Barcelona surrenders to Shanghai: 45nm AMD heart beating 20% faster?

Too bad. Another discussion ended becuase of fanboy voting. Waaa.

Message edited by turpit on 10-02-2008 at 04:16:18 AM

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

http://www.fudzilla.com/images/stories/2008/November/General%20News/amd_shanghai.jpg

AMD's MHz run faster than anyone else's - e-kenmac (courtesy from The INQUIRER)

Register or log in to remove.

Even if this is true, they won't take the lead, though they will finally have a good strong competitor to intel's Q9000 series quads if this is the case. I'd love to see them take the lead as much as anyone, but I don't see it happening with this revision. This could definitely start to close the gap though.

Reply to cjl

Well the AMD has have been too optimistic sometimes... But, when they have compared to their own older products, the numbers have been in line of their claims. But offcource this is from inquirer, not official...

Reply to hannibal

I just hope to see AMD competitive. It does everyone good to have them putting out quality products.

Reply to dirtmountain

Even with this advance, AMD is likely to fall even further behind.

The reason is Intel's forthcoming server line will have significantly improved memory performance due to adding the IMC and switching to DDR3.

Intel was way ahead in processing power, but in very memory intense functions such as an ESX Virtual Server, AMD did pull ahead.

However, Intel will likely claim the memory performance lead.
The Improved CPU performance for AMD would only close the gap they had with Intel's previous generation chips, but would still lag perhaps even further behind the upcoming server chips.

Right now, we buy AMD Servers for our VMWare ESX Server farm, but Intel for most other jobs. I suspect that in 3-6 months, we will be all Intel for our servers.

It's nice to see AMD progressing, but instead of getting there in Q1 '09, they needed to be there in Q2/Q3 '06.

------------------------------ If its good in theory but not in practice,
its not good theory.
Reply to zenmaster

dattimr wrote :

Well, actually, it's not from The Inquirer, but from CNET news. Even Daily Tech reported it: http://www.dailytech.com/Shanghai+ [...] e13090.htm

Yes, but it's from an interview with an AMD GM.

This could be very good for AMD in the server space, but remember Randy Allen's claims of 40% over Cloverton? Of course, Mr. Plata is comparing with Barcelona, this time around, and not the competition. So, hopefully, it does what is being claimed.

------------------------------ 1:

http://valid.canardpc.com/cache/banner/481029.png

2:

http://valid.canardpc.com/cache/banner/464866.png

Reply to NMDante

Show it, don't tell it.

------------------------------ Intel will not take the top spot, or probably the top 3 spot back for the forseeable future. Not even with 32nm and more cores will intel be able to beat Jaguar. - JennyH the AMDiot, Nov 2009
Reply to yomamafor1

We can only hope AMD can pull all this off. However, many of us don't put much stock in what AMD says since their last major launch.

If AMD returns to being competitive with Intel consumers stand to win big time.

------------------------------ jennyh wrote: AMD break-even Q4 2009. *Gauranteed*

RabidFanboysSpreadingFalse.Info
Reply to TechnologyCoordinator

Whatever AMD releases, it has to be a quad core that performs better than a Q6600 at 3 GHz for $189 AND put out less heat which shouldn't be hard. Otherwise they are still up **** creek. We won't see any of that from AMD until closer to christmas for desktop parts but at least we get an idea soon.

------------------------------ If you don't know what OS/2 is, you don't understand.
Reply to rockbyter

and that my friends is all I have to say about that.....

------------------------------ Athlon 64 AM2 6000+
Gigabyte M61P-S3
4 GB OCZ Fatal1ty DDR2 800
Asus 4850 512mb
Reply to caamsa

^+1

Reply to BadTrip

http://news.cnet.com/8301-13924_3-10054038-64.html

Quote :

As a result, the schedule for Shanghai has been pulled in. "Originally the plan was that Shanghai would launch in Q1 of '09 and we were able to pull that into Q4," according to Patla, adding that the product will not only be announced in the fourth quarter but vendors will be shipping servers in the fourth quarter.

"We're in full production right now in the factory," he said. "People will start getting first silicon from the final production very shortly."

