Is Barcelona in trouble?

[Nathelion]

It has definitely been stated that AM2 boards will be supported - albeit with a somewhat impaired functionality, such as no HT3

 

[Sable]

Originally posted by: tuteja1986
No.. thats anandtech being angry with AMD as whole. I remember many article of Opertron in trouble and they were running at very low speed just few months before launch.

If it was just Anand saying Barca was in trouble then maybe you'd have a point but when I see the same thing reported on one of the least biased tech sites I tend to get a bit suspicious.

http://www.techreport.com/onearticle.x/12640

 

[harpoon84]

Originally posted by: Hulk
I don't understand the statement that "K10 will come alive at around 2.6GHz."

"Come alive?"

The best a processor can do is 1:1 scaling as clockspeed increases AFAIK right?

So scaling clockspeed up will show what potential performance at higher clocks will be.

It may be the clockspeed AMD would require to be competitive or exceed current Intel performance levels... at least in the short term. Of course, Intel is not standing still either.

 

[lopri]

Yup. Barcelona doesn't look good. And AMD will be in trouble starting with the 2nd quarter financial results. As much as I hoped the Barcelona to be a 2nd-coming of Jebus for AMD, things are getting clearer day by day. I do indeed worry the consequences will not be for everybody's (even for Intel stock holders') interest. The only relieving (at least for now) factor is that without a revolutionary break-through in programming, today's CPUs are pretty much overkill for every other desktop application. In other words, today's desktop applications are no where near close to take advantage of what AMD/Intel are offering. People like to blame (argue) Intel on having sit on Netburst in the past and AMD on K8, but if you take a moment and think about it - did your everyday applications improve or change to the degree of how much CPUs have evolved?

My personal answer is a resounding no. I use the same/similar applications that I used on P4 3.0GHz / i850 / RDRAM 5 years ago and probably will use the same apps the same time next year. I think one of the reasons that Intel is busy presenting updated road maps as of late is this very fear: Until there is a major breakthrough in application coding, vast majority of computer users may think there is no need for faster CPUs. For example, think about MS Office 2000 vs MS Office 2007. (yes, 7 years of gap) Will an average user benefit much by upgrading from, say, PIII 733MHz to C2Q 6600? By how much? If you happen to read latest tech news, UMPC has been getting a huge spotlight and guess what, the CPU in that thing is like something like PIII 733MHz. (OK, maybe a little faster than that. But you get the point)

Now it makes it easier to understand why Intel's been pushing multi-cores (making demos, showing off a game that uses 4 cores, etc.). It's not just about competing against AMD. It's about the fear that people won't consider buying faster/more-core CPUs because they're content with what they have now.

What will happen to AMD after the failed (as in not competitive enough) launch of Barcelona? I don't know for sure. But I am pessimistic and at the same time optimistic. Will I remain as a hardware enthusiast without fierce competition, and innovation that comes with it? Probably not. (Just look at the GPU market today) But at least I have a feeling that I won't get charged ridiculous amount of dollars for a CPU in anytime soon. Even with AMD out of the picture.

 

[GFORCE100]

AMD has problems with its Barcelona/Agena period. There is no sound reason for AMD's secrecy act that's been going on for all too long now. AMD didn't spare one opportunity to kick Intel's Pentium 4/Pentium D offerings in the corner claiming how superior their K8's really were not so long ago, they even did a pathetic marketing competition asking Intel to a duel and when Intel refused, made fun of them giving 10 reasons why Intel didn't agree. It all started with the AMD Me theme. Honestly, if AMD had the goods they would have made a real loud impression of the fact. 2007 will not be a good year for AMD at all and it's best if all just face these facts instead of beating around the bush. Even if AMD does eventually push out something so so competetive then comes Intel with their 45nm mass production and its panic stations all over again. This is why leaving a product launch too late hurts because every day you delay a product, is one more day of profits for your competitor(s) and opportunity to be one up on you in terms of terchnology R&D. AMD thought too much of itself and hence the ego of recent years. They should remember the scale of their company in comparison to Intel and focus on that because quite frankly, if it wasn't for the monopoly lawsuits waiting in the wings, Intel could destroy AMD either by a) a price war, b) marketing or c) its product portfolio.

Intel made a mistake with their Prescott Netburst in 2003 (should have stayed with evolving Northwood instead) but needed to wait until a 5 year evolution gap elapsed (Netburst 2001-2006) in order to justify to investors massive R&D into a new architecture and strategy for the next 5 years or so. If a company changes its operating strategy too often, investors otherwise see this as an indication that there's some uncertainty in the company. Also, spend too much on R&D too often, and you're eating into your profits which won't please investors either. It's a balancing act. However once this window of opportunity surfaced they bit the bullet and then came Conroe/Merom/Woodcrest and so forth. Intel used its past experience to strike back because for the technically minded, these Core 2 Duo's and the like are actually based off what was the P6/Klamath/Deschutes/Katmai/Coppermine/Tualatin/Banias/Dothan/Yonah architectures with some new enhancements. In some ways today's Conroe actually dates back to 1995's Pentium PRO known otherwise as the P6.

