AMD Unveils "Barcelona" Architecture

Status
Not open for further replies.
If intel takes 50 second to compress a file with winrar, then AMD would take around 40-45 seconds.

I don't think anyone would get a heart attack if he waits for 5 or 10 extra seconds.

Depends on the size of the file. Sometimes I need to compress large files with winrar, some GBs. Taking in account the numbers you used this equals a reduction of 12%/25% of the compress time, this is great. I would say that taking 8m instead of 10m is excellent.

Some AMD fan boys speculated about the K10 performance, but you are also speculating about the Penryn performance...
 
Depends on the size of the file. Sometimes I need to compress large files with winrar, some GBs. Taking in account the numbers you used this equals a reduction of 12%/25% of the compress time, this is great. I would say that taking 8m instead of 10m is excellent.

Some AMD fan boys speculated about the K10 performance, but you are also speculating about the Penryn performance...

my point exactly... we haven't seen benchmarks for Phenom or YorkField.... enough of the freaking fanboy war!!!
 
No

Most of the Intel motherboards nowadays support 1333 FSB, like P35 motherboards

The motherboards from Nvidia like nforce 680 have 1333MHz FSB support by bios update. These motherboards were out long time ago
But the point is, people who don't have a 1333FSB motherboard, will still have to get one. And there are people who don't.

So it is hypocritical to criticise K10 because some people don't have an AM2+ motherboard.

Those who don't have an AM2+ motherboard, can still get one. but they don't have to.

The clock speed of the CPU's isn't going to decrease going from HTT 3.0 to 1.0, so the effect isn't as great as an Intel 1333FSB CPU in a 1066FSB motherboard.




Intel X38 are going to support PCI-E 2.0, and it will come out this month

Does AMD have any motherboard that support PCI-E 2.0 ? I think the answer is no
actually, the answer is yes. To name one, there's the Gigabyte GA-MA790-DQ6

But I think that gap is going to be less when 45nm Xeon comes out
So you're actually admitting that Barcelona is faster than 45nm Xeon in this?
that must be a first.



wow, two games, in a very unspecific benchmark that contradicts other benchmarks.

You said that Barcelona is better for gaming
based on trends from K7 and K8 architectures.
K10 will be better in gaming. I just don't know how much yet.

Then you said we haven't actually seen gaming benchmarks for Barcelona
To point out that your comparision between Barcelona and Penryn in gaming is invalid.

Your are contradicting yourself
see above.

Then I think you don't know anything about penryn
That's just a childish attack.

So, not all that much more than what I said.
Except for the 12MB cache, I missed that. That can make a significant difference in a few applications, such as Superpi.
 
Except for the 12MB cache, I missed that. That can make a significant difference in a few applications, such as Superpi.

wouldnt a 12mb cache perform alot better? having a bigger cache means that itll store more and more instructions for the cpu and being able to access them at the internal clock rate of the cpu. which means getting less instructions from the ram and accessing them at the slower fsb speed.
 
based on trends from K7 and K8 architectures.
K10 will be better in gaming. I just don't know how much yet.

But how can you know that it is better than Intel in gaming if you haven't saw any gaming benchmark as you claim ?

actually, the answer is yes. To name one, there's the Gigabyte GA-MA790-DQ6

I have searched in newegg and google, I haven't found any place that sells them


So you're actually admitting that Barcelona is faster than 45nm Xeon in this?
that must be a first.

So ?

Do you admit that the current Xeon can beat Barcelona in Windows media Encoding at the same clock speed. If the current Xeon beats it in that thing then what will happen when 45nm comes out ?

wme9.gif



Just because Barcelona wins in one thing, it doesn't mean it can win in all things

So, not all that much more than what I said.
Except for the 12MB cache, I missed that. That can make a significant difference in a few applications, such as Superpi.

What about the improved Wide Dynamic Execution and Virtualization technology ?

What about the reduced cache latency ?

What about 24-way associative cache ?


The clock speed of the CPU's isn't going to decrease going from HTT 3.0 to 1.0, so the effect isn't as great as an Intel 1333FSB CPU in a 1066FSB motherboard.

Can you use AM2 processors on Socket 939 ?

As far as I know, all the 1066FSB motherboards don't support 45nm processors anyway. So, we won't see any people having penryn with decreased clock speed. but we are going to see people having phenom with lower HTT speed

By the way, Intel planing to release X48 motherboards on early 2008 and these motherboards have 1600MHz FSB support.

X38 motherboards are supposed to come out this month
 
Can you use AM2 processors on Socket 939 ?

Can you use a 775 on a 468? Same comparison.

Let's chill with the whizzing contest and wait to see what happens.
 
wouldnt a 12mb cache perform alot better? having a bigger cache means that itll store more and more instructions for the cpu and being able to access them at the internal clock rate of the cpu. which means getting less instructions from the ram and accessing them at the slower fsb speed.
Firstly, the cache still has to get its data from somewhere. (FSB)
Secondly, not all that many applications benefit more than maybe a percent or two from having more cache.

