Official AMD Barcelona Thread

[Shintai]

http://www.informationweek.com/news/...leID=199501467

anyone know the validity of this statement (on page 2), refering of course to the IMC supporting both DDR2 and DDR3

Since DDR3 is for AM3 with serial VIDs I would say no.

[Face]

Real World Tech has just updated their article on Barcelona:

http://www.realworldtech.com/page.cf...WT051607033728

Excellent article with very high technical details.

I admit that it's mostly chiense to me, So if you can crack it down it or draw any performance estimations from this, Please do.

[madcho]

Real World Tech has just updated their article on Barcelona:

http://www.realworldtech.com/page.cf...WT051607033728

Excellent article with very high technical details.

I admit that it's mostly chiense to me, So if you can crack it down it or draw any performance estimations from this, Please do.

i readed it too, very nice

AMD show us all please

[gOJDO]

@Face
A very educative article, but not enough detailed to make perfromance estimations. IMO, K10 will be a FP monster. It will outperform any x86 for FP per clock per core. For non-SSSE3/SSE4 optimized applications it will perform roguhly same as Core2 at same frequency. It will be slower than Core2 for arithmetic and logic calculations, at same frequency.
On 4P+ it will dominate for sure and maybe on 2P. On 1P, Core2 will remain holding the crown.

[informal]

Hmm let's see if what the last "expert" said will hold true when K10 arrives.I think not.

[madcho]

Hmm let's see if what the last "expert" said will hold true when K10 arrives.I think not.

if you think in 1P K10 is better i'm with you

[madcho]

...
David Kanter once again struck a gold vein for hard-core enthusiasts and engineers, by posting an article that goes in-depth with AMD K10. While Phil Hester claims that AMD never called Barcelona/Agena/Griffin processor line K10 (nor K8L, which was invented Charlie and then "confirmed" by another AMD executive), their own software guide calls for 10h family of processors. You do the math, and while you're doing it - head over to RealWorldTech and enjoy.
...

source

amazing

[eXceeded]

erm... don't get too excited *cough*R600*cough*

[Face]

According to the writer of the article, Barcelona is a "solid improvement across the board" over AMD's current chips, and it should give the firm decisive advantages in the high-performance computing and multi-processor server markets.

With 2P, 4P I do agree and think it will dominate - Caneland\Tigerton should narrow the gap though.
I think 1P and desktop varients would(and should) edge current Conroes, even if just by a marginal gap.
When Penryn is out my guestimation is that they will perform toe-to-toe and that should be a nice battle to witness.

[informal]

Madcho, I truly believe 10h will spank Conroe/Penryn across the board in 1P configs (including int besides the fp domination) and all that while consuming less energy and working at lower clocks.
Intel however could try to pump Penryn derivatives to well above 3 Ghz and try to counteract.

[madcho]

Madcho, I truly believe 10h will spank Conroe/Penryn across the board in 1P configs (including int besides the fp domination) and all that while consuming less energy and working at lower clocks.
Intel however could try to pump Penryn derivatives to well above 3 Ghz and try to counteract.

10h is low TDP, , oc will be great

[LowRun]

source

amazing

This is just an Inq link to the Real World Tech article previously mentioned, how that makes it amazing

[gOJDO]

I bet Penryn will wipe the floor with K10 on the 1P configuration as much as Core2 did to K8. Anyone interested about placing bets PM me.

[Shintai]

NDA finish the 14, but don't dream to see Barcelona on stock

a french web site :



translate by google ( seems good )

Time caught up..no k10!

[Shintai]

source

amazing

Just like theinqs track record

[LowRun]

I bet Penryn will wipe the floor with K10 on the 1P configuration as much as Core2 did to K8. Anyone interested about placing bets PM me.

Maybe try with David Kanter, from what he says this remains to be seen, he looks like he knows much more than you tho'

[frankR]

K10 better beat Core 2 with overclocking not just at stock for me to consider it a success. Some how I doubt we'll see K10 much above 3 Ghz.

[adamsleath]

c'mon amd; i want a barcelona / X4 ; intel are way too smug at the moment; they're annoying me.

[madcho]

K10 better beat Core 2 with overclocking not just at stock for me to consider it a success. Some how I doubt we'll see K10 much above 3 Ghz.

i think K10 at 2.6 will match 3ghz c2q.

In the better case intel will release this year penryn 3.3ghz. And i don't think intel could do it before early 2008.

I'm very confident about oc of K10, TDP are low and vcore too

[EternityZX9]

erm... don't get too excited *cough*R600*cough*

Who created the R600? AMD or ATi?

Now, who's fault was the R600 really?

[madcho]

8.38 drivers show us XT is faster than GTS640 in evry games, wait 3 or 4 months and the XT should beat a GTX :love:

[Cooper]

Let us not forget that this thread is not about graphic cards

[KTE]

I know which one I'm getting :p

Hopefully DT doesn't get a sample before official reviews are out...

