Printable View
At the International Solid State Circuits Conference in San Francisco, which kicks off Feb. 11, the company's engineers will demonstrate new technology that will make AMD's first quad-core chip more power-efficient than previous dual-core processors.
http://www.channelinsider.com/articl.../200885_1.aspx
Independent cores mmm sweet :banana:
Sounds very exciting! With all those power enhancements I wonder how low Barcelona can go in terms of power usage (7W??). Definitely this will be major selling point in server world because whole platform will be able to cut power significantly (switching off 14 from 16 memory modules can shave a bit of power;) ).
I can't wait for tomorrow :slobber:
true but they seemed to have missed the fact that the 2 power levels are one for the Memory controller and the other for the cores.
Thus the voltage across all the cores are identical.
I don't know what do they mean by "demonstrate Barcelona"?
AMD demonstrated Barcelona in decembar 2006. Do they mean that we will see some ES and benchmarks soon?
Showing a task manager is like showing a concept car covered with a blanket. No offense, but nobody really cares. Like you said presley, we want to see something other than a meaningless application that shows number of cores.
Not necessarily !Quote:
Originally Posted by brentpresley
This is original article:
http://www.channelinsider.com/articl.../200885_1.aspx
And it's titled:
AMD Reveals More 'Barcelona' Secrets
And if you look AMD's agenda for ISSCC, you'll see that they have planed only to talk more about Barcelona features, and not to run sPi ;)
man , you have got to help this old blind man , i went to yhe page that you linked too , and did not see there what you postedQuote:
Originally Posted by Nedjo
where does it say amd's agenda for isscc
i even did a search on that page for what you posted
maybe we will get lucky and amd will multitask
Heh, if I had one of those chips all 4 cores would be at 100% utilization all the time running F@H :fact:Quote:
If Core Zero is working a 80 percent utilization, but Cores 1, 2 and 3, are not very busy, they can lower their power frequency down
I hear ya...
Ryan
I cant wait for Barcelona!
~Mike
nice, but does the new platform still memory timing dependant ??like 4-4-4-12 vs 5-5-5-15 can have 5 to 10% difference
especially with DDR2 i havent seen some 3-3-3-1 like it was in the DDR 1 days ... ... to achieve low timing is hard 0.o today
so yeah this platform can be expansive, but hey RAMs r cheap nowadays GSkill 4-4-3-5 DDR2 1000 2GB PHU HZ is only $299 now vs $400+ b4
but anyway i cant wait... i hope clock for clock this is 20 to 30% faster than conroe
cuz if u want quad core + multi GPU u dont have a lot of options
QX6700 + nforce 680i sux... no fsb overclocking beyond 310fsb
in real world performance? i beg to differ :DQuote:
Originally Posted by theteamaqua
just like you said , straight from amdQuote:
Originally Posted by brentpresley
In addition to the estimated 40-percent performance advantage native Quad-Core AMD Opteron™ processors can offer over the competition in benchmark testing
40% core to core between clovertown and AMD's Quad :slobber: :banana:
think you've read something in that wrong. If anything AMD are saying a 40% between k10 & k8. This is only AMDs estimation on certain apps known to itself :nono:Quote:
Originally Posted by madcho
So in others words, until someone who we can trust benchmarks with an app we know, it's all smoke and mirrors
40% is best case scenario, and theyll need it, as the highest clocked chip will be at 2.5ghz, compared to clovertown 3ghz. overall, itll probably be around 10% faster.
according to brents link , they are going to be 40% faster then the competitionQuote:
Originally Posted by Motiv
In addition to the estimated 40-percent performance advantage native Quad-Core AMD Opteron™ processors can offer over the competition in benchmark testing
This is the problem with using AMD as a source. They can pick and choose whatever 'Benchmark' they want, to prove their performence.Quote:
Originally Posted by The Ghost
We have AMD in 1 corner saying 40% and unknown sources saying 10%. Personally I will go for an average of 15% on release.