 

[sdsdv10]

Originally posted by: lopri
(snip) did your everyday applications improve or change to the degree of how much CPUs have evolved?
(snip)

Hey lopri, did you see this article. http://hubpages.com/hub/_86_Ma..._Wont_Believe_Who_Wins

They compared a '86 Mac Plus to a '07 AMD Dual core machine in various MS Word operations. Guess which one won! ;-)

Now before anyone gets their underwear in a bunch, I realize the comparison was done a little ?tongue in cheek?, but I think it illustrates lopri's point. In the productivity arena (i.e. MS Office), the advances in hardware have not been matched with advances in software. Yeah, they have added more and more features to Word, etc., but the underlying code has not improved to the same extent. Games, well that is an entirely different story. However, the vast majority of people don?t make a living playing games at work. The bulk do use some type of productivity software in their day to day job.

 

[lopri]

sdsdv10: checked out the link. Interesting comparison. Thanks.

 

[BigDH01]

Originally posted by: lopri
sdsdv10: checked out the link. Interesting comparison. Thanks.

Agreed . I wonder how linux compares. Maybe Damn Small Linux.

 

[jazkat]

Barcelona thermal dissipation is fine at high clocks Even over 2.8 GHz


Our knowledgeable sources have confirmed that Barcelona is going to be an overclockable part. Even at 2.8 GHz to 3 GHz this processor won't have any TDP dissipation limitations. TDP stands for Thermal Design Power and is a very important figure as this is the power that has to be removed from a CPU. You need to cool this thing of to make it stable.

The new revision that a few sources referred as B1 shouldn't have any TDP problems and it can work up to 2.8 if not higher to meet the standard TDP numbers.

The main problem now is to get enough of these chips out on the market and even if AMD launches Barcelona in Q3 we are talking about a few chips in late September. Realistically this chip is Q4

thought id share this for those interested , i also read somewhere there is a b2 revision but
b1 will be good enough and be able to hit decent clock speeds, i think the b2 revision will be able to achieve even higher clocks and still stay within the TDP limit

 

[nyker96]

Originally posted by: jazkat
Barcelona thermal dissipation is fine at high clocks Even over 2.8 GHz


Our knowledgeable sources have confirmed that Barcelona is going to be an overclockable part. Even at 2.8 GHz to 3 GHz this processor won't have any TDP dissipation limitations. TDP stands for Thermal Design Power and is a very important figure as this is the power that has to be removed from a CPU. You need to cool this thing of to make it stable.

The new revision that a few sources referred as B1 shouldn't have any TDP problems and it can work up to 2.8 if not higher to meet the standard TDP numbers.

The main problem now is to get enough of these chips out on the market and even if AMD launches Barcelona in Q3 we are talking about a few chips in late September. Realistically this chip is Q4

thought id share this for those interested , i also read somewhere there is a b2 revision but
b1 will be good enough and be able to hit decent clock speeds, i think the b2 revision will be able to achieve even higher clocks and still stay within the TDP limit

First I'd like to ask how reliable is your source on this?

Second, very nice to know that current Barcelona revision is hitting 2.8 and OCable. If this is true, AMD could have a real winner at hand considering how power efficient Barcelona design is over generic Intel glue solutions at this point.

 

[bryanW1995]

Originally posted by: sdsdv10

Originally posted by: lopri
(snip) did your everyday applications improve or change to the degree of how much CPUs have evolved?
(snip)

Hey lopri, did you see this article. http://hubpages.com/hub/_86_Ma..._Wont_Believe_Who_Wins

They compared a '86 Mac Plus to a '07 AMD Dual core machine in various MS Word operations. Guess which one won! ;-)

Now before anyone gets their underwear in a bunch, I realize the comparison was done a little ?tongue in cheek?, but I think it illustrates lopri's point. In the productivity arena (i.e. MS Office), the advances in hardware have not been matched with advances in software. Yeah, they have added more and more features to Word, etc., but the underlying code has not improved to the same extent. Games, well that is an entirely different story. However, the vast majority of people don?t make a living playing games at work. The bulk do use some type of productivity software in their day to day job.

Gotta run, need a mac to run my office apps faster!

 

[Fox5]

Originally posted by: Hulk
I don't understand the statement that "K10 will come alive at around 2.6GHz."

"Come alive?"

The best a processor can do is 1:1 scaling as clockspeed increases AFAIK right?

So scaling clockspeed up will show what potential performance at higher clocks will be.

Actually, I'm not sure if performance scales 1:1 with clockspeed.

2ghz = 1/2 billion s cycle time (.5 nano seconds per cycle), right? Let's say it takes 5 minutes for something to finish at 2ghz.
2.6 ghz = 1/2.6 billion s = .38 nanoseconds per cycle.