Patla asserted that Shanghai is a "very power efficient product" and will perform much better than Barcelona because the smaller 45-nanometer process yields "a lot more (clock) frequency."

At the same frequency (speed), Shanghai will outperform Barcelona by about 20 percent, Patla said.

AMD is also boosting the size of the cache memory, which typically speeds performance, from 2 megabytes to 6 megabytes. Another speed improvement will come from increasing "instructions per clock," Patla said.

"We're also turning on HT3 (HyperTransport 3) and you'll see partners start to validate that in the Q1 time frame," Patla said. HyperTransport is a high-speed communication link technology between silicon.

Shanghai will be followed by a 45-nanometer desktop processor code-named Deneb, which is due to launch in the fourth quarter of this year or first quarter of 2009, AMD said.

In the fourth quarter of 2009, AMD will add a six-core processor. "We'll take what we've learned from our 45-nanometer process and Shanghai core and bring out an Istanbul six-core product," Patla said. Like Shanghai, this will be targeted at servers with up to eight processor sockets.

Better to hear it from an engineer than the AMD marketing dept (unless the engineer is Casey Jones :ouch: )

Reply to wisecracker

Hope somebody gets one and "leaks" its performance

------------------------------ I went drifting, thru the capitols of tin, where men cant walk and cant freely talk, and sons turn their fathers in
Reply to jaydeejohn

Quote :

In the fourth quarter of 2009, AMD will add a six-core processor. "We'll take what we've learned from our 45-nanometer process and Shanghai core and bring out an Istanbul six-core product," Patla said. Like Shanghai, this will be targeted at servers with up to eight processor sockets.

This is my favorite part. They have barely had much more than 6 month to tinker with their 45nm process and they think they have learned a lot from it? They haven't had it long enough to have learned a lot from it. That and they have yet to see it in the public hands which is much different from their controled labs.

Well still feels like hype. The few Deneb benchmarks I have seen don't show 20% over Phenom clock per clock. But maybe in server they will be able to increase memory bandwidth 20% and give a boost in performance there.

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

http://www.steamcalculator.com/76561197970703804/camo_sig.png

Reply to jimmysmitty

Yeah ... lets see some ES benchies please??

Some of you must have one by now.

I heard you had to marry an ugly daughter of an AMD exec to get some ES silicon ... whereas at Intel half the forklift drivers in the loading bay have engineering samples.

I have only ever held one AMD ES in my hand ... bit like a four leaf clover.

The guy who showed it to me had an ugly wife too ... hmmm.

Message quoted 1 times

------------------------------ Great spirits have always encountered violent opposition from mediocre minds

Reply to reynod

I'm going to give AMD the benefit of the doubt here and believe these press releases. AMD needs to re-establish it's relationship with the media. An on-time (or ahead of schedule) release, plus honest statements about performance, will do wonders.

Plus, Dirk desperately needs to distinguish himself from Hector/Henri.

Message quoted 1 times

Reply to exit2dos

Im just hoping IF the 20% is true, it DOESNT include server. Tahtd be good to see, as itd put it in close to i7

Message edited by jaydeejohn on 10-01-2008 at 02:59:09 AM

------------------------------ I went drifting, thru the capitols of tin, where men cant walk and cant freely talk, and sons turn their fathers in
Reply to jaydeejohn

exit2dos wrote :

I'm going to give AMD the benefit of the doubt here and believe these press releases. AMD needs to re-establish it's relationship with the media. An on-time (or ahead of schedule) release, plus honest statements about performance, will do wonders.

Plus, Dirk desperately needs to distinguish himself from Hector/Henri.

I gave AMD the benefit of the doubt with Phenom. Only because I know K8 was a great design so I expected the same with K10. Core 2 I was hoping was decent but was still very worried based on Prescott.

So as of now I doubt the claims until I see a benchmark.

I also doubt that doesn't include server side. The 40% that was spun to the desktop side was actually server side and in 4P+ servers. This is probably the same thing TBH.

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

Reply to jimmysmitty

reynod wrote :

Yeah ... lets see some ES benchies please??

Some of you must have one by now.