Simulations conducted in AMD laboratories indicate that certain database applications will see performance improvements up to 70 percent and certain floating point applications will experience performance gains of up to 40 percent over platforms powered by current dual-core AMD Opteron processors
I hope they will also have the same Overclocking headroom as C2D/Q. I wouldn't bet on it though...
nn_step and loe are not too bad compare to the ghost.Quote:
Originally Posted by brentpresley
Well when they 40% advantage over the compention, do you think they mean just the processing power or as well power consumption into that figure?
It should overclock similar to the Brisbane (same production process). Probably 3 GHz on air if you are lucky.Quote:
Originally Posted by Poodle
Maybe better though, by then they should be able to have gotten 65nm more refined. Kind of like the switch from 130 to 90, originally the overclocks weren't as good as 130nm but we can hit 3GHz give or take with 90nm. Once 65nm is more refined I don't see why we shouldn't be able to see 3.2-3.4GHz range overclocks. Maybe not right away though. But really, no one knows yet, we'll just have to wait and find out.Quote:
Originally Posted by Scimitar
umm besides 65nm the two are anything but equalQuote:
Originally Posted by Scimitar
"Simulations conducted in AMD laboratories indicate that certain database applications will see performance improvements up to 70 percent and certain floating point applications will experience performance gains of up to 40 percent over platforms powered by current dual-core AMD Opteron processors."
This is almost certainly based on data presented by AMD several months ago at their "Virtual IT Experience." At InvestorVillage, they have screenshots (linked below).
"Certain database applications will see performance improvements up to 70 percent"
The Virtual IT Experience shows gains of 70% for OLTP. They do not specify what benchmark they used, but it has been widely speculated (and assumed?) that this is TPC. This is probably the source for the rumors (e.g. in this RWT article) that AMD claims a 10-15% TPC-C advantage with Barcelona.
"Certain floating point applications will experience performance gains of up to 40 percent"
In the Virtual IT Experience AMD predicted gains of 40% for specfp_rate2000. This is almost certainly where this number comes from.
What this tells us
The thing that interests me about both of these slides is that they suggest 10-15% better OLTP performance and ~30% (Some say 40%) better specfp_rate performance than a Xeon 5355. That's a 2.66 GHz Clovertown. There has been a lot of speculation that Intel will release a 3 GHz Clovertown this year, although to my knowledge Intel has never stated this. If they do, and we assume perfect scaling, the 3GHz Clovertown would be about 13% faster than the 2.66 GHz Clovertown. That would make it about the same as Barcelona in AMD's OLTP benchmark, and still significantly behind in specfp_rate.
What this does not tell us
AMD does not say what OLTP benchmark they used. It's been speculated that it's TPC, but then why wouldn't AMD just say TPC?
AMD also does not tell us the clockspeed of the Barcelona processor used for comparison. At this point, it's probably safe to assume that the Barcelona is the highest clock they plan on releasing in "mid-2007." We also don't know how the rest of the simulated platforms were set up.
These are quad-core server benchmarks, most likely selected to highlight Barcelona's strength. We don't know how Rev. H should do in other environments, although the rumor on these forums is ~10% faster than C2D per clock.
all i did was take a quote out of a page that you linked too , i did not start any rumors , what i linked to was in black and white , i have not seen anything else in black and white that proves that the info is false , or a rumorQuote:
Originally Posted by brentpresley
SO did they demonstrate it??
well 3GHz of this vs 3GHz of kentsfield.. i like to see that
as with 680i board u cant really fsb OC higher than 3GHz with kentsfield Q6600 or Q6400
unless nvidia some how find a fix for that
Yes AMD demonstrated it. They demonstrated it saving power while running task manager! :rolleyes:
Hey, someone had to say it! :D
or they saved more power by not running it at all :):banana:Quote:
Originally Posted by freeloader
umm i was talking about performance vs Xeon 5300s ... lol
It has been discussed by your beloved Scientia that the 40% is multiple of 4 cores. That the performance of Barcelona will not 40% or even close to 40%....... What say you to that?Quote:
Originally Posted by The Ghost
God forbid you say Scientia is wrong.
