AT Barcelona previews are up

[NoSoup4You]

I'm waiting until a huge game is released that I can't run at high settings strictly because I'm cpu-limited. Maybe Crysis will be that game. If it is, a G0 Q6600 will be on it's way to my door through First Class shipping the next day.

Unfortunately, it doesn't look like K10 will be worth checking out. At least not at this point...

 

[Phynaz]

Originally posted by: Viditor

Originally posted by: covert24
i love how people are making assumptions on phenom before it has even came out. hope everyone know as people have said that they are using a barcy proc to give a general performance idea, so all you people saying phenom is gunna blow need to chillax and lay off the keyboard since apparently you have no idea what day it is.

Thank you...this is exactly what I've been saying for awhile now. There are so many differences between Phenom and Barcy that I honestly think it was a mistake for Anand to post the article (though I do understand the pressure to do so...and AMD has been anything but forthcoming).

1. Phenom must necessarily be a new revision because it will be HT 3.0. That can introduce many other variables besides just the HT bandwidth (for instance they must have a different memory controller).

2. They already have a 2.5 GHz Barcy, so a 3 GHz Phenom is quite likely...

3. The chipsets and platform will be quite different as well

4. Launch is in 2 months, not 6 months...so it will coincide with Penryn's launch.

5. We are seeing a review of Engineering Samples of Barcy using a first generation bios. Those of you who have worked in the server world know exactly what I mean here...
It's true that AMD is rushing this out and this is probably their worst marketed launch ever, but drawing conclusions about the technology because of that is a BIG mistake.

6. None of the review sites has had much of a chance yet to work on these new systems (they only arrived a few days ago). Think back to previous reviews and recall how many "gotchas" they have found after running their initial series in the past...

1. MUST have a new memory controller? Hardly, since HT has nothing to do with memory. HT 3.0 will provide no benefit to a desktop anyway.

2. Maybe at some point, but is not currently on any roadmap.

4. Maybe, maybe not. The review systems shipped with Nvidia chipsets. What's Nvidia's chipset plans?

5. If these are not representative of shipping products, then you don't show them. This is a very big deal for AMD, you want to show themselves in the best possible light. That would make this the best they can do for the near future.

An observation I have made is that it is clear now Why Henri left. He knew he had a problem promoting these things. Cnet has very nice article calling AMD to task for hype they didn't live up to here.

Here's a quote:
"But AMD will not deliver--at least not yet--on promises made by Randy Allen, corporate vice president of AMD's server and workstation division, in January. "We expect across a wide variety of workloads for Barcelona to outperform Clovertown by 40 percent," Allen said. In May, Allen told reporters that Barcelona "will be the highest-performing x86 chip out there. It will blow away Clovertown."

There was no proof to those statements in the test results AMD distributed ahead of the Barcelona launch. In its briefing materials, the company touted only benchmark results that emphasized floating-point performance and memory bandwidth, which have always been strengths of the Opteron processor but do not cover the entire spectrum of the server market."

I think Henri had a personal problem with telling lies to the public - which AMD has done. He has the integrity to quit his job instead continuing to do so.

 

[Phynaz]

Originally posted by: Viditor

I would disagree with your conclusion here...Intel (rather blatantly) used their dominant position to prevent OEMs from using AMD as a second supplier for many years. This is what the anti-trust suit is all about.

Let me point out (again) that this is an allegation, not fact.

 

[Phynaz]

Originally posted by: Amaroque

Anand just knew that most of us don't give a crap about servers and care much more about desktop systems.

That's not true. Their is alot of IT professionals that read AT.

Depends upon your definition of IT professional. Is that somebody that works in IT? Then yes, there are many readers.

If you mean people in corporate IT who are resonsible for server purchasing, that's a much, much smaller number.

 

[firewolfsm]

Everyone keeps forgetting, the new chipsets by AMD and nVidia will support AM2+, which means lower power consumption and independent overclocking and voltage of the memory controller and HT links. It should improve performance A LOT.

 

[bryanW1995]

Originally posted by: Phynaz

Originally posted by: Viditor

Originally posted by: covert24
i love how people are making assumptions on phenom before it has even came out. hope everyone know as people have said that they are using a barcy proc to give a general performance idea, so all you people saying phenom is gunna blow need to chillax and lay off the keyboard since apparently you have no idea what day it is.

Thank you...this is exactly what I've been saying for awhile now. There are so many differences between Phenom and Barcy that I honestly think it was a mistake for Anand to post the article (though I do understand the pressure to do so...and AMD has been anything but forthcoming).

