AMD Unveils "Barcelona" Architecture

[Nagasama]

ooo..........i c. Do u think that i should get 1066 instead of 800 or stick with 800 n OC them to save me money??

i think most decent ddr2 800 will overclock to that speed. you may have to use 5 timings, though.
depends on how expensive the 1066 is for ya.

 

[maroon1]

From what I heard, all the current AMD motherboards (AM2 socket) don't support HTT 3.0. Only AM3 motherboards are going to support HTT 3.0, right ?

In other words, people with current motherboards will not have HTT at full speed if they upgrade to phenom

please correct me if I'm wrong

 

[Wildside]

From what I heard, all the current AMD motherboards (AM2 socket) don't support HTT 3.0. Only AM3 motherboards are going to support HTT 3.0, right ?

In other words, people with current motherboards will not have HTT at full speed if they upgrade to phenom

please correct me if I'm wrong

AM2 doesnt support HTT 3.0, only 2.0. That means if the Phenom quad-cores r on a AM2 mobo, it will use only 2.0, not 3.0, meaning less performance then how much it can utilize on a AM2+ or AM3 mobo.

AM2+ will support HTT 3.0 n DDR2. AM3 will support HTT 3.0 as well, but will support DDR3 only.

*EDIT*

Barcelona supports HTT 2.0, so that is big downfall of it IMO.

 

[Apokalipse]

The current Core 2 Duo beats K8 by around 15% in gaming at the same clock speed. In other things like DivX encoding, photoshop, office applications intel beats K8 by about 25% and 30%

So it looks like Agena has a good shot at taking the gaming crown.

Penryn is going to be around 10% faster than the current Core 2 Duo in gaming. In other things it will be around 0 to 10% faster.
In programs that use SSE4 instruction like DivX6.6, Penryn is 100% faster than the current Core 2

And there are basically very few, if any programs that use SSE4.

So, in the other words if you compare Penryn with Barcelona, then Penryn would beat it in almost everything.

How about we compare Desktop to Desktop, and Server to Server chips here?

Phenom is going to have HTT 3.0, but how much difference HTT 3.0 will make ? 3% or 5% ?

Depends on the application. Sometimes it may give a 20-30% increase, in applications that are very dependent on memory speeds.

And there is the DDR2-1066 non-ECC RAM it will use.

HTT 3.0 is not going to save AMD

1. It doesn't really need to "save AMD".
2. How do you know how much performance it will give? you haven't even seen HTT 3.0 benchmarks.

Not to mention that K10 performs very bad in non gaming application like encoding, office application etc. Just look at Barcelona performance in DivX 6.6 encoding compared to K8 (only 11%). Pernyn is going to be twice as fast as the current Core 2 in DivX 6.6, if Barcelona can't beat the Core 2 in DivX6.6 encoding then how it will able to compete with Pernyn

So, you're saying that because there are a few areas that Barcelona isn't as good in, that the whole K10 architecture is uncompetitive?

The only place where AMD might beat Intel in is the synthetic benchmarks like 3dmark and PCmark.

I'm sure you want that to be true.

But who cares ? 3dmark and PCmark are noting.

They are programs that only score points and they don't tell you which processor is better

Maybe not absolutely, but they can give a good indication.

Few moths ago many rumors, and many AMD fanboys (like Apokalipse) claimed that AMD Barcelona is going to be around 40% faster than Intel processors.

Depends on the clock speeds. They do have an IPC advantage in a lot of cases.

I remember many Intel fanboys who, although have a few valid points here and there, will cherry pick and sometimes even lie to make K10 look bad.

Oh wait, that's now.

But AMD Barcelona just came out, and they are not as good as what AMD fanboys claimed. Not even close

Depends on what the claim is.
I, for example, never claimed that Barcelona was going to have a 40% better IPC. I did say it should have a better IPC (which it does in a lot of cases), but not 40%. That would be overall performance, dependent on what clock speeds they're able to reach.

If you read older threads here you will see many AMD fanboys making posts like "AMD is going to beat Intel", "K10 is 40% faster than Intel", "AMD is better because there processors are "true" quad core" blahbalbalabala etc...............

You say this as if somehow, the cherry-picking and lies by Intel fanboys never occured.

If you read those rumors and lies, you would thing that k10 is going to be the ultimate processor that noting will be able to touch it !!

Ha, now that's severely taking things out of context.

Next time I hope that AMD fanboys wait for the actual results before they open their mouths

Irony?
Penryn isn't even out yet. Neither is Agena.

EDIT: And in Anandtech conclusion, it says that the new architecture from AMD is not powerful enough to compete with Intel Penryn

We haven't even seen what Agena can do yet. Did you forget what you said just before this?

And it also says that the Phenom chip is large chip, and Penryn is a small chip. In other words, intel have the ability to reduce the price of Penryn without losing a lot of money.

Really? Phenom is bigger?
That's it, now I'm convinced K10 is the devil.

Read this
AnandTech: AMD Phenom Preview: Barcelona Desktop Benchmarks

Barcelona has a higher IPC on average.

What about DDR3 ?

Some of the current intel motherboards already support DDR3 memory

it's on its way.
Plus, K10 gets better use from DDR2 than Core 2 gets from DDR3, using the Hypertransport bus and onboard memory controller.

From what I heard, all the current AMD motherboards (AM2 socket) don't support HTT 3.0. Only AM3 motherboards are going to support HTT 3.0, right ?

