AMD states K8L aka Barcelona faster than all Intel cores

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
(Reply to BM)

I think when mem usage is low Kentsfield will win but when you load past physical RAM, then 4x4 will more than likely outshine it because of the FSB.

PS
I load more than 2GB RAM at times and I need something that won't choke. When I go over to Server 2003 x64, I will load it even more. i hope that the FSB won't cause problems when overloading the physical RAm but FSB is uni-directional IIRC and for 4x4 disk IO will go over the South Bridge and memory IO won't. That may turn out to be huge.

You idiot!

How many times do I have to slap you, the FSB will have no impact on the performance once you use all of the physical RAM and are swapping



He still doesnt understand that a computer swapping is only going to be as fast as the slowest component involved (erm, the hard drive)? What a knucklehead.
 
http://enthusiast.hardocp.com/article.html?art=MTE3NCwxLCxoZW50aHVzaWFzdA==

“With our next-generation architecture we expect to continue our track record of delivering industry-leading performance. When you combine this with 4X4 you change the rules of the game, and provide an enthusiast-class platform that is upgradeable, scaleable and unparalleled in terms of performance.”

But then I think that a slow FSB will cause a swapping bottleneck so what do I know.

You know nothing so quit with the FUD, moron.
 
http://enthusiast.hardocp.com/article.html?art=MTE3NCwxLCxoZW50aHVzaWFzdA==

“With our next-generation architecture we expect to continue our track record of delivering industry-leading performance. When you combine this with 4X4 you change the rules of the game, and provide an enthusiast-class platform that is upgradeable, scaleable and unparalleled in terms of performance.”

Wow really reaching there pope. AMD did say that conroe would just be a done up P3 and be slower then K8. I'll care when they have something of substance.
 
Your forgetting the Koenigsegg CCR. It had the title before the Bugatti. When they declare a a car the fastest they always test in a straigh line.
Correction to the post though, the ME412 willl be faster than the Bugatti.
 
Funny how the Intel fanboys are crying already over a stupid article.
Everyone knows K8L is gonna smoke Intels quads. Its native quad, Intels are only 2 core 2's on die.
Now they are already trying to pull the power per watt crap they so adamantly ignored when AMD had the lead in it.

Intel fanboys should rejoice, this means competition and this means lower prices. If Intel continues to hold the lead, you will see prices continue to rise on their higher end parts, eventually eclipsing the prices we have seen for EE's and FX's. ($1200+)

And you'd be basing this off nothing as per usual FUDie?
 
Dasickninja, now that you've been smacked down, is there any other unintelligent comments you'd like to make to the forum?

Learn to actually read an article, before posting comments about it and making a fool out of yourself. :roll:

Oh you mean like you've done for the majority of your posts? Then yes I'll try to be less like you.
 
I really highly doubt that K8L would be 10-15 % faster then core 2, that would mean realistically they would have to be 30-35 % faster then their current k8 core. That big of a leap is not possible with simple optimisations to an old core design. Putting in extra fpu units isnt going to boost overal performance by 30 % or more. But really does anyone think amd would come out saying our new K8L is slower then all of the new intel cores ?..... I wouldnt think so.
 
The speed increase is the word from AMD and like I said the specs seem to bear it out. As far as 4x4, I don't care which socket but they have to let it use nonECC as that is a MAJOR expense.
I wonder if Socket F's Register DDR2 memory controller is also backwards compatible with unregistered DDR2 memory. Otherwise they would have to design a separate chip which isn't very convenient. Opteron 1xxx which use unregistered DDR2 are use AM2 not Socket F so they don't really have an example of that yet.

I load more than 2GB RAM at times and I need something that won't choke. When I go over to Server 2003 x64, I will load it even more. i hope that the FSB won't cause problems when overloading the physical RAm but FSB is uni-directional IIRC and for 4x4 disk IO will go over the South Bridge and memory IO won't. That may turn out to be huge.

At least for us megataskers. :wink:
In terms of disk IO, 4x4 isn't a saviour either. In Kentsfield's case, both processors are just 1 hop from the NB. However, in 4x4s case, the second chip is 2 hops away having to go through HT to the 1st chip and then HT again to the NB. That adds latency and even worse, differential latency between the chips.

Similarly, in terms of memory bandwidth through 2 IMCs, if it happens that the 1st chip needs data that is stored in the 2nd chips memory banks it'll have to go over HT to get it. That adds latency and also starves the 2nd processor of bandwidth needed to access the NB. People call for "true" multithreading, but in the event that 4 cores are running a single program this problem is actually increased since each core will need data from each other's memory banks more frequently. In this case, 2 or 4 completely separate programs each having a dedicated chip and it's memory banks would be better at avoiding conflict.