1. Phenom must necessarily be a new revision because it will be HT 3.0. That can introduce many other variables besides just the HT bandwidth (for instance they must have a different memory controller).

2. They already have a 2.5 GHz Barcy, so a 3 GHz Phenom is quite likely...

3. The chipsets and platform will be quite different as well

4. Launch is in 2 months, not 6 months...so it will coincide with Penryn's launch.

5. We are seeing a review of Engineering Samples of Barcy using a first generation bios. Those of you who have worked in the server world know exactly what I mean here...
It's true that AMD is rushing this out and this is probably their worst marketed launch ever, but drawing conclusions about the technology because of that is a BIG mistake.

6. None of the review sites has had much of a chance yet to work on these new systems (they only arrived a few days ago). Think back to previous reviews and recall how many "gotchas" they have found after running their initial series in the past...

1. MUST have a new memory controller? Hardly, since HT has nothing to do with memory. HT 3.0 will provide no benefit to a desktop anyway.

2. Maybe at some point, but is not currently on any roadmap.

4. Maybe, maybe not. The review systems shipped with Nvidia chipsets. What's Nvidia's chipset plans?

5. If these are not representative of shipping products, then you don't show them. This is a very big deal for AMD, you want to show themselves in the best possible light. That would make this the best they can do for the near future.

An observation I have made is that it is clear now Why Henri left. He knew he had a problem promoting these things. Cnet has very nice article calling AMD to task for hype they didn't live up to here.

Here's a quote:
"But AMD will not deliver--at least not yet--on promises made by Randy Allen, corporate vice president of AMD's server and workstation division, in January. "We expect across a wide variety of workloads for Barcelona to outperform Clovertown by 40 percent," Allen said. In May, Allen told reporters that Barcelona "will be the highest-performing x86 chip out there. It will blow away Clovertown."

There was no proof to those statements in the test results AMD distributed ahead of the Barcelona launch. In its briefing materials, the company touted only benchmark results that emphasized floating-point performance and memory bandwidth, which have always been strengths of the Opteron processor but do not cover the entire spectrum of the server market."

I think Henri had a personal problem with telling lies to the public - which AMD has done. He has the integrity to quit his job instead continuing to do so.

henri is a job hopper. he stayed longer at amd than he ever has at another job. my experience with job hoppers is that they typically make good resume-builders and poor employees. His marketing dept was always substandard at best. Also, he tried to quit last year and for some crazy reason they convinced him to stay for one more year, which is exactly what he did.

btw, since when are marketing people bothered by ethics?

 

[Idontcare]

Originally posted by: Phynaz
An observation I have made is that it is clear now Why Henri left. He knew he had a problem promoting these things. Cnet has very nice article calling AMD to task for hype they didn't live up to here.

Here's a quote:
"But AMD will not deliver--at least not yet--on promises made by Randy Allen, corporate vice president of AMD's server and workstation division, in January. "We expect across a wide variety of workloads for Barcelona to outperform Clovertown by 40 percent," Allen said. In May, Allen told reporters that Barcelona "will be the highest-performing x86 chip out there. It will blow away Clovertown."

There was no proof to those statements in the test results AMD distributed ahead of the Barcelona launch. In its briefing materials, the company touted only benchmark results that emphasized floating-point performance and memory bandwidth, which have always been strengths of the Opteron processor but do not cover the entire spectrum of the server market."

I think Henri had a personal problem with telling lies to the public - which AMD has done. He has the integrity to quit his job instead continuing to do so.

I 100% agree with the spirit of Phynaz's message.

Henri of all people would have precisely known the shite was about to hit the fan and he clearly negotiated his career path so as to minimize his exposure to the downwind-side of said fan.

I really do not understand what the hell AMD's design team has been doing all these years. This K10 appears to be a warmed-over K8, hardly a step above what would be classified as a refresh for anyone else.

Look at Penryn versus Conroe...talk about spending as few resources as possible, hardly what anyone would call a refresh, and they bag 5% improvement. AMD spends years on taking K8 to K10 and squeaks out a 15% IPC improvement on what desparately needs to a microarchitecture representative of their finest hour?

All I can say is YIKES! If you attempted to intentionally engineer a worse outcome for AMD in their execution and planning of the K10, how could it be worse than the reality they have so deftly created?