AM2+ will, which is backward-compatible with regular AM2.
regular AM2 uses HTT 1.0, Barcelona uses HTT 2.0, and Agena and Kuma will use HTT 3.0.

Barcelona supports HTT 2.0, so that is big downfall of it IMO.

Yeah, I think so. HTT 3.0 is twice the speed of HTT 1.0. Or if not twice, then nearly that.

here's an article about HTT 3.0:
» HyperTransport 3.0 Released to the Public » Soft32.com News
which states that HTT 3.0 gets about 86% more bandwith than HTT 2.0

 

[Invisible]

that post,
must've taken up 15 min of your time

 

[Wildside]

i thought AM2 also supports HTT 2.0? Hmm guess not; same goes with 939.

 

[Poizen22]

i just cant wait for these new chips to be released.

 

[maroon1]

How about we compare Desktop to Desktop, and Server to Server chips here?

Agena is just Barcelona with HTT 3.0

Depends on the application. Sometimes it may give a 20-30% increase, in applications that are very dependent on memory speeds.

The 1333MHz FSB in some cases performs 5% better than 1066MHz FSB, but that just the maximum. On average the performance gain is just around 1%

I'm talking about the average performance. I highly doubt the average increase that HTT3.0 would give is more than 5%

5% is not enough to beat Penryn


And people who have AM2 motherboard will not be able to take the advantage of HTT3.0 anyway

Penryn isn't even out yet. Neither is Agena.

Anandtech said that Agena is not powerful enough to compete with Intel Penryn. Anandtech siad that, not me

Anyway, many AMD fanboys have judged the performance of Agena even before it comes out.

How many Intel fanboys claimed that penryn is going to kick AMD but ? Zero or one ?

How many AMD fanboys here in this forum claimed than Agena will wipe the floor up ? 10, 11, 20.... ?

 

[Apokalipse]

Agena is just Barcelona with HTT 3.0

and it uses socket AM2+, and doesn't have to use ECC RAM, and uses DDR2-1066 - which, given its HyperTransport bus and onboard memory controller, lets it scale better with memory speed.

Plus, we haven't actually seen gaming benchmarks.

Barcelona did have an average of about a 10% IPC advantage in the testing, in which most of the applications used are normally better on the Core architecture.

If Penryn gets a 10% advantage over current C2D's, that's only a match for Barcelona, clock for clock.

Let's say that Agena does get only 5% better. That still would put it ahead of Penryn.

The 1333MHz FSB in some cases performs 5% better than 1066MHz FSB, but that just the maximum. On average the performance gain is just around 1%

I'm talking about the average performance. I highly doubt the average increase that HTT3.0 would give is more than 5%[/quote]Firstly, FSB is different to HTT.
Secondly, I think the HyperTransport bus has more than just a frequency increase for HTT 3.0

5% is not enough to beat Penryn

See above.

And people who have AM2 motherboard will not be able to take the advantage of HTT3.0 anyway

But they won't have to buy a new motherboard.

Anandtech said that Agena is not powerful enough to compete with Intel Penryn. Anandtech siad that, not me

We'll see about that.

Anyway, many AMD fanboys have judged the performance of Agena even before it comes out.

Which is one reason why you can't just automatically say that Penryn is going to beat it.

Furthermore, many Intel fanboys have judged the performance of Penryn even before it comes out.

 

[maroon1]

Barcelona did have an average of about a 10% IPC advantage in the testing, in which most applications are normally better on the Core architecture.

Actually you are really wrong

Here is a comparison between E6750 and 6000+

http://img405.imageshack.us/img405/3961/untitledel4.jpg

Even though E6750 is clocked lower, the performance difference between E6750 and 6000+ is higher than performance difference between Barcelona and K8 (I'm talking about the non gaming application)

So, in other words even the current Core 2 can easily beat Barcelona in non-gaming application like DivX, encoding, Itunes

In gaming, they only tested oblivion and HL2 in the Barcelona vs K8 benchmark. So we only need to pick oblivion and HL2 and compare them

Barcelona on average is 18.6% faster than K8 in gaming at same clock speed

E6750 is around is 12.9% faster than 6000+ in gaming, but don't forget that E6750 is clocked lower.

So based on my calculation Barcelona is less than 5% faster than the current core 2 in gaming !! Maybe 3% which is noting

In non gaming applications Barcelona is much slower

If Penryn gets a 10% advantage over current C2D's, that's only a match for Barcelona, clock for clock.

You are assuming that Barcelona is faster than C2D, but thats incorrect. read the above.

If Penryn gets only 5% increase (not 10%) then that would put it ahead of Barcelona in almost everything

Let's say that Agena does get only 5% better. That still would put it ahead of Penryn.

Thats still not enough

Maybe it is enough to match Penryn performance in gaming

But in non gaming gaming applications, no it is not enough

.........................


As far as I know the top end phenom is going to have around 2.6-2.8GHz clocks, while Yorkfield top end is going to have more than 3.0GHz

Intel released the spec of some of the Intel Wolfdale and Yorkfield processors. The slowest one is 2.5GHz


So, in other words, low end AMD processors are going to compete with higher clocked low end intel processors

High end AMD processors are going to compete with higher clocked high end intel processors

If AMD processors performs worse than Intel per clock (I have a strong faith that Penryn is going to be better per clock), then you need to say bye bye for AMD