I disagree.Quote:
Originally Posted by Scimitar
Both are on 'same' production process, but Brisbane Revision G1 is using OLD transistor design (done for 90nm) which is severely slowing it down and incrasing power consumption. Barcelona is using new transistor design, more metal layers and every process addition AMD mastered in Dresden factory.
Best example is to look at voltage both cores are running on (for Brisbane it is 1.25V-1.35V@1.9GHz-2.6GHz):
http://www.xtremesystems.org/forums/...&postcount=541Quote:
A 65nm silicon-on-insulator process is used for producing the near-450-million transistor device, with dual stress liners and a silicon germanium process is used to speed up the pFETs. Eleven layers of copper and low-k dielectrics connect the device.
At 95 degrees Celsius, modelling suggests the processor will run at between 2.2 and 2.8GHz at 1.15 volts. Each of the four cores include eight temperature sensors. The on-chip northbridge contains a further six.
I will expect better overclocks from K10, especially dual core variants ;)
PS. Off topic THIS IS MY 500 POST ON XTREMESYSTEMS!!
Rumors: the production process used for Barcelona is a little different from the one used for Brisbane. So, maybe a better OC-er, maybe worse or maybe same.
TPC-C and SPEC_FP are bandwidth and FPU dependent benchmarks and are not representing the real world. Especially SPEC_FP is not adequate benchmark to conclude how Barcelona will perform in server environment. Servers use software which depends mostly of the ALU performance. IMO K8L AL performance will be 10% faster clock for clock than K8, but slower than Core2.
As you said it gojdo,it's your opinion and since you saw none of the tests and benchmarks of the K10,it's worth zero.Quote:
Originally Posted by gOJDO
True. This phoney war has lasted too long, it will be a blessed relief to see the first benches and know for sure.Quote:
Originally Posted by brentpresley
Regards
Andy
if you have something to say to Scientia , by all means go to his blog and post itQuote:
Originally Posted by n91htmare
but nothing that any one has to say positive about amd would be OK with you
all we have is what amd has said for sure , anything else is not a proven fact
funny when intel said said how good it was going to do with the conroe , , what did iontel say it was going to be 40% better ? 80% better? but then agan intel was talking about better then it's own cpu
didn't intel say the same thing about conroe , before it demonstrated conroe ?Quote:
Originally Posted by brentpresley
how long before conroe was demonstrated was the numbers coming out on how good conroe was suppose to be ?
there was no flamebaiting in my post , but any one sho posts anything good about amd to you , is flame baiting
well, regardless of how they perform against one another... both AMD 65nm and 90nm are proven not to overclock nearly as well as intel 65nm.
So, for the enthusiasts (like us) it should be an interesting war. Intel must be nervous though if they want to get 45nm out the door 6 months early.
you are comparing a 12 stage to a 14 stage processor on the field of max clock speed, while ignoring the fact that Intel has 6 months advantage in the 65nm Process.Quote:
Originally Posted by irev210
Nice comparison.
Werd™
I defiantly cant wait until Intel is back down on their knees.
Everything that Intel and I have been through, Im REALLY glad to see Intel the low dog.
~Mike
AMD has SiGe and DSL-SOI-3 :stick: and the new 65nm transistors, they will allow less power leakage. Brisbane use 90nm transistors :slapass:Quote:
Originally Posted by nn_step
I expect that Barcelona will overclock at 3.5ghz on air, like kentsfield. :toast:.
Shangaih may do it at 4ghz. :slobber:. Shangaih is on classic AMD's 45nm, but the next one will use imersion 45nm and High K :eek:. AMD is gonna clock very high :clap:
AMD said that +40% on clovertown benchs are not rare. ;)
I can't wait 4 see a sandra benchmark of the Barcelona's FX :slobber:
I'm just wondering me if the FX will use HT3 or old hypertransport :confused:. It Would great If AMD Get out a RD790+ bord E-ATX for the FXs with separate powerplanes.