I heard you had to marry an ugly daughter of an AMD exec to get some ES silicon ... whereas at Intel half the forklift drivers in the loading bay have engineering samples.

I have only ever held one AMD ES in my hand ... bit like a four leaf clover.

The guy who showed it to me had an ugly wife too ... hmmm.

ROFL.

I have high hopes (but low expectations) for AMD. Their management seems to better now and hopefully they have learned from past mistakes. Perhaps it's time to stock up on a little NYSE:AMD?

Message edited by TechnologyCoordinator on 10-01-2008 at 04:20:41 AM

------------------------------ jennyh wrote: AMD break-even Q4 2009. *Gauranteed*

RabidFanboysSpreadingFalse.Info
Reply to TechnologyCoordinator

Hopefully theyve learned to do with their cpus as theyve done with their gpus. Sneak up and slay em. Im not saying this is going to happen, but they did do it right with ATI products, and hopefully when they say 20%, its at least that, sorta like how the 4850 was to compete with the 8800GT heheh. But who knows? Im hoping the best as well, just as Im sure since they released their 4xxx series gpus, nVidia is working on something good, if AMD does good here,you can bet Intel will crank it up as well, as I dont see them going back to the P4 days unless AMD squalors

------------------------------ I went drifting, thru the capitols of tin, where men cant walk and cant freely talk, and sons turn their fathers in
Reply to jaydeejohn

the 4850 is not in competition with the 8800GT. They're an entirely different generation. They're in competition with the 9800GTX+. The 8800GT is $130, vs the 4850's $200, and their time in release is so great it's not a valid comparison.

I'm interested in whatever drives competition. I don't care who takes the lead as long as I get the best prices and products. Screw the fanboys.

Reply to resonance451

The announcements and slides from ATI said it was to compete with the 8800GT, throwing nVidia off as to its true performance. It was ATI/AMDs call, no one elses. It was part of their marketing strategy, and its paid off handsomely, as it kept nVidia to going to 55nm sooner rather than later, and ATI has much better marketshare because of this

------------------------------ I went drifting, thru the capitols of tin, where men cant walk and cant freely talk, and sons turn their fathers in
Reply to jaydeejohn

This will help AMD remain competitive in the server space, but right now I'm more worried about the consumer space Of course since Core 2 parts are generally at least 20% faster this will, at best, allow AMD to catch up. I'll wait to see bench marks.

The core 2 parts aren't 20% faster than Barcelona. 20% will put it right in Kentsfield territory.

Message quoted 2 times

Reply to sseyler

A 20% increase sounds good to me, it's just a shame the B2 stepping muddied the waters (which personally I think the TLB errata got blown way out of proportion) because with the Phenom's current pricing they're good chips. Last I checked you could pick up a 9850 for $144 with a hard drive combo on Newegg. That's a steal, especially when you go back and look at old benchmarks for the 9850 back when it was $235. For most tasks (with the exception of 3D modeling and power consumption) it's just as good as a Q6600 but cheaper.

I'm pretty interested in seeing what their FX line can do as well.

Message edited by Ryun on 10-01-2008 at 05:58:59 AM

Reply to Ryun

sseyler wrote :

The core 2 parts aren't 20% faster than Barcelona. 20% will put it right in Kentsfield territory.

I guess I should have said two Core2s slapped together since I meant to compare them to the Intel Quads. :lol:

I expect they'll be competitive with Nehalem if they get an FX CPU out. They'll be competitive vs. Core 2 quads with Deneb. Competitive doesn't mean faster in everything, but it means fast enough in important applications to merit a look at a total AMD platform.

What I'm wondering about on the desktop are the 45nm processors without L3 cache. Leaked roadmaps claimed Deneb quads and Heka triples will have L3 cache (shared with Heka). Propus quads and Rana triples will not have L3 cache. Will lack of L3 cache significantly affect the hoped for 20% performance gain over B3?

Their 45nm duos also are interesting, but they look like 2 cores are disabled in a quad, as they still have the same power requirements. AMD needs to do something about power and thermals, especially in a dual, but that might not happen until their second generation 45nm.