Perkam

And add [H] on that list of "don't want them to get at it early" for me too

And yeah, AMD is "saying" to release those boards you were seeing with their own chipsets along with the Agena FX release (after Barcelona for server).

I would "guess" (because even inside the industry you cannot do more than that right UNTIL the very last minute) mid Q3 for the Agena FX, the AMD boards and the 'ol 65nm XTX. But I know Computex is hot on the agenda so you never know.

Additionally competition is so rough 'n aggressive right now, with Intel on the upper edge obviously and already shooting its prices down, let alone then. Price wise AMD cannot match that with their new offerings already (unless some lower clocked ~1.9-2.3GHz Kuma better a E6750/E6850 ) and would therefore rely on the mid-higher range for fame 'n fortune I feel.

We enthusiasts are not even 0.1% of their intended market. The world is mostly based on price and then performance along with local availability and "trends" when buying, with the server/workstation/OEM market the largest and most profitable by far, and yet they look to good performance, business relationship, low TDP of the biggest factors.

I have no idea how they perform, too much gobble wobble around, not just online, even inside the industry professionals. Though AMD vs Intel processors have always been top notch when released, whereas ATi vs nV is a different story altogether, more usual to see them on par with each other.

People have to realize the Core 2 Duo architecture is damn very good for it's time to level, let alone beat. Early stages I reckon the Penryn: Wolfdale/Yorkfield at 3.33GHz will not be much of an improvement over a QX6800/QX6850 at 3.33GHz. And I have seen early performance to base this on, although they were not tuned and tweaked at best for release, they were ES samples that Intel saw fit to pimp.

But I predict that the Barcelona cores will be hits in the OEM/Server/Workstation (Home/Business/Governmental) markets (like SP/DP/MP - SMP, MIMD, NUMA, MISD, ASMP and multi-node configs for supercoms due to low TDP etc). I suspect their highest will do well for desktop range too (no idea how well), but how the middle and low end stack up is probably what will decide their outcome in the desktop market I feel.

I'll wait as I deem it only sensible to and hardly believe much around until I see it tested by someone trustable and with credibility beyond site hit gathering. Could easily turn out just as what happened with R600 now. (the highest main one not released - although we haven't seen the XT's stable performance yet, but promises from AMD for the new 8.38 driver and the one after it that supposedly fixes the AA issues- hopefully in the next 2 weeks a decent "valid" review should be out rather than a preview, as far as I'm hearing from some reviewers).

[adamsleath]

ho hum; i hope something turns up in the next 6 months.

multicores was/is supposed to reduce the "need" for higher clocks as the only performance measure...but it is now more cores and higher clocks as fab processes continue to improve and shrink.

Which makes me hungry for more cores and more speed.

...waiting like a drooling vulture...

it seems in the intel camp that only a few motherboards are capable of achieving high fsb's with current quad core chips. (eg. EVGA 680i, p5k)...so i'm not rushing into an intel quad at this stage.

I reckon AMD X4/barc/agena/some variant of...will be a quad core option. I hope they overclock well...or i might just update to a newer sli board when it arrives and plonk in a yorkfield....impossible to predict...i just hope that AMD come up with the goods.


...and then intel will bring out Nehalem in 2008 and the tables will turn again probably.

[Face]

AMD shows sunny side of Barcelona:

http://news.com.com/AMD+shows+sunny+...l?tag=nefd.top

AMD demonstrated its Barcelona quad-core server chip for reporters and analysts here Monday, comparing its performance to one of AMD's dual-core Opteron processors. This marked the first time AMD shared Barcelona performance information with anyone outside of its server partners and internal folks, said Randy Allen, corporate vice president for server and workstations at AMD.

The company ran a demo comparing the performance of two four-socket servers, one using the quad-core Barcelona chip and one using a dual-core Opteron chip. The demo measured the performance of the chips on an imaging benchmark called POV-Ray, and as you might expect, the quad-core chip finished its task quicker than the dual-core chip.

The quad-core chip processed about 4,000 pixels per second in rendering the image, while the dual-core chip could only hit around 2,000 pixels per second.

I'm surprised that only Cnet has reported about this as of yet and I couldn't find it mentioned anywhere else. Im sure they weren't alone while demonstrating. Maybe tommorow.. But at least it seems like wer'e getting close to seeing some numbers at last.

--------------------------------------------------------------------------

PS, This doesn't indicate any performance levels, But it's a good sign to see they're out there.
Please take in mind that it could be complete BS, And I cant validate any of this so don't take this for granted.

It's not a benchmark, just the average plys\KNs reads from the game.

Zappa on a 32 threads AMD (Barcelona core):

Zappa Zanzibar X64: 28.7 ply; 24,375kN/s Quad-Core AMD Opteron(tm) Processor 8270 2612MHz, (32 threads)

http://216.25.93.108/forum/viewtopic.php?t=13936