Barcelona can do separate power planes, but actual motherboards F/AM2 can't. F+/AM2+ can ;)
what part of current Barcelona chips are about 3 months old and are missing a feature or two of the ones that you listed. You have to remember AMD always puts yields ahead of minor advantages. And the difference between AMD and Intel is that AMD is a Design company and Intel thinks of itself as a Manufacturing companyQuote:
Originally Posted by madcho
Shouldn't a smaller process have MORE leakage...Quote:
Originally Posted by madcho
That's why people are adding new features to the materials, like SOI, stranded silicon, high k dielectrics, etc, to offset some of the leakage problems.Quote:
Originally Posted by aznblueeboi
it largely depends on many factors, however just a dumb shrink would cause about double the leakage (for 90nm to 65nm)Quote:
Originally Posted by aznblueeboi
People aren't guessing when they say that, they are reading what you write... so when you got 5000 posts and at the 5001 you say "you don't know me" it's kind of funny.Quote:
Originally Posted by brentpresley
Unless someone comes out and says they hate amd, it's alls speculations and guesses.Quote:
Originally Posted by DoubleZero
I use an intel chip, because it gives me more bang for my buck. That doesn't mean I hate AMD or he hates AMD.
You know what's REALLY pathetic? When you offer more loyalty to a company, then to your wallet. Our biggest factor is $$$ as consumers. What gives me the best service at the lowest price is what I look at.
My last system was a 165 opteron, because it gave me the best performance at the lowest price. Now it's Intel E6300, again best performance at lowest price for me.
If a third company comes out and beats both intel and amd, ill go with them if their prices are lower.
If you chose to pay more $ for less performance just because you "hate" the competition is idiotic, and bad decision making IMO.
i wouldn't consider it to be idiotic to support the underdog so you good decision makers can continue to have choices to make those good decisions....
i do consider it to be idotic to call someone idiotic...wait, i think i just called myself idiotic....
in other words, ignorance is bliss... :D now can we get back to the discussion at hand?
was Barcelona demonstrated, or just talked about?
Sure mate, I can be 100% racist, but until i come out a say i hate black people or whatever it's all speculation and guesses. :slapass:Quote:
Originally Posted by n91htmare
Did i said that he was this or that based on what he uses? I think not... that must be the reason for you to need someone to come out and say... your intrepetation skills are weak...Quote:
Originally Posted by n91htmare
Excuse but... why are you talking to me?Quote:
Originally Posted by n91htmare
Oh i see, backing up a fellow fanboy...Quote:
Originally Posted by n91htmare
I fail to see where that dumb post by your fellow fanboy could own me in any way...Quote:
Originally Posted by brentpresley
im still waiting for AMD to beat intel's xeon 3200...
i dont want to see intel dominate and rise the price .. etc
Quote:
Originally Posted by brentpresley
Oh, don't stoop to their level. ;)
Hiroshige Goto has written an article about their presentation. It sounds like there was no demonstration. Not terribly exciting, but among the new information is that AMD confirmed the die size is 283mm^2, about what was expected.
Goto also labels the various parts of the Barcelona die, but I think the labels he applied to the top left core should have been applied to the top right core. (They are mirror images.)
yea , i didn't see anything about a demonstration either
Built on a 65 nm low-k SOI process with 11 copper interconnect layers, Barcelona contains 463 million transistors and has a 283 mm2 die size. The presenter was only willing to commit to a "2 GHz and up" performance estimate along with noting that the chip operates at core voltages ranging from 0.8V to 1.4V, but the published paper provides more details, "In a 95W max power envelope, the target frequency is 2.2 to 2.8GHz at 1.15V." Each core contains separate 64 KByte two-way set associative caches for instruction and data, along with 512 KBytes of L2 unified cache. A core-shared 2 MByte L3 cache completes the on-chip memory picture.
http://www.edn.com/blog/400000040/post/1400006940.html
I am sorry for not speaking forest gump. But i promise to study your dialect so that i won't fail to see no more...Quote:
Originally Posted by brentpresley
except that the L2 are NOT unified And the Vcore range is high, unless they are referencing the Memory controllerQuote:
Originally Posted by The Ghost