• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

AMD Roadmap 09-10.

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Kuroimaho

Member
Joined
May 26, 2004
Location
Japan, Tokyo, Ueno.
Watch.impress compiled a roadmap for AMD again. Link

The article is in Japanese but the images tell a tale, not much but still worth to take a look.

The native dual with 1MB L2 is certainly an interesting move (in the picture quad is written but that's a dual).
 
The maps are optimistic and great so I hope AMD pulls it off. Why I think 2011 seems a little too soon but assuming that the NY FAB is operational in mid 2010 then it might be doable.
 
Well, I hope that they stay afloat and have the capital to pay for the building. If they can get the ny fab with 32nm they will be golden. But, until then... Intel will just start to get a head. IMO, Amd really caught up to intel with the release of the PII. They did so much better with these compared to the PI.
 
Caught up...almost. They have caught up with 45nm Yorkfields, and now need one more step for core i7, and 32nm will be here soon. Amd seems like it will be one step behind intel, but at its price/performance it is in an ideal position for the mainstream market who, for intel's side will stay on 775 for a while before core i5 comes.
 
Yeap AMD is covering the gap, would be nice to see more competition. I think 12 cores are far in the future for now.
 
Looks like the 6 core monster is reserved for workstation and server. That makes sense since I hardly use 4 cores let alone 6. I am very intriged by the dual cores with 1MB L2 and the Quad core with 0 L3 cache. Both are interesting as an alternative for simple home PCs.
 
I say Amd just goes right from 45nm to 21nm. That would really make Intel poop themselves.

It isn't that simple, it costs huge R&D, and yet the economy isn't as good so it wouldn't be wise to drop that low so fast. Beating Intel to 32nm or at the same time will be good enough, although intel is already ahead on that. Mainstream is where most the money is, and they can keep their prices though and make money now then there isn't much point in jumping that far.
 
I'm seeing a lot of PII's around lately. Still not seeing a ton of i7's, even on enthusiast forums! Intel is still clearly the speed champion, but that is far from everything in todays economy.
 
I'm seeing a lot of PII's around lately. Still not seeing a ton of i7's, even on enthusiast forums! Intel is still clearly the speed champion, but that is far from everything in todays economy.

That's right. An i7 setup is really a stab in everyones pocket, even for the enthusiast.
 
I want to clarify that it was meant a joke when I said shrink right to 21nm. That said, I agree that intel is ahead on the 32nm race as the ny Fab hasn't begun building yet... On a good note, I like to see things built in the US!!!
 
That's right. An i7 setup is really a stab in everyones pocket, even for the enthusiast.

i7 isn't too expensive, it is the motherboard which are slowly getting cheaper. Hopefully the price will drop so that it is more affordable. Core i5 is more promising though, and looks to be a competitor to the phenom II, and core i7 for the enthusiast market. I do not regret my build at all :)
 
No doubt, I wish I had the cash burning a hole in my pocket for 3x2gb ddr3, an i7, and a good i7 mobo. That's 600 minimum all together though, with the cheapest poorly reviewed mobo and cheapest ram. Had I had the money I probably still would have gotten the phenom II, with say a 4870X2.
 
If I have the dough, I would get a PII and an nvidia 780a chipset and run three 285's. But that's just me and I'm into the whole sli thing.
 
The maps are optimistic and great so I hope AMD pulls it off. Why I think 2011 seems a little too soon but assuming that the NY FAB is operational in mid 2010 then it might be doable.

Take a look at this, I am quite sceptical about it but backs up the roadmap on fabrication process. Link.

“The next-gen SOI 32nm bulk fabrication tech will be available for AMD, the main consumer of the Globalfoundries, sometimes in H2 2009. Bulk 32nm manufacturing technology should become online even later,” a high-ranking statesman said.

I wonder what are they going to do with 32nm SOI if AMD does not use it until 2011, especially why use it sooner than bulk with ATI and NV standing in line for a good 32nm bulk process.
The roadmap also shows with purple color that AMD will be able to use 32nm process end of 2009 and throughout 2010. So who knows a surprise attack ?

