CPUs: the near future.

[maverik-sg1]

I agree with K404 and think that AMD's BArcelona product due this year is an exciting prospect (opertons due out June, Desktop in October), if the numbers we see are anything go by then we will see neck and neck performance between Intel and AMD multi core cpu's.

The interesting part of this cycle of products will be the performance comparison of the AMD Desktop Barcelona chip versus Intel's 45nm product, which is rumoured for release around the same time as AMD's desktop Barcelona chips.

As for eDRAM - At this point I don't see either AMD/Intel using eDRAM, at present this 65nm tech is being supplied exclusively to Nvidia which strongly suggests that they are moving ever closer to releasing a 65nm GPU this year - although this will probably be a low end product to start with.

I'd love to see 65nm Geforce G81 8900's but more likely means that 65nm will be with us with G90 (9800's?) before year end.

 

[Master_G]

I agree with K404 and think that AMD's BArcelona product due this year is an exciting prospect (opertons due out June, Desktop in October), if the numbers we see are anything go by then we will see neck and neck performance between Intel and AMD multi core cpu's.

Barcelona is interesting, but IMO not that exciting, if they are only now coming back at Intel with performance about the same, and with 45nm on the horizon for Intel that doesnt sound too good for AMD to me.

I rather subscribe to the view of this editorial:

http://www.overclockers.com/tips01117/

AMD isnt sharing real numbers with people, and when AMD doesnt share good news, that tends to mean there isnt any. When intel were in the same position last year they released Conroe numbers, and that hardly did them any harm, why doesnt AMD do the same, especially since they are so needy of cash at the moment? At a point when they are trying to raise extra cash by expanding the number of shares in the company, surely you would want to release good numbers if you had them?

This to me seems to point to a lack of winning performance, which means they are only about catching up with Intel a year later, while Intel is moving on with 45nm and Penryn. To be a year behind is a lot in this business, especially from where AMD was when K8 first knocked Prescott's socks off. All rather disappointing for me tbh.

As for eDRAM, i welcome anything that might reduce power consumption in the next gen of GPUs, we might even get one that is a reasonable size with a reasonable power envelope (but i doubt it given current trends).

G

 

[maverik-sg1]

I agree with MasterG on all points reference the eDRAM

On CPU performance -the difference is nowhere near the gaping crevice it was (90nm am2 vs 65nm Core2) and compared to how long Intel took to comeback from the drumming netburst cpu's had (4yrs?), I think its hats off to AMD - in so far as on same size tech nodes, AMD may comeback with a competitive product in less than a year after it lost the crown.

Granted Intel will step to 45nm by then which means they are still in front, AMD should be able to take comfort that it took a die shrink for the gap to be extended again - although there is no comfort in being so far behind in the die shrinkage either at this point

On the point with AMD raising capital by selling shares - who knows what they did with all the cash they made when they were top of the tree and a top end cpu was a margin busting £700+, it's hard to understand why they purchased ATI when they are not cash rich???

AMD has never really been forward coming with ES or pre-release information - but I agree with most of that that guy in MasterG's link says as well, but consider that the first batch of barca's are clocked considerably lower then their intel counterparts so would it be unfair to bench them against Intel at this point? People will only ever look at the benchmarks rather then the clock for clock performance.

 

[Master_G]

I think another problem for AMD is that the Penryn shrink is not just a size shrink, they are going to start using those fancy new high-k transistors they were showing off a while back, and that is something that AMD probably wont get for a while yet.

As for making money at the top of the tree, when they were the best performing CPUs around, market share never rocketed up, Intel still had over 70% of the market, and AMD were limited by capacity and were desperate to use their performance to get OEM deals, so they never made a tonne more money.

I think the ATI deal comes from the expand or die viewpoint, either AMD increases its market share, or Intel will use its bulk to starve it of cash until it drops out (just as it is doing at the moment with the current price-wars). It could also reflect AMD conceding the battle for the moment and maybe hoping to redefine the battle in the future as a combined CPU and GPU effort.

I take the point about the lower clockspeed, but in the industry people can guess how much extra clockspeed will be gained by the launch, so numbers would still help them, or they could construct a benchmark with a Conroe underclocked to give the same MHz. I still think if they had a really exciting product in Barcelona then they would find some way to show it off to at least give their stock a boost. Hopefully they will prove me wrong, but i am not terribly optimistic about Barcelona atm.

I think that these are some interesting issues that are worth discussing.

G

 

[K404]

To take the tangent from another thread and plant it in here..the points coming out were balanced and good, as long as we can not have it descend into fanboi-ism (which seems pretty non-existant here) I dont see why it cant continue

IMHO AMD arent in a position to get wild with press releases regarding how great the new stuff will be...they cant afford to kill sales of AM2, and i`m guessing they`re still making the chips so backstore is *quite* heavy. I`ve also been told from a source that AMD are having trouble getting ES out to the necessary people because of how tight supply is at the mo.

I do think Intel and 45nm will take the crown back. Intels die-shrinking has been shown to scale better from launch, whereas again, AMD are cautious because they cant afford mistakes.

Actually...Intels 45nm may not tkae off as fast because the launch chipsets bring DDR3, and like last time a CPU, mobo and RAM switch will put a lot of ppl off until prices come down. Obviously, the performance crowd will jump faster if the 15% jump from DDR3 is promised.

What IS beginning to annoy ppl is the mobo design for Intel CPUs. A new PLL revision is planned for the P35/DDR3 mobos I hope revision includes a serious look at droop as well as just re-doing DIMM V supply. nVidia chipsets suffer worse droop than Intel ones...is that built into Intel design spec I wonder?

Ill stop there

K

 

[stocky]

Hehe, look at you lot with your intellectual essays. I'd better add a line or two

It appears to me that Intel are going to hold the cup for the larger part of this year until AMD pull something out of the depths of their R&D later in the year. I still hope that the future of CPU technology will allow increased efficiency rather than relying on brute force. Prescott comes to mind!

 

[maverik-sg1]

I still hope that the future of CPU technology will allow increased efficiency rather than relying on brute force. Prescott comes to mind!

Prescott was unfortunately an Audley Harrison - blows hot air and delivers nowt (so was neither forceful or efficient)!!

I know what you mean though Stocky mate - I am sure that now more than ever that tech developments in cpu's will lean toward fast and efficient - Intel are claiming that the 'leaks' that were identified in the 65nmm process are resolved for 45nm so they should use even less power (unless they identify a completely new set of leaks lol).

Unfortunately, the CPU is leaps and bounds ahead of GPU's in this area and it really is Nvidia and AMD (formerly ATI) that need to address those issues more seriously now. I guess the move the eDram is part of that process - we need to remind ourselves that the transistor count in a GPU is much more than a CPU iirc.

Now look I feel another OT coming on - truth is that current GPU technology outscales the computing power that normal clockers can squeeze from todays cpu's - therefore it is an ideal time for the two big guns to start producing more efficient products rather than looking to make it go even faster than it already does. Don't forget that Intel made it clear in Autumn of 2006 that it was srious about re-entering the performance GPU market - this could be the catalyst (no pun intended) to start the effieciency wars of 07/08.