 

[leexgx]

this time round unlike intel core2 was bosted that it be Very fast abd bad for AMD it was and thay lost an big market when that happend

if one of these chips (X4 @ 2ghz+) can out perform or be about the same as my AM2 5600X2 @3.1ghz i probly get it but i find that unliky whats disapointing, but as i say its best to wait and see what the fully rels product is like or we get what happend with core 2

i like to have an Quad core for the stuff i do but i do not need it yet

 

[Viditor]

Originally posted by: Phynaz

1. MUST have a new memory controller? Hardly, since HT has nothing to do with memory. HT 3.0 will provide no benefit to a desktop anyway.

But cHT does have a lot to do with the memory controller...the coherent links are what allow access of the current CPU's memory channels to the other CPUs. In the case of a single socket Opteron, there are still the cHT links to other devices (Torrenza)...
So yes, it MUST have a new memory controller.

2. Maybe at some point, but is not currently on any roadmap.

Neither is most of the Phenom line...in fact the 2.5GHz Barcy that they shipped wasn't on the roadmaps either...

4. Maybe, maybe not. The review systems shipped with Nvidia chipsets. What's Nvidia's chipset plans?

Do you mean will they be shipping desktop boards with server chipsets? No...

5. If these are not representative of shipping products, then you don't show them. This is a very big deal for AMD, you want to show themselves in the best possible light. That would make this the best they can do for the near future.

As I said, this was a terribly marketed launch...but more to the point, these chips are for servers, which means they will of necessity be going through a 6 month+ qualification phase anyway (the same as Woodcrest, Opteron, and every other server chip).

An observation I have made is that it is clear now Why Henri left. He knew he had a problem promoting these things. Cnet has very nice article calling AMD to task for hype they didn't live up to here.

Henri first gave notice over a year ago and was paid a bonus to stay for a while longer...
If you go through his resume', you will note that he has stayed at AMD almost twice as long as he has with any other company (some of which he was CEO for...).

[/quote]

 

[Phynaz]

Henri first gave notice over a year ago and was paid a bonus to stay for a while longer...

News to me...Got a link?

 

[Viditor]

Originally posted by: Phynaz

Originally posted by: Viditor

I would disagree with your conclusion here...Intel (rather blatantly) used their dominant position to prevent OEMs from using AMD as a second supplier for many years. This is what the anti-trust suit is all about.

Let me point out (again) that this is an allegation, not fact.

So is concluding that OEMs didn't sign a second source because it was "easier"...
Personally, I think my hypothesis is much more reasonable (and has been attested to by many...including Anand in his blogs at the time).

 

[Phynaz]

But cHT does have a lot to do with the memory controller...the coherent links are what allow access of the current CPU's memory channels to the other CPUs. In the case of a single socket Opteron, there are still the cHT links to other devices (Torrenza)...
So yes, it MUST have a new memory controller.

Please see the architecture diagram of Barcelona Here. HT links are not directly connected to the memory controller, nor have they ever been in any AMD cpu.

 

[Regs]

Originally posted by: keysplayr2003

His point (I think) was that a rather large percentage of PC users do not need the best of the best. For what most people do with their PC's, a Celeron or Sempron is waaaaay more than enough and perhaps overkill. Depends on what they do. If all they do is browse the web and watch movies, Play their mp3's, burn CD's DVD's, etcetera, they don't need crazy super powered juggernaut CPU's. Gamers "might" have an occasion to need such power.
Distributed Computing folks like Mark, Duvie and others may need the most power they can get their hands on. Points are points after all .

The old saying is more is not always better but just more. However in today's buyers world, people who make over 100k a year will take out a 30 year loan just so they can spend money on a 6 million dollar house.

It's the marketers job to sell you a need with an item and not just an item. How many things have you bought in the past that you realized you didn't really need? I can think of plenty, like a new car.

We should stop kidding ourselves and realize that there is a performance craze out there. If people don't want or need performance, they want what's new and exciting and the Penryn and Core Duo are precisely that. It drives prestige and attention to the product.

 

[gOJDO]

Originally posted by: Viditor
But cHT does have a lot to do with the memory controller...the coherent links are what allow access of the current CPU's memory channels to the other CPUs. In the case of a single socket Opteron, there are still the cHT links to other devices (Torrenza)...
So yes, it MUST have a new memory controller.

Jesus. If you know nothing it is better to be quiet. The memory controller has nothing to do with HTT, cHT links or Torrenza. The controller on the Phenom is exactly the same as the one on Barcelona! It is capable for running registered and unregistered ECC and non-ECC DDR2.

