more than 10% performance improevments in a while and doubt there will be unless there is a magical new way to move data even faster.

Besides in gaming the biggest improvement will always come from a new GPU unless the game engine takes advantage of multicored CPUs like Lost Planet does.
 


Bah, my Toast is wonder bread toast, so that means it owns all with it's bleached, super preservative goodness..... 😉

That doesn't say much though, a lot of bench's show Nehalem being slower than C2Q in games.
 


Nope, I think you're confusing it with some other article, probably the one about stock 4GHz+ Deneb FXs beating 5GHz Kentsfields or whatever...

This article looks legit, although as I said earlier its on an earlier stepping so the performance numbers aren't final, but its a good guide to the general improvements we can expect from Deneb anyway.
 


It's called NDA's: Non Disclosure Agreements. Until those expire, anything is just a rumor, or an unconfirmed leak.

AMD is much better nowadays about not tooting their own horn prior to release. Deneb is expected to be better than B3 because it's a die shrink. It's not a new architecture like Nehalem. Intel's CPU will be better than Deneb, but that better is still relative as Deneb should be priced much lower. B3 and Core 2 prices should come down to as those processors move towards EOL.



I like toast with raisins and cinnamon.



Sometimes (gasp!), it's not all about gaming, even though games will support more cores in the future. Look at the article on gaming with an Athlon 64. Back when that was a top of the line AMD processor up against the P4 Northwoods, people said dual cores weren't necessary. Now, the dual core Athlon X2's in the article handle newer games much better than the old Athlon 64 4000+.

People do so much beyond simple business programs from Microserf or single threaded games nowadays that quad cores are good, just so long as they have decent clocks. That's what AMD needs to improve with Deneb.
 


The i7 920 costs $284, for all intents and purposes this will be the i7 chip most people will buy as very few can afford $500+ for CPUs.

Unless it absolutely stinks, I doubt Deneb will be priced 'much lower' than the i7 920. It should be competitive with Yorkfields (maybe slightly slower per clock) but close enough that it shouldn't fall below $200 unlike current Phenoms.
 


Total platform costs for AMD should be less. That will counter the inexpensiveness of the 2.66 gigahertz Nehalems. If someone doesn't go Deneb, then I'm sure they'll do quite well with a Core 2 at a similar platform cost. Initially, Nehalem looks to be an expensive build for budget gamers and video fans.
 


Yes, but it was implied. After all, CPU prices don't exist in a vacuum. I'm hoping that AM3 will have more tolerant DDR3 kit standards than Nehalem's picky boards.






 
I think the real improvement for Deneb will be clockspeeds, not IPC. AMD's 65nm process stinks IMO, they went lower but couldn't raise clockspeeds at all, even AMD's fastest processor, the Athlon 64 X2 6400+, is a 90nm model. I'm just hoping they do better at 45nm, get some 3Ghz Denebs out that do as well as the Q9550 at stock, and make some money. The AM2+ Denebs could be a cheap, but fast combination, since DDR3 is still so expensive (which is what I don't like about Nehalem, even if the processor is cheap, DDR3 is going to raise platform costs significantly over a Q6600 build).
 



It should be fairly tolerant, almost the same IMC that's in the current Agena core. If you look at certain programs like everest it shows current phenom's IMC supporting DDR3, but obviously the physical socket can't connect to it. Which leads to something I was thinking about earlier today.

I can kinda see the point in making the AM3 socket not pin compatible with am2 cpu's. As older AM2 cpu's that aren't based on k10 or later don't have an IMC that can support DDR3. That would effectively prevent anyone from plugging an old Athlon CPU into the socket and frying the whole setup. But, Agena/barcelona's IMC does support DDR3, so it's kinda a catch 22. I just wonder how many uninformed people will try and plug one of the am2+ denebs into an older AM2 board and fry em.

But back to the memory voltage tolerance. It should be fairly tolerant, the one of the Agena cores is fairly good as long as you don't go much over 2.3v on DDR2. Doesn't effect me much since I almost always pick the stuff with the best specs at the lowest voltage reqs though. Thats how I ended up with the DDR2 800 g.skill kit I had, and the current ddr2 1066 mushkin kit.
 



See that was the whole thing I was getting at. Its not that the Intel product was that much better, architecturally, it was that their product would withstand much more RPM than the AMD product. Now we can argue for years that, well the product is better because of the speeds it can attain, but that's like saying a mustang GT is better than a LS460 because of the higher MPH. Keep in mind I don't know if my auto facts are correct, i'm just tryin to make a point.

I'm thinking that if AMD can attain the same Ghz as the Intel product we'll be in business.
 


Low clockspeeds are part of the reason, but AMD is significantly behind in IPC as well. Yorkfield is some 15 - 20% faster in IPC, which combined with the clockspeed advantage are the main reasons why AMD is currently so far behind the performance curve. Deneb addresses both clockspeed and IPC, maybe not sufficiently to overtake Yorkfield but it should at least be competitive.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom