AMD Barcelona/Shanghai Quads to top out at 2.5GHz initially?

[red]

http://www.pcpop.com/doc/0/167/167267_1.shtml
http://www.hkepc.com/bbs/news.php?tid=708150&starttime=0&endtime=0
http://img2.pcpop.com/ArticleImages/0x0/0/377/000377897.jpg
http://www.rubyworks.net/quarkimages/Barcelona.jpg
http://www.planet3dnow.de/news_images/barcelona_cloverton.jpg

:) Isn't that cute that they show Clovertown's increased TDP..while they're still lower than AMD's for the most part?:p:

[FUGGER]

Max "idle" TPD?

Yay more slides and no product to back up the marketing spew.

[VulgarHandle]

:) Isn't that cute that they show Clovertown's increased TDP..while they're still lower for the most part?:p:

well, they actually show +%, so the dual core would be 120w not 80w, and quad-core would be 180w not 120w, and so on down the line with their respective +%

[metro.cl]

lol amd is just getting stupid with the comparatives slides.

same cpu cores same watts or lower and they :banana::banana::banana::banana::banana: about it, they should really get to work and improove the arch... to beat intel

[red]

Didn't you claim K8L to be 25% better? :DIt better with with those quad clocks..
Vulgar, I think your math is wrong?:(

[Qkjhfhaiguihfma]

lol amd is just getting stupid with the comparatives slides.

same cpu cores same watts or lower and they :banana::banana::banana::banana::banana: about it, they should really get to work and improove the arch... to beat intel
yea no kidding.

but ALL companies do it, if you don't have the best product you need to market whatever you do have.

[uOpt]

http://www.planet3dnow.de/news_images/barcelona_cloverton.jpg

:) Isn't that cute that they show Clovertown's increased TDP..while they're still lower than AMD's for the most part?:p:

Yeah, that. I am seriously considering framing this :)

And the guy who made this doesn't understand what "Thermal Design Power" is. Note: there is a reason why it is not called "Power" or "Thermal Power".

[p0tter]

wow stop throwing processors at mobo's and start adding cores to cpus.

[VulgarHandle]

Vulgar, I think your math is wrong?:(
Cloverton
Dual Quad % +
80W 120W 50%

80W + 50% = 120W for dual core
120W + 50% = 180W for quad core


their claim, not mine.. math is spot on..

[uOpt]

wow stop throwing processors at mobo's and start adding cores to cpus.

That would be fine if we could have low-cost 4 socket boards. But we don't.

As far as we can tell the number of CPU sockets in board classes will not change for AMD quad-cores. That means more than 2 sockets you go really deep into server class pricing, server class space requirements, same as socket 940 and same as Intel.

Right now Intel doesn't have Core2s for more than 2 sockets so AMD is still leading.

But if AMD maxes out at 2.5 GHz and keeps the same socket number as in 940 and 1207 then they are in trouble.

[red]

The 'Extreme' dual core Woodcrest 3GHz 5160 is 80W. The 'Extreme' quad core 2.66GHz Clovertown is 120W.:stick: You aren't reading the chart correctly. On the left side are dual cores, separated into 3 rows of performance, mainstream, energy efficient. On the right side is quad cores..:)

[nn_step]

umm that isn't an AMD slide

[VulgarHandle]

oh i see, it's the percentage of watts increased to go from dual to quad...oops

[red]

Umm.. Once again, thanks for the very insightful comment:rolleyes: I totally thought that was an official slide from AMD, greyness and all.

So anyone have have any idea how much of an improvement clock for clock these should be..combined with the low clocks..for total guesstimated performance:D

[VulgarHandle]

So anyone have have any idea how much of an improvement clock for clock these should be..combined with the low clocks..for total guesstimated performance:D
First, remember that the angle of the dangle is perpendicular to the heat of the meat, and proportional to the mass of the a$$, making it better for the eater and tighter for the .....pipelines

so i'd like to predict it to be at least 1% better clock for clock...

[perkam]

Max "idle" TPD?

Yay more slides and no product to back up the marketing spew.AMD will be a laggard till K8L, but if Intel did not believe K8L was a major threat, then it would not have needed to make plans for yorkfield to launch simultaneously so soon after core2 launch.

Perkam

[mr_knowitall15]

i like the chart. Saying that their native quad core will perform better than the kentsfield. Who knows maybe it can, what i DO know is that a little slide of a chip that doesnt seem to yet exist surely isnt out performing squat!

[Poodle]

I never liked the MCMs as I don't like hot chips. If Amd will be first with a native quad then hey, they've done a good job. I'm not impressed however how long it took (is taking?) for the K8 65nm shrinq. They better be faster with their 45nm chips or they will be in trouble.


I would consider a Kenty if they were cheaper though. Imo, $500 or so and it would be a nice price. :)

[metro.cl]

Didn't you claim K8L to be 25% better? :DIt better with with those quad clocks..
Vulgar, I think your math is wrong?:(

its gonna be 25% over C2D, i just don't like the wattage claims

[alpha_tot]

http://img2.pcpop.com/ArticleImages/0x0/0/377/000377897.jpg


:confused: So this means @ the same clock barcelonas twice as fast core for core??? Or am I missing something here??? WTF is a tick?

[Shadowmage]

Anyone else see the 2.5GHz limitation as a "good thing"?

AMD is currently in a rut in terms of performance. There's no way they'll release K8L without taking back the performance crown, or at least being very competitive (ie either win or tie). This means that AMD thinks that 2.5GHz will be enough against Kentsfield.