Neither is most of the Phenom line...in fact the 2.5GHz Barcy that they shipped wasn't on the roadmaps either...

The 2.5GHz Barcy was an B2 ES with 1.52v V_core, hot as hell(almost 200W). We need to see 2.5GHz with reasonable TDP, before we can talk about higher clocked K10. Their new "wattage" method just proves the K10's high power consumption and heat dissipation.


BTW, What happened to your .5 million $ bet on K10 "outstanding" performance?

 

[Viditor]

Originally posted by: Phynaz

But cHT does have a lot to do with the memory controller...the coherent links are what allow access of the current CPU's memory channels to the other CPUs. In the case of a single socket Opteron, there are still the cHT links to other devices (Torrenza)...
So yes, it MUST have a new memory controller.

Please see the architecture diagram of Barcelona Here. HT links are not directly connected to the memory controller, nor have they ever been in any AMD cpu.

If you look at that diagram again, they are...through the crossbar (which is why it's called DCA).

Edit: Let me flesh this out a bit...the cHT request can go directly to the memory controller or the cache, depending on the flag given to the SRI. This is all done through the crossbar switch (notice that this is a switch and not a hub like the MCH).

 

[PetNorth]

Originally posted by: gOJDO

Originally posted by: Viditor
But cHT does have a lot to do with the memory controller...the coherent links are what allow access of the current CPU's memory channels to the other CPUs. In the case of a single socket Opteron, there are still the cHT links to other devices (Torrenza)...
So yes, it MUST have a new memory controller.

Jesus. If you know nothing it is better to be quiet. The memory controller has nothing to do with HTT, cHT links or Torrenza. The controller on the Phenom is exactly the same as the one on Barcelona! It is capable for running registered and unregistered ECC and non-ECC DDR2.

Neither is most of the Phenom line...in fact the 2.5GHz Barcy that they shipped wasn't on the roadmaps either...

The 2.5GHz Barcy was an B2 ES with 1.52v V_core, hot as hell(almost 200W). We need to see 2.5GHz with reasonable TDP, before we can talk about higher clocked K10. Their new "wattage" method just proves the K10's high power consumption and heat dissipation.


BTW, What happened to your .5 million $ bet on K10 "outstanding" performance?

How do you know 1.52v is correct? I've seen a lot of errors (and BIG) with CPU-z reading vcore (a lot, and BIG difference with real vcore in cpus launched months ago -X2 or C2D-, so I can't imagine how incorrect can it be with one new).

 

[Pederv]

I'll probably get a Phenom, sometime next year or so.
Am I a fanboi? I don't consider myself one.
I do have name brands that I usually purchase, whether it's my Darigold milk, my Ford car or my AMD CPU. Do I stop buying Darigold because I get a bad gallon? No I just buy another gallon of milk.

 

[Keysplayr]

Originally posted by: Pederv
I'll probably get a Phenom, sometime next year or so.
Am I a fanboi? I don't consider myself one.
I do have name brands that I usually purchase, whether it's my Darigold milk, my Ford car or my AMD CPU. Do I stop buying Darigold because I get a bad gallon? No I just buy another gallon of milk.

I don't think I've ever bought a CPU that was past it's expiration date.

 

[BigDH01]

Any ideas as to when we might see a more extensive review from Anandtech? There doesn't seem to be much information out there about a product that launched yesterday.

 

[Pederv]

Originally posted by: keysplayr2003

Originally posted by: Pederv
I'll probably get a Phenom, sometime next year or so.
Am I a fanboi? I don't consider myself one.
I do have name brands that I usually purchase, whether it's my Darigold milk, my Ford car or my AMD CPU. Do I stop buying Darigold because I get a bad gallon? No I just buy another gallon of milk.

I don't think I've ever bought a CPU that was past it's expiration date.

Good one Keys.

My wife didn't look at the expiration date on products until she saw me making a point to do so. As a result, the bad gallons are few and far between.

Just as my auto maker of choice went from AMC to Ford, when AMC went out of buisiness. My CPU manufacturer of choice will change if AMD goes out of buisiness, no biggie. Until that happens the CPU I'll purchase is what AMD has that I can afford, when I can afford it. It's not like AMD is going to charge me twice the price for the same performance as Intel would. I just won't have the fastest system on the planet, again no biggie.