At any rate, I can't wait to see the benchmarks on the new desktop parts in December or January. If it's worth it, I'll consider an upgrade. If it's not much faster than B3, it may be like replacing my 3870x2 with a 4870, not enough of a boost to warrant a change.

When (not if) AMD gets out parts that match or beat Intel at most price ranges, will they really gain market share or will it be like in the Netburst days when people buying at big box stores bought Intel due to marketing, not performance.

Will AMD ever market their CPU's to the masses? Right now, they should consider touting their total platform, because the 780G is a nice IGP and that's all the average person needs. Yesterday, I tested LOTRO on the IGP, and it didn't look all that bad, not as nice as with the 3870x2, but not bad at all. More people play WoW, LOTRO etc. than Crysis, so all they really need is a good AMD or Nvidia IGP. That's where AMD's marketing should be concentrated.

Maybe Thunderman's on to something with "AMD for life!" (or maybe not).

With me, it's AMD for price, chipsets and GPU's (ATI for life? )

Message edited by yipsl on 10-01-2008 at 08:40:38 AM

------------------------------ Phenom 8750, ASUS M3A78T
4 gigs Kingston DDR2 800 two 1T SAMSUNG HD103UI
Sapphire 4870x2, Sony BDU-X10S BD-ROM
Antec Neo 650 PSU Antec Nine Hundred, Acer H213H 1080p LCD
Reply to yipsl

Benchies to come.

Message quoted 1 times

Reply to randomizer

You shameless hyper you heheh

------------------------------ I went drifting, thru the capitols of tin, where men cant walk and cant freely talk, and sons turn their fathers in
Reply to jaydeejohn

Thats true, it was stretched to mean more than a few apps, but still a disappointment in overall performance that PEOPLE WANTED. They let the people down, add the tlb, and the rest id history. If its 20% accross the board this time, which is more likely, as there isnt the great changes seen in the arch, just like i7's great change was mostly aimed at server, it will give a decent showing.
Some people are tired of Intel ocing and want something different, and if this performs, Im thinking itll have a positice effect. A new way of doing things, at an affordable price. If this all comes true, AMD has their foot wedged tightly in the door.

------------------------------ I went drifting, thru the capitols of tin, where men cant walk and cant freely talk, and sons turn their fathers in
Reply to jaydeejohn

Reply to sedaine

courtosey xtreview

Reply to sedaine

20% that is a very good number.

With Upgraded Hypertransport they should continue to lead in the server market. Nehalem this, nehalem that, ever compared Xeon vs Opteron pricing ? Nehalem will be good, but i doubt it will have a large acceptance. People in the server bussiness don't upgrade their plataforms 6 in 6 months. More like 3 to 5 years cycle.

So pricing/performance will still be a question. In Virtualization, well, we still have to test them. Server benchies are far more complicated than 3dmark or Sandra. So im not worry here.

On the Desktop side, Nehalem isn't a desktop chip. Sorry guys. Most of you that will buy nehalems, will by to see some many CPUs (logical) on windows. Windows Kernel can't juggle too well past 4 CPUs, and multi threaded apps won't appear from scratch momentarily just because you bought that CPU.

And don't forget that are appearing apps that are using the GPU also. So having a massive CPU might not do any good.

20% on Shanghai ? Possible. Kuma already showed a nice bump. Not that i'm buying a CPU soon.

Edit: Nice graphs, seems that the 20% bump is real.

Message edited by radnor on 10-01-2008 at 01:36:18 PM

------------------------------ Rock journalism is people who can't write interviewing people who can't talk for people who can't read - Frank Zappa
Reply to radnor

Looks like 7- maybe 20%. But in some needed areas its higher than 7%. Looks as tho theyve shored it up nicely. Main thing is core speeds at this point, plus any other optimisations thatll hopefully come with final shipping stepping.

Message quoted 1 times

------------------------------ I went drifting, thru the capitols of tin, where men cant walk and cant freely talk, and sons turn their fathers in
Reply to jaydeejohn

jaydeejohn wrote :

Looks like 7- maybe 20%. But in some needed areas its higher than 7%. Looks as tho theyve shored it up nicely. Main thing is core speeds at this point, plus any other optimisations thatll hopefully come with final shipping stepping.