[kl0012]

This means that AMD thinks that 2.5GHz will be enough against Kentsfield.
Or may be AMD can't reach higher frequencies?
It may be very competitive against QX6700 but what about higher clocked Kentsfield or even Yorkfield?

[Shadowmage]

It's not a question of being able to reach higher frequencies. AMD can obviously do so if it's designed to go to 2.9GHz. It's a matter of yields. 2.9GHz is doable, but yields are obviously lower. Therefore, going higher is to be avoided unless necessary.

[gOJDO]

It will be another slap in AMD's face if the "glued"/MCM/whatever Intel's quadcore outperforms the "unglued"/native/whatever AMD's quadcore.

[zakelwe]

Well I think the following probably holds true, or at least I guess more people will assume roughly this

1) AMD Barcelona will be a lot more powerful than K8
2) It will have a great punch per Mhz
3) It will be at least a match for Kentsfield.

I guess the unknowns are how high it will clock and how it does against Yorkfield for price and performance. I still think that the 2 core versions of Barcelona or Yorkfield might be a more clever bet for the thrifty home user though. I know it's not as sexy though.

I guess we all wish it was closer though. It's not even 2007 yet and 2H2007 seems a long way away.

Regards

Andy

[Nanometer]

Be nice to get more ES around here so we have somthing to look forward to. As of right now, all we have is numbers on paper as far as I'm concerned.

[xVeinx]

Both AMD and ATI have been keeping a lot more under wraps lately, it's a little odd...

(yes, I realize they merged, I think you get my point though :p:)

[rcofell]

:confused: So this means @ the same clock barcelonas twice as fast core for core??? Or am I missing something here??? WTF is a tick?
A tick is the same thing as a clock cycle and yes that would imply each core is twice as powerful as a k8 core.

[LOE]

well now it seems a K8L core is twice as powerful as a K8 core

K8 cores get 4 gflops per core
K8L cores get 8 gflops per core

so a single 2.5Ghz K8L chip will perform like a 5Ghz dual socket dual core server - now thats hard to believe

if this is true then K8L will be more powerfull than everyones thinking, and intel will have problems matching it even with 45nm quad cores

[accord99]

well now it seems a K8L core is twice as powerful as a K8 core

K8 cores get 4 gflops per core
K8L cores get 8 gflops per core

so a single 2.5Ghz K8L chip will perform like a 5Ghz dual socket dual core server - now thats hard to believe

if this is true then K8L will be more powerfull than everyones thinking, and intel will have problems matching it even with 45nm quad cores

QX6700 is 42.67 gigaflops/s for double-precision. Core 2 Duo already went to 4 DP flops/cycle using SSE, K8L is just catching up.

[LOE]

last time I looked at server benches the only reason core based xeons ware faster was the higher clockspeed, unlike desktop where it is better clock for clock

so IF core with 4 flops/cycle = k8 with 2 flops/sycle

then what would it be against a k8l with 4 flops/cycle?

[accord99]

last time I looked at server benches the only reason core based xeons ware faster was the higher clockspeed, unlike desktop where it is better clock for clock
Like which server bench? Most server benches don't use vector math in any significant manner.

[kl0012]

Like which server bench? Most server benches don't use vector math in any significant manner.

Very true. 3.0GHz Woodcrest is up to 147% faster then 2.6GHz Opteron in intensive math calculations (like matrix multiplication):
http://www.cse.clrc.ac.uk/disco/Benchmarks/matrix.shtml

[FUGGER]

It should be impressive to see AMD on its own chipset, I dont think the hold out is intentional but more growing pains.

I hope they like the cold, we may see dual head coolers become more popular.

[uOpt]

The 'Extreme' dual core Woodcrest 3GHz 5160 is 80W. The 'Extreme' quad core 2.66GHz Clovertown is 120W.:stick: You aren't reading the chart correctly. On the left side are dual cores, separated into 3 rows of performance, mainstream, energy efficient. On the right side is quad cores..:)

The TDP is NOT the actual power consumption. Your assumptions are not useful for decision making.

[alayashu]

3) It will be at least a match for Kentsfield.

i'm betting it will not match the Kentsfield. It will surpass it in multithreaded apps like a Mercedes vs Trabant ;)

And the 2.5 Rhumors are just... rhumors.

[uOpt]

It should be impressive to see AMD on its own chipset, I dont think the hold out is intentional but more growing pains.


Exactly. My hope was AMD bought ATI to get back into the chipset business. So far we see nothing.

NVidia is really not the vendor of my choice for chipsets.

[LOE]

2.5 Ghz could be the clock of the first ES / tapeout 1 bad chips ;)

[LOE]

the quad core demo amd did was with A0 chips - so thats not too bad at all - booting windows in a 16 core config

A1 chips should be ready by january, man I hope we'll see some benches soon

[LOE]

hmm... well could it be the make many revs with one mask?

I heard there is another tape out comming shortly

[XS Janus]

A tick is the same thing as a clock cycle and yes that would imply each core is twice as powerful as a k8 core.

TICK :p:

[Lightman]

That was A2 silicon. Not A0.

I have friends w/ A1 and A2 silicon first hand. A0 was never sent out.

Brent do you know clocks of this A1/A2 revisions (approximately?) and if they are usable/working or partially usable only?
Many thanks in advance for any hints ;) .

[el rolio]

ahahah, what else besides alcohol!?!?!? dont tell me you are date raping your friends for stepping info!

[red]

Are they not good enough for benches because the performance doesn't match the paper specs or are they simply bugged?

[LOE]

even a chip running at 1Ghz will provide enough info to caculate raw instructions per cycle and by how much it differs from conroe and K8

hope they find one without bugs soon :)