News AMD's AM2 - Conroe to get a free ride on the desktop?

And yet you clearly cannot grasp the concept of architecture revisions making a significant change. Willamette to Northwood is our example today. The performance improvement was comperably massive. AMD isnt sitting on its laurels, and will have an answer of some sort - if not at launch, later on in late 2006 or early 2007. Dont forget that Conroe doesnt launch until Quarter three, which is still five months out.

 

Let's not start a DDR2 Speed/Latency war. I've played with ENOUGH of the damn stuff to tell you over time that in essence, PC2-6400 @ 4-4-4-10 is the performance equal of PC3200 @ 2-2-2-5. You're looking at double speed + double latency, so the problem goes away. Now, considering that both platforms sit at the performance dawn of DDR2 (after all, how many Uber performance DDR1 kits did you see when only AMD toyed with it ). Right now, the fastest DDR2 modules you can buy are Corsair's 1066ers (of which i have a set). That's just right now. I wouldn't be surprised to see DDR2-1066, 1GB Sticks, at CAS4, by year's end. Maybe Q2 '07 at the LATEST.

After all, Intel is pushing up 1333MHz FSB chips, and companies WILL produce DDR2-1333 to go with it. This memory will, of course, start at the traditional (read slow ass) 5-5-5-15 timings. Now just where do you think the chips that can't bin that high of a clock will end up ?

 

yeh, i was saying

as you said, "PC2-6400 @ 4-4-4-10 is the performance equal of PC3200 @ 2-2-2-5."

So basically, ATM upgrading to DDR2 doesnt give any memory speed increases really

the higher latencies cancel out the higher clock speeds

 

Yes, but then you forget.

DDR2-1000 @ 5-5-5-15 is = to DDR-500 @ 2.5-2.5-2.5-7.5 lol.

 

huh

what relevance does that have

 

thaat DDR2-1000 performs better than most DDR500?

 

what clock rating are DDR500 rated for

anyway, its not amazingly relevant, the performance gains arent going to be massive

 

For AMD users, maybe not. For Intel users? Pushing that FSB wide open will give Intel the best gains they'll ever see, that is until an IMC hits the plate.

 

Not quite. Latency is a very different thing to speed, and you'll find that the 1GT/s memory will probably be a fair bit faster.

The 1GHz stuff will be able to give a very fast stream (well, not quite a stream) of data, but a while after it's requested. On the other hand, the DDR500 stuff won't be able to supply the information at the same rate, but it'll be delivered more quickly to the controller.

 

Nope ... i'm not talking about APX at all!!!! I compared 939 to AM2!!!!!!

 

Just saw an article over on Dailytech (http://www.dailytech.com/article.aspx?newsid=1702) that has the release dates as 6/6/06 (666) which reminded me of an article a while back that had a plane flying with the banner 6/6/06 (this article http://www.news-journalonline.com/NewsJournalOnline/News/Local/newEAST01032506.htm) but it says that its just for a movie.

Could be cool if it was actually AMD

 

And yet you clearly cannot grasp the concept of AMD working frantically, but that doesn't mean results in time.

Whatever they're working on, we're not going to see it implimented soon, or else we'll already of heard about it. Just like how conroe's amazing performance was ADVERTISED before it's on the shelf. since we haven't heard of anything new from AMD, it must be that whatever they're going to bring to the table they're not going to be able to offer it soon. AMD hasn't said anything, meaning that for the next few months, they're not going to even close to launching anything comparable to Conroe. And time matters, for people like me, who's waiting for summer to rebuild....

and uh... how is it 5 months away? If it's launching in Q3 then it's going to be from either april to june... and june is just 2 months away...

http://en.wikipedia.org/wiki/Fiscal_year

AMD's AM2 is not anything new, it's the same old architecture using DDR2, and it's not going to offer any incredible speed bumps over intels "new" (re-hash of PIII) architecutre... so I doubt we'll see anything from AMD which can pump out 20% performance increase soon.

And are you kidding me? AMD in my opinion is kinda in a mess right now, you got Fab 36 opening with 90nm with Fab 30 almost to a closing but still no 65nm in sight. That means they might have trouble just switching to 65nm let alone diverge resources to create a new architecture. And even if they do, if they're going dual core at 90nm, they're going to be outsold by Intel's 65nm dual core as it's simply cheaper, the same reason why I'm probably going to swtich to the blue side if Conroe is able to come out at June.... and it's relatively cheap too. A X2 3800 today costs roughly 340CAD, while a conroe E6600 is projected to cost only 360CAD and looking to see that an FX-60 got handled by a E6700, I would expect the E6600 to offer immense performance increase from the X2. So, which am I going to buy. AMD or Intel by the end of (launch date for conroe) Q3?

 

no, you said that 939 with a Performance rating of 3200 should have the same performance as a AM2 with a performance rating of 3200.

But that's clearly incorrect as I've stated. Same Performance rating doesn't mean same ammount of performance. Nothing to do with platforms.... I mean, don't make me compare sempron 3000 to Z64 3000

Same Performance rating does not mean same performance. Period.

 

with the move to am2 the cpu are exactly the same as socket 939 cpus except for the on die memory controller so the am2 cpus performance is equivalent 2 the 939 parts n that is probably why amd is keepin the rating the same so overall performance should only benefit from the memory which will be 10%. however, this dont mean that conroe will hav an easy ride because as hitman102 pointed out this is only a new platform and not a new marchitecture n i think that around the time intel starts launching their next-gen desktop parts we will start to hear more about amds next marchitecture.

 

How exactly is conroe miles ahead in performance?
http://sharikou.blogspot.com/2006/04/clovertown-scores-revealed.html

I'll answer that for you... it's not.

 

thats clovertown - not conroe ...

 

and that blog entry is a load of rubbish.

 

I'm well aware that it's clovertown. Take a wild guess what clovertown is. Once you figure it out then please try to post a comment that means something.

P.S. - that blog is hardly rubbish.

 

the clovertown is a much bigger proccesor then conroe

this means the 1ghz FSB is much more of an issue

conroe has 2 (or 4 with HT) cores right ?

well clovertown has 8 (if that includes HT - i guess it does)

That means clovertown needs twice as much bandwidth to be on par with conroe

Different problems

 

intels 'core' architecture (previously known as intel next generation microarchitecture) does not support hyperthreading. Clovertons 8 threads are accomplished through two sockets, with each processor consisting of two dual core dies, making 8 cores split over the two sockets. It is essentially two Conroe2M's on a MCM.

Considering cloverton will be using a 1333mhz dual independant bus (1333mhz bus per socket), each die will have 667mhz of bandwidth the same as merom will have in notebooks.

This blog has been discussed in depth over at xs, and to be fair it is rather a load of rubbish.