Looking at those graphs at 2.3Ghz we see a 17% to 25% improvement. I call that a decent speed bump jay.

Message quoted 1 times

------------------------------ Rock journalism is people who can't write interviewing people who can't talk for people who can't read - Frank Zappa
Reply to radnor

Amiga500 wrote :

How the f**k do you know?

You don't have a clue about fabrication yet you happily second guess those that work in the area?

Easy there bro! Set phasers to stun, not kill! Very few of us have an 'inside' perspective on chip fabs, and any of us that do should not be posting here (NDA). Jimmy is probably making prediction based off of the perception that AMD hasn't got 65nm right, so going even smaller may pose more issues. Keep in mind that an entirely new product line was born because of 65nm manufacturing issues; the Defectoron\Tripple Cripple\Tri-Core\X3.

My gut tells me that AMD has everything right this time, because they can't afford another disappointment. Seriously, this may be AMD's last chance to survive. They know that, and they know that the military would consider this a zero fail mission. That's why I have high hopes, but I won't be exhaling for AMD until I see some solid third party benchmarks. And even then, I want to see this stuff shipped in large quantities at advertised speeds. We all remember AMD showing off 3 Ghz Phenom's before launch, and guess what, as much as Baron Matrix said there would be 3 ghz Phenoms in January of this year it still has yet to come even close to happening, and I wouldn't be surprised if Agena never makes it to 3Ghz.

Amiga500 wrote :

People did the same with Barcelona/Phenom.

AMD stated that there would be up to 40% difference between Barcelona and Clovertown clock-for-clock. Guess what, in certain applications, it is far more than 40%.

But people took that to read across the board and including the desktop parts.

Hopefully AMD has learned from this mistake. One thing I've learned in life is to UNDERPROMISE and OVERDELIVER. I do it at work all the time. When something is requested I give myself a buffer of time to get it done. If I say I'll have it done tomorrow noon, I'll be planning on doing it today.

So if AMD promises 40% clock for clock they better not mean in extremely specific workloads that aren't always typical. When there are expectations and investors involved you shouldn't advertise something that won't be perceived as true.

When AMD throws around percentages like this they need to be speaking in average and general terms, not cherry picked benchmarks.

Message edited by TechnologyCoordinator on 10-01-2008 at 02:05:57 PM

------------------------------ jennyh wrote: AMD break-even Q4 2009. *Gauranteed*

RabidFanboysSpreadingFalse.Info
Reply to TechnologyCoordinator

Different source... seems to tally with other results.

[img]http://img.inpai.com.cn/2008/8/5/74c966f3-d032-4f81-b3fc-82c11f710554.gif[img]

[/img]http://img.inpai.com.cn/2008/8/5/57e23f7f-3ad2-4024-baba-52bc32f8dbaf.gif[/img

Quake Wars

Crysis

Blu-Ray (Pirates O. T. C)

Blu-Ray (007 - CR)

Idle Power

Power Load

Reply to sedaine

I like the gaming numbers, not so much the encoding numbers. Better is better tho. To see this jump in Crysis is phnominal heheh. If it wwere HL2 I wouldnt be as impressed, as these numbers should be higher for that game

------------------------------ I went drifting, thru the capitols of tin, where men cant walk and cant freely talk, and sons turn their fathers in
Reply to jaydeejohn

This is for the above ES Deneb Tested.

Full Report is HERE:

http://translate.google.com/transl [...] h-CN&tl=en

Reply to sedaine

sedaine,

Thanks for posting up all those pics, you rock.

I'm actually getting kind of excited about this launch!

