AMD Unveils "Barcelona" Architecture

[Talrinys]

Might beat it in memory bandwidth but it will beat it in not much/nothing else.

I know that, never claimed anything else, but that doesn't take away from my point, at all

And memory bandwidth is important, his PC does feel a little smoother for everyday use.

 

[mosfet]

Penryn is die-shrink to 45nm, so there won't be any massive increase in performance over conroe/allendale, but they'll have a lower TDP and will be cheaper for Intel to produce.

The Barcelona benchmarks so far are inspiring a whole lot. 15% or so faster than K8 clock for clock, but not blowing away the competition.

 

[sargatanas]

I know that, never claimed anything else, but that doesn't take away from my point, at all

And memory bandwidth is important, his PC does feel a little smoother for everyday use.

Never had a problem with my 6750 i have to say, but ill take your word for it .

 

[Gashman]

Penryn is die-shrink to 45nm, so there won't be any massive increase in performance over conroe/allendale, but they'll have a lower TDP and will be cheaper for Intel to produce.

The Barcelona benchmarks so far are inspiring a whole lot. 15% or so faster than K8 clock for clock, but not blowing away the competition.

thing is though. core architecture is only about 10% or so on average faster than K8 clock for clock, so it brings barcelona equal/above core in that respect, albeit not by much, i heard they scale VERY well though, all in all not looking to bad

 

[Owenb]

thing is though. core architecture is only about 10% or so on average faster than K8 clock for clock, so it brings barcelona equal/above core in that respect, albeit not by much, i heard they scale VERY well though, all in all not looking to bad

The problem is that intel already have a 3ghz quad core, while AMD might not have that for 6 months. Yes i know that the top end is just a niche market, but gives a good indication of who is producing the better products. Intel will also have the advantage of 4.5nm vs 6.5nm, giving bigger profit margins to squeeze AMD with. On top of that AMD are bleeding money and have just wasted a huge chuck of cash on the failed ATI merger. Add all that up and I don't think AMD will see next christmas, which is a sad thought.

 

[simonnance]

the most important question hasnt yet been answered.....

how do they overclock?

Coz if those 2Ghz beasties can give 3.5Ghz goodness or higher, then with the scaling that they seem to have AMD will have an OCer's pick of a chip for sure.

If they top out at 3Ghz then things might be a little hard on them with C2Qs hitting the mid 3Ghz and Penryn giving us less heat to worry about.

 

[Final8y]

The problem is that intel already have a 3ghz quad core, while AMD might not have that for 6 months. Yes i know that the top end is just a niche market, but gives a good indication of who is producing the better products. Intel will also have the advantage of 4.5nm vs 6.5nm, giving bigger profit margins to squeeze AMD with. On top of that AMD are bleeding money and have just wasted a huge chuck of cash on the failed ATI merger. Add all that up and I don't think AMD will see next christmas, which is a sad thought.

Can you post a link to the failed merger..im having trouble finding it?

 

[Gunther]

Who states that pernyn will give less heat worries, possible the max temp can be lowered from 61c to perhaps 51c. The pentiums could also handle 100 c, and the C2d's only 61, now 71 c. So unless you've got accual proof of the termal limit of the penryn. The evolution shouws that the limits only go down seldom up. He for haven sake perhaps the max temp for a penryn can be only 45c. So where is the added headroom then???

 

[simonnance]

smaller manufacturing process => less voltage => less TDP => less heat output

what the designed temp limit for the chip will be i dont know, but i doubt they will drop it that much if at all. 45C would be insane, imagine one of those on a hot day (which in the USA can be 40C easy) on a stock HSF! Chip designers apply a sensible range of operation temps based on the needs of the chip, and the environment in which they wish them to operate. 61C isnt even the Tjunction for early C2Ds, its 85C (now 95-100C with the newer ones i think), and that's with chips that run pretty cool (excepting the quads of course).

Pentiums NEEDED 100C max temp, or the EEs would have self destructed when you turned them on.

 

[Justintime]

Who states that pernyn will give less heat worries, possible the max temp can be lowered from 61c to perhaps 51c. The pentiums could also handle 100 c, and the C2d's only 61, now 71 c. So unless you've got accual proof of the termal limit of the penryn. The evolution shouws that the limits only go down seldom up. He for haven sake perhaps the max temp for a penryn can be only 45c. So where is the added headroom then???

You're on about Tcase max maybe, Tjunction is 85-100C depending on which stepping.