 

[Phynaz]

Originally posted by: Viditor

Originally posted by: Phynaz

But cHT does have a lot to do with the memory controller...the coherent links are what allow access of the current CPU's memory channels to the other CPUs. In the case of a single socket Opteron, there are still the cHT links to other devices (Torrenza)...
So yes, it MUST have a new memory controller.

Please see the architecture diagram of Barcelona Here. HT links are not directly connected to the memory controller, nor have they ever been in any AMD cpu.

If you look at that diagram again, they are...through the crossbar (which is why it's called DCA).

Edit: Let me flesh this out a bit...the cHT request can go directly to the memory controller or the cache, depending on the flag given to the SRI. This is all done through the crossbar switch (notice that this is a switch and not a hub like the MCH).

Jeez, think man!!

There is a switch between the HT lanes and the MC. They are not connected. Changing one does not affect the other.

Got it yet?

 

[Arkaign]

Originally posted by: Pederv

Originally posted by: keysplayr2003

Originally posted by: Pederv
I'll probably get a Phenom, sometime next year or so.
Am I a fanboi? I don't consider myself one.
I do have name brands that I usually purchase, whether it's my Darigold milk, my Ford car or my AMD CPU. Do I stop buying Darigold because I get a bad gallon? No I just buy another gallon of milk.

I don't think I've ever bought a CPU that was past it's expiration date.

Good one Keys.

My wife didn't look at the expiration date on products until she saw me making a point to do so. As a result, the bad gallons are few and far between.

Just as my auto maker of choice went from AMC to Ford, when AMC went out of buisiness. My CPU manufacturer of choice will change if AMD goes out of buisiness, no biggie. Until that happens the CPU I'll purchase is what AMD has that I can afford, when I can afford it. It's not like AMD is going to charge me twice the price for the same performance as Intel would. I just won't have the fastest system on the planet, again no biggie.

That goes beyond idiotic, to purchase based on brand in the computer world. It would have meant that you were using the crap K5 when the Pentium was kicking it's butt. Or using a PMMX when K6-2 was walking all over it. Or a K6-3 when The Celeron 300a@450 made it look like a 386, or perhaps worse.

The *only* logical choice is to pick the best deal at the time that you buy.

EDIT : Removed needlessly hostile remark, sorry for that btw

 

[Arkaign]

Originally posted by: keysplayr2003

Originally posted by: Arkaign

Originally posted by: fishmonger12
Someone may have already said this, but on the desktop side, few people really NEED the power of a core 2 duo cpu. I would say gamers and other enthusiasts are the only ones that actually take advantage of a chip like that, and they make up what, like 10% of the market? I just bought my girlfriend a laptop with a 1.9 ghz turion x2 and 1 gig of ram, and it plows right through vista, multitasking, movies, etc. Which is what most of the market uses a pc for.

So I don't think this as big of a disappointment as people paint it to be, nor will it have a large effect on the business side of amd in the desktop space. It will matter in the server space, and guess what, it's actually competitive there.

Yeah, because everyone wants to buy the 2nd best processor :roll:

With C2Ds already so affordable (quad-core for under $300, top-notch dual-core for $100ish), I don't see the point of bringing out something worse than what they already have (3.2Ghz K8).

His point (I think) was that a rather large percentage of PC users do not need the best of the best. For what most people do with their PC's, a Celeron or Sempron is waaaaay more than enough and perhaps overkill. Depends on what they do. If all they do is browse the web and watch movies, Play their mp3's, burn CD's DVD's, etcetera, they don't need crazy super powered juggernaut CPU's. Gamers "might" have an occasion to need such power.
Distributed Computing folks like Mark, Duvie and others may need the most power they can get their hands on. Points are points after all .

I do agree with that assessment, but question the value of releasing a new product that is not much better than the old one, and certainly worse than the competitor's offering. I mean, a 2.4Ghz Phenom probably won't outpace a hypothetical 4-core 3.0/3.2Ghz K8, so why bother? They should have scrapped K10 and gone for something better. For those who didn't need 'the best', they could have continued to sell K8. As is, K10 isn't going to help them raise their ASP, as the faster Intel products are so cheap.

Look for AMD to lose another 500-800 million dollars next quarterly report.

 

[Regs]

Even if Barcelona was going to be a major hit Arkaign, AMD was going to lose money anyway. They won't have these things in channels for at least another month and they would likely have to sell out of them 3x over to cover enough expenses making them in the FABs.

 

[Arkaign]

Hmm well I bet the old rumors about IBM or the like buying what's left of AMD have a higher probability of happening now.