------------------------------ jennyh wrote: AMD break-even Q4 2009. *Gauranteed*

RabidFanboysSpreadingFalse.Info
Reply to TechnologyCoordinator

randomizer wrote :

Benchies to come.

http://i176.photobucket.com/albums [...] b_ES_1.png

They really aren't messing around with the numbers game...20550? I wonder what happened with four numbers being adequate. :??:

------------------------------ Antec P182, i7 920 3.7Ghz @ 1.3V, Xigmatek 1283, Asus P6T X58, 3 x 2048MB OCZ Plat DDR3 1600 RAM, 2 EVGA GTX260 Core 216 in SLI, WD 160gb,320GB 1TB WD Black. Corsair 750TX. Acer 24" Monitor. Vista x64 Home Premium.
Reply to one-shot

Youre getting more for your money heheh

------------------------------ I went drifting, thru the capitols of tin, where men cant walk and cant freely talk, and sons turn their fathers in
Reply to jaydeejohn

radnor wrote :

Looking at those graphs at 2.3Ghz we see a 17% to 25% improvement. I call that a decent speed bump jay.

What plots are you looking at? I see *one* (in the list) that shows a ~20% bump clock-for-clock. The rest are ~5%.

And 5% is not out of line with a process shrink.

Reply to ryman554

Note these are Engineering Samples... but what caught my eye was load power consumption. That is almost 50% that used by current 2.3ghz phenom.

They managed to overclock the 2.3ghz CPU to 3.2ghz. Impressive... hopefully not cherry picked.

I'll be the first to say this... AMD is back (if not cherry picked).

Message quoted 1 times

Reply to sedaine

I'd say the Deneb numbers might look a bunch better if they got the HT up to 2600MHz and ran DDr2 1066.

And since they used the MSI K9A2 790FX - disk I/O will get a nice bump with a sb750 mobo over the sb600.

edit: yup - that power consumption drop looks to be great

Message edited by wisecracker on 10-01-2008 at 04:14:59 PM

Reply to wisecracker

The ones that aren't displayed properly...

The other thing I picked up from the article is that the chips will allow you to over-clock by 1ghz and realise a 40% performance boost.

I had not realised that they also think the 3.2ghz part will over-clock by the same margin i.e 4.0ghz plus.

Message quoted 1 times

Reply to sedaine

epsilon84 wrote :

Crysis, like most games, perform much better on CPUs with big caches (at least thats the case with Core 2), I think thats why we see a relatively large jump in Crysis performance compared to other benches, I'd say the jump from a 2MB L3 to a 6MB L3 would have a large say in that.

I'm not sure I'm entirely conviced of that. I think that Crysis will see a lot big difference on high performance CPUs. Most games seems to bottleneck on the GPU, but Crysis with its physics (especailly on the high settings) utilized the CPU a lot.

I'm not convinced it has to do with large cache; I'm sure more cache helps, but I don't know if it is the deciding factor.

------------------------------ jennyh wrote: AMD break-even Q4 2009. *Gauranteed*

RabidFanboysSpreadingFalse.Info
Reply to TechnologyCoordinator

epsilon84 wrote :

I wouldn't bank on it... look at the benches from the previous page, does that look like 20% higher performance per clock to you? Only in ONE bench does it come close to the 20% mark (POV-Ray) which is why I say its possible in a best case scenario but it won't be a 20% improvement across the board. If you average out the results from the Hardspell Deneb preview we end up with a ~7% improvement per clock.

7% would be awesome for a shrink. Combine that with (hopefully) higher clocks and AMD might be back in the game. I don't think they'll be able to take the crown, but if they can even just get back into the $500 range it would do wonders for them.

I notice the benchies showing the 45nm @3.2ghz. I can only hope that they can hit these speeds in quantity! Remember last year when AMD was showing off cherry picked Agena(65nm Phenom) chips at 3ghz just before launch (with a fist full of NDA for the press)?

------------------------------ jennyh wrote: AMD break-even Q4 2009. *Gauranteed*

RabidFanboysSpreadingFalse.Info
Reply to TechnologyCoordinator

BaronMatrix wrote :

But again, with C2 being launch the rumors about a 4GHz chip may not be so far-fetched.

At what kind of TDP? 250W with a Vapochill as standard? :lol:

Message edited by epsilon84 on 10-01-2008 at 08:17:29 PM

Reply to epsilon84

Forum CPU & Components : CPUs - Barcelona surrenders to Shanghai: 45nm AMD heart beating 20% faster?

Please mind