 

[mosfet]

On top of that AMD are bleeding money and have just wasted a huge chuck of cash on the failed ATI merger. Add all that up and I don't think AMD will see next christmas, which is a sad thought.

Failed is a little harsh. Neither companies are performing well, both have had to deal with severely delayed silicon and channel issues. The merger has at least allowed them to streamline operations and lay off staff to get costs down.

Who states that pernyn will give less heat worries, possible the max temp can be lowered from 61c to perhaps 51c. The pentiums could also handle 100 c, and the C2d's only 61, now 71 c. So unless you've got accual proof of the termal limit of the penryn. The evolution shouws that the limits only go down seldom up. He for haven sake perhaps the max temp for a penryn can be only 45c. So where is the added headroom then???

I think you're misunderstanding how TDP and minimum channel length correlate. By reducing the FET channel length, switching can occur at lower voltages, meaning less heat is dissipated. This reduces the stock TDP, but heatsinks still have the same thermal conductivity, so you effectively have more temperature headroom.

 

[Owenb]

Can you post a link to the failed merger..im having trouble finding it?

Sorry i should have said takeover not merger.
What have ATI brought to AMD expect a sub standard GPU, 2*00 series, and run of the mill chipsets? All i can see is a huge hole in cash in hand, and a massive loan to pay back. Maybe in the longterm it would have worked out if they could have got fusion off the ground, but AMD will be belly up by then.

 

[Final8y]

Sorry i should have said takeover not merger.
What have ATI brought to AMD expect a sub standard GPU, 2*00 series, and run of the mill chipsets? All i can see is a huge hole in cash in hand, and a massive loan to pay back. Maybe in the longterm it would have worked out if they could have got fusion off the ground, but AMD will be belly up by then.

Can your link me to the failed takeover as that would imply that the bid did not go through noire exchange of money or control of one being given to the other.

 

[Owenb]

Can your link me to the failed takeover as that would imply that the bid did not go through noire exchange of money or control of one being given to the other.

If you are going to be a smart ass you might as well spell 3 letter words correctly.
Everyone knows that AMD are in huge trouble. You don't make a quarterly loss of $600m on $1.4b turnover and turn it around overnight. They got a few quarters to turn it around before they are out of cash, as they spent a large sum of their reserves on ATI. They are already up to their eyeballs in debt so i doubt another loan is in the equation.

 

[tsinc80697]

was hoping AMD would be able to pull something out the bag really, even if it was just for competition sake and lowering Intel prices.

Still remember the days of using AMD and refusing to go to "the dark side" of Intel lol, they're miles behind now compared to what Intel seem to be knocking out

 

[Emlyn_Dewar]

Miles behind?
From some benchmarks of the initial server CPU releases, which weren't too bad given the clock speed.

It's all doom and gloom here...

 

[tsinc80697]

But how many months have the Intel CPU's that perform similarly been out for and how long until the new batch of Intel CPU's?
Seem to be staying 1 step ahead on release times

 

[Tute]

So if Penryn is 5% faster than Conroe, and Barcelona is 20% faster than K8, surely that would put AMD on an even level, or right behind Penryn?

Good overclocking and/or good pricing *could* see Intel come unstuck. Bit early to be thinking AMD will "collapse".

 

[cavemanoc]

We're all making the same mistake here. Overclockers are a complete irrelevance (us not the company!) We're not a financially important part of their budget - we don't buy enough - same goes for ATI/nVidia - which do you think sells more 8800Ultra or 8600? Amd have been doing something they've never done before in any serious way - breaking into the server market - this is where the big bucks are, and newsflash - they don't care who has the speed crown this week - they want best economy - performance at a small price tag, with... a low thermal footprint and low power rating... sound like any chip you know about to come to market (1st to say Penryn gets an e-slap!)

As sad as I am that AMD can't reclaim the speed crown - I do at least realise that it's not the top end enthusiast chip that makes a company - the advertising doesn't hurt, but it's not the stuff of long term survival (just look at ageia's physx card - fastest on the block... financial success?). IT managers don't care how well a chip clocks - they just don't - they want stable, cheap, and energy efficient with best performance per pound.

The real question is "Will barcelona give better performance/pound/watt than Penryn?" I believe it will, and I'm betting a lot of server 'farmers' will agree!

 

[Emlyn_Dewar]

Will Dell be shipping the new AMD desktop offerings (when they're released) in it's desktop boxes?