Well, I hope that they stay afloat and have the capital to pay for the building. If they can get the ny fab with 32nm they will be golden. But, until then... Intel will just start to get a head. IMO, Amd really caught up to intel with the release of the PII. They did so much better with these compared to the PI.

Mubdala will pay for it not AMD and if the above said statement is true they better go for 25nm in the sates.

The 950 will catch up with the top end cpus of Intel's last gen and will come about 1Q earlier than Intel's mainstream Neha. So if the roadmap is true we get a new stepping slightly higher clocks from AMD while Intel threws out a new mainstream arch and shrinks it to 32nm and maybe even Sandy Bridge reaches the shelves. Where would that put AMD ? Close to 2 gen behind Intel ?
 
Only one gen :p, Yorkfield will only be one gen old then, and AMD will be roughly on par (QX9770 is 3.2Ghz, and not to mention overclocking ...). i5 Will gain I am estimating 10-30% in performance depending on the type of task, multithreaded will be faster due to 8 threads. So intel will still be ahead, and looks like for the meantime always will be.
 
I just hope Bulldozer arrives before Sandy Bridge, and performs at par or better. Sandy Bridge is scheduled for Q4 10, but it may be shifted to H1 11 to compensate for shorter Westmere cycle. Plus Intel doesnt save costs by switching to a new architecture in these times.

Phenom II might look like revival and may offer better cost/performance ratio, but thats only for the customer. The company is actually having to sell them for less to maintain cash flow.
And Core i5 looks promising. From what I've seen it performs ~90% of Core i7 clock for clock. And Intel decision to delay i5 by another quarter just confirms that it wants to drive more and more crowd into the uber i7 camp which might otherwise consider i5.

AMD might just be competitive enough, but I want some real good stuff frome them, not silly low priced stuff that eats huge power and must be run at 3.8 Ghz to match the rival at less.

I say Amd just goes right from 45nm to 21nm. That would really make Intel poop themselves.
There's probability than 32->22 might not be possible in a single jump(intel docs) and another gap might be needed (26?), leave alone 45->22

The native dual with 1MB L2 is certainly an interesting move (in the picture quad is written but that's a dual).
Umm, you think its going to be better than Intel Wolfdales ? I think it should atleast compete at the 100 $ pricepoint if not higher. Or are these 45W HTPC types ?

That's right. An i7 setup is really a stab in everyones pocket, even for the enthusiast.
No, its just than for the price of super high-end Dragon (PhII X4 BE+ 790FX) you get the lowest i7 which performs better :p
 
Only one gen :p.i5 Will gain I am estimating 10-30% in performance depending on the type of task, multithreaded will be faster due to 8 threads.

1. Neha. 2008
2. Sandy bridge 2010.
3. Bulldozer 2011.

You can also add a few more % for the I5 next year after it comes from 32nm.

Plus Intel doesnt save costs by switching to a new architecture in these times.

If it has higher performance it means it can be sold for more than the competitor's product, so regardless of the economy being ahead of AMD can pay off.

Umm, you think its going to be better than Intel Wolfdales ? I think it should atleast compete at the 100 $ pricepoint if not higher.

Which wolfdale the 5K or the 7K or the 8K ? ;)

With the Deneb based X3s selling for 140-120$ Propu(o)s should be between 150 and 50.
So these L3 less duals should go as low as 50$ while the Deneb based ones closer to 100, with this rough price estimation you can match them to wolfdales.
4G capable 50$ semprons sound like fun.
 
If it has higher performance it means it can be sold for more than the competitor's product, so regardless of the economy being ahead of AMD can pay off. .
I think high performance products only target enthusiast segments that have a small market share. Intel will try to hurry to 32nm because switching to smaller process will save money in the long run. Unless AMD can hurry up Bulldozer, Intel can easily relax with their 32nm mainstream parts.

Which wolfdale the 5K or the 7K or the 8K ? ;)

With the Deneb based X3s selling for 140-120$ Propu(o)s should be between 150 and 50.
So these L3 less duals should go as low as 50$ while the Deneb based ones closer to 100, with this rough price estimation you can match them to wolfdales.
4G capable 50$ semprons sound like fun.
I assumed E7X00 series.
$50 is too low, is that a nettop product ? (Intel is disallowing Ion, so why not put in AMD ;) )
 
Back