Doing a ratio of
.5/.38 = 5/x, you'd get that x equals 3.8 minutes.

2.6 is 30% higher frequency than 2ghz, but 3.8 minutes is only 24% faster than 5 minutes.
So in actuality, the best a processor can do is a bit less than 1:1 scaling.

 

[dmens]

Originally posted by: Fox5
Actually, I'm not sure if performance scales 1:1 with clockspeed.

2ghz = 1/2 billion s cycle time (.5 nano seconds per cycle), right? Let's say it takes 5 minutes for something to finish at 2ghz.
2.6 ghz = 1/2.6 billion s = .38 nanoseconds per cycle.

Doing a ratio of
.5/.38 = 5/x, you'd get that x equals 3.8 minutes.

2.6 is 30% higher frequency than 2ghz, but 3.8 minutes is only 24% faster than 5 minutes.
So in actuality, the best a processor can do is a bit less than 1:1 scaling.

math is a bit off... frame of reference is switched when you go from frequency to clock cycle.

under ideal (academic, hehe) conditions, it should be a 1:1 speedup.

 

[TuxDave]

Originally posted by: Fox5

Originally posted by: Hulk
I don't understand the statement that "K10 will come alive at around 2.6GHz."

"Come alive?"

The best a processor can do is 1:1 scaling as clockspeed increases AFAIK right?

So scaling clockspeed up will show what potential performance at higher clocks will be.

Actually, I'm not sure if performance scales 1:1 with clockspeed.

2ghz = 1/2 billion s cycle time (.5 nano seconds per cycle), right? Let's say it takes 5 minutes for something to finish at 2ghz.
2.6 ghz = 1/2.6 billion s = .38 nanoseconds per cycle.

Doing a ratio of
.5/.38 = 5/x, you'd get that x equals 3.8 minutes.

2.6 is 30% higher frequency than 2ghz, but 3.8 minutes is only 24% faster than 5 minutes.
So in actuality, the best a processor can do is a bit less than 1:1 scaling.

You have got to be kidding me. You failed because you rounded.

2GHz = 1/2e9 second cycle time that takes 5 minutes to finish something
2.6GHz = 1/2.6e9 second cycle time.

(1/2e9)/(1/2.6e9) = 5/x you'd get that x = 10/2.6

5 minutes / [10/2.6] minutes = 130%
2.6GHz/2GHz = 130%

1:1 assuming clock frequencies alone determined performance.

 

[jazkat]

hi nyker,

well yesterday there was only a handfull of news sites with the information but theres many more today.(google it)

The launch revision B1 barcelona, New Opteron and new desktop Phenom or Agena FX CPUs will be able to hit 2.8 GHz, clocks between 1.9 - 2.8 ghz

Q3/Q4 fx-80: phenom FX/AGENA FX 2.2-2.4, 3.2ghz??? AM2+

the B2 revision budapest is K10 core but with native HyperTransport 3.0 and word is that the fx series should go upto 3.6ghz.

Q1/08 fx-90: PHENOM FX/AGENA FX 2.2-2.6ghz, 3.2ghz 1207+
Q1/08 fx-91: PHENOM FX/AGENA FX 2.4-2.6ghz, 3.6ghz 1207+

it looks like the NEW opteron will be available next month:
http://www.softlayer.com/servers_dpmc_compare.html
check under Opteron 2000 QC Series.

its all i have for now.

 

[lopri]

However they come 'alive' @2.60GHz, it seems pretty clear to me that they will be eaten alive by Q6600 OC'ed by mere 200MHz. ($266 next month)

But herein lies the truth: Every stepping is extremely important for AMD. Without such extravagant financial surprlus, AMD has no choice but betting on the every new revision. If one stepping hits high clock frequencies, that'll be the shipping/flagship stepping. It's sad but that's where AMD is now.

 

[coldpower27]

Originally posted by: jazkat
hi nyker,

well yesterday there was only a handfull of news sites with the information but theres many more today.(google it)

The launch revision B1 barcelona, New Opteron and new desktop Phenom or Agena FX CPUs will be able to hit 2.8 GHz, clocks between 1.9 - 2.8 ghz

Q3/Q4 fx-80: phenom FX/AGENA FX 2.2-2.4, 3.2ghz??? AM2+

the B2 revision budapest is K10 core but with native HyperTransport 3.0 and word is that the fx series should go upto 3.6ghz.

Q1/08 fx-90: PHENOM FX/AGENA FX 2.2-2.6ghz, 3.2ghz 1207+
Q1/08 fx-91: PHENOM FX/AGENA FX 2.4-2.6ghz, 3.6ghz 1207+

it looks like the NEW opteron will be available next month:
http://www.softlayer.com/servers_dpmc_compare.html
check under Opteron 2000 QC Series.

its all i have for now.

I don't see 3.6GHZ on 65nm at all, your talking about HyperTransport speeds, which chipset to CPU communication link.