But how can you know that it is better than Intel in gaming if you haven't saw any gaming benchmark as you claim ?
Because of the architecture. K10 is just an upgrade of K8, which is an upgrade of K7. They share much of the same base architecture.
The pipeline, and instruction decoders on AMD's K7 and up work better with gaming.

Similarly, the Core 2 shares a lot of its architecture with the Pentium 3.

I have searched in newegg and google, I haven't found any place that sells them
I haven't seen any place that sells Intel boards with PCI-E 2.0.

So ?

Do you admit that the current Xeon can beat Barcelona in Windows media Encoding at the same clock speed. If the current Xeon beats it in that thing then what will happen when 45nm comes out ?
It'll be a little bit faster, of course.

But I don't have a problem admitting that Core 2's are better in some areas.
K8 is better in some areas than others. In some benchmarks the difference to Core 2's is less (specifically, GAMES, Cinebench, 3dmark, and some memory-dependent applications), and sometimes K8 even comes out on top.
K10 is just an upgrade of K8, and so should have similar strengths and weaknesses.

But you seem to simply ignore any and all advantages AMD has.

Barcelona still has a better average IPC, in applications where the Core architecture is usually better.

wme9.gif



Just because Barcelona wins in one thing, it doesn't mean it can win in all things
Yes, but you seem to just ignore any and all advantage AMD has.

What about the improved Wide Dynamic Execution and Virtualization technology ?
Do you even know what that does?
Furthermore, does it even state what these improvements actually are, and how they specifically affect performance?

Also, Virtualisation is only helpful if you're using multiple OS's, or multiple computers on a single operating system. K10 has a vastly improved virtualisation over the previous generation.

What about the reduced cache latency ?
it's a minor difference.

What about 24-way associative cache?
Will only make a difference in cache-dependent applications. Which is not many, except for programs like superpi.

What about the almost double bandwith of HTT 3.0 over HTT 2.0?

and the DDR2-1066 RAM Agena will use, instead of the DDR2-667 used on Barcelona test rigs?

and the lack of ECC memory used

Memory speed plays a much bigger role in performance on K8 and K10 than on Core 2's, due to the much faster Hypertransport bus.
the high amounts of cache is probably just a way to compensate for the much slower FSB. And it doesn't always work.

Can you use AM2 processors on Socket 939 ?
Can you use socket 775 processors on socket 478?

As far as I know, all the 1066FSB motherboards don't support 45nm processors anyway. So, we won't see any people having penryn with decreased clock speed. but we are going to see people having phenom with lower HTT speed
well that's something in AMD's advantage.

At least people on AM2 don't have to buy a new motherboard.
But they still can buy a new motherboard. People are not forced to stay on regular AM2.

By the way, Intel planing to release X48 motherboards on early 2008 and these motherboards have 1600MHz FSB support.
verses a 3600MHZ, or possibly 4000MHZ HyperTransport bus, and an onboard memory controller.

X38 motherboards are supposed to come out this month
That's nice.
 
okay lemme join in on the quote war lol...

then you need to say bye bye for AMD

this just makes me laugh. ha hah ha.

Just because Barcelona wins in one thing, it doesn't mean it can win in all things

omfg dude i have used that same argument with you countless times to defend AMD. lol.

What about the almost double bandwith of HTT 3.0 over HTT 2.0?

and the DDR2-1066 RAM Agena will use, instead of the DDR2-667 used on Barcelona test rigs?

oh. snap....

wouldnt a 12mb cache perform alot better? having a bigger cache means that itll store more and more instructions for the cpu and being able to access them at the internal clock rate of the cpu. which means getting less instructions from the ram and accessing them at the slower fsb speed.

the reason for the bigger cache (from what i understand, but heh i dont know everything) on intels is BECAUSE they use a FSB.
cpu access from the cache is faster than cpu access from the fsb accessing the memory. lol a lot of access there, sonny. but, it still isnt as fast (in principle. yes i know c2ds are faster than x2s) as an on-die memory controller. by far.
thats my
2cents.jpg
 
okay lemme join in on the quote war lol...



this just makes me laugh. ha hah ha.



omfg dude i have used that same argument with you countless times to defend AMD. lol.



the reason for the bigger cache (from what i understand, but heh i dont know everything) on intels is BECAUSE they use a FSB.
cpu access from the cache is faster than cpu access from the fsb accessing the memory. lol a lot of access there, sonny. but, it still isnt as fast (in principle. yes i know c2ds are faster than x2s) as an on-die memory controller. by far.
thats my
2cents.jpg

nice argument.. and loving the "2 cents" lol!!
 
Status
Not open for further replies.
Back
Top Bottom