Both 4x4 and FSB have their inherent problems. So we'll have to see. I guess a factor is also the cost of 4x4. It is definitely looking more expensive as a complete system than Kentsfield right now. If 4x4s price is too high, Cloverton may become another option in addition to the cheaper Kentsfield.

4x4 can be directly compared to Opteron 2xx. That doesn't have any bottleneck issues that have been exposed. According t the 4x4 specs each chip has it's own RAM. WIth NUMA, WIndows can put data closer to the correct processor.

They have said that 2 chips would be less than Kentsfield (not the fastest two) but the FX 7000 will probably come in around $400. I think that the reason for using an engineering sample is to disable the ECC pin of SOcket F. There are no two sockets AM2 mobos, so I guess they found it easier to remove the ECC pin than to add dual socket functionality to AM2.

They would be fools to use ECC for an enthusiast product. I expect the mobo to only have PCI\PCIe.

Again I think it will be close between the two but Barcelona will be in another league.(IMO)
 
I thought only Windows 2003 was NUMA aware at this point, am I mistaken?

The latest versions of Windows XP (SP2) and Windows Server 2003 fully support NUMA systems (Physical Address Extension must be enabled in 32bit versions (/PAE in boot.ini), which is fortunately enabled by default in AMD64 platforms, as it's required by Data Execution Prevention). What concerns applications, it first of all means that a program shouldn't deploy its data in the memory of one processor and then access it from the other processor.

http://www.digit-life.com/articles2/cpu/rmma-numa.html

People looking at the 4x4 platform, should probably give this link a read through.
 
Funny how the Intel fanboys are crying already over a stupid article.
Everyone knows K8L is gonna smoke Intels quads. Its native quad, Intels are only 2 core 2's on die.
Now they are already trying to pull the power per watt crap they so adamantly ignored when AMD had the lead in it.

Intel fanboys should rejoice, this means competition and this means lower prices. If Intel continues to hold the lead, you will see prices continue to rise on their higher end parts, eventually eclipsing the prices we have seen for EE's and FX's. ($1200+)

Mike,

Nobody knows absolutely nothing. I cant believe you. Here I was , hoping that you were growing up, yet you stand up and spew this hypocritical BS. When intel was spreading its Conroe hype, you were right here on THG under you old MMM login whining "theres no proof"," the benchmarks are cooked" "everyones being paid off by intel"---yada yada yada.

Now someone at AMDis running their pie hole, and its exactly the same situation, except you're an AMD fanboy on the other side of the fence. I dont give a crap what Phil Hester says, how much he touts the technology, Uarch or native vs shared bus BS. Until "Barcelona" is in the consumers hands, it doesnt outperform anything, it is going to out perform anything, and its nothing more than a paper tiger. There isnt even a single ES benchmark yet and here you are right back to the "AMD rulz, you'll see" BS.


The ONLY worthwhile peice of information in that article was that AMD is not going to muck with ATI and there are still going to be ATI branded video cards, which is contrary to a report from a month or 2 ago.
 
"record breaking track times" not, no record breaking times.

And shit, I actually meant to mention to mention the S7 Turbo (which has a very very minimal turnbo lag, because it was built well) as a car that nearly recreates the same straight line times, and also kills on the track.

The saleen S7 twin turbo 1000 edition does 1.12 g's on a track.

I also want to mention that the hennesey vipers post faster 0-60's and faster top speeds.

There is no "top sports car bud" each great car has it's niche.

The Enzo is still the fastest slalom car I believe

The CCR's and CCX's are neck and neck with the saleen twin turbo, and lighter smaller cars will perform better on smaller tracks.

The Bugatti really has nothing going for it, all it can do, is in a straight line, people do that to muscle cars all the time. And the usually don't look like hotwheels from mcdonalds.

You obviously have a very limited concept of what makes an exotic sports car.
 
I'm still not convinced that a 1333MHz FSB provides no benefits over a 1067MHz FSB as THG implies.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2795&p=6

Anandtech ran Conroe at 2.67GHz (266MHz x 10 and 333MHz x 8 ) and they found a 1333Mhz FSB provides an average 2.4% performance increase peaking at 7.5%. And this is only 2 cores with no need for cache coherency traffic through FSB. I would think Kentsfield should show at least an average 5% performance increase using a 1333Mhz FSB.

I'm not quite sure of THG's method's, but I commented on this at the time. If they used a i975X chipset and the 1333Mhz strap, then that is unconclusive since the 1333Mhz FSB strap has relaxed timings over the 1067Mhz setting which means that at the same clock speed the 1333Mhz FSB would be slower.

THG said they use a P965 chipset which doesn't have a 1333Mhz strap. The problem I have with the P965 is that the chipset isn't mature yet and it's performance is still generally inferior to the i975X chipset. I also wouldn't find data on that chipset conclusive.

Still it's quite possible that Kentsfield isn't that limited by the 1067MHz FSB. I'm sure the large 4MB caches are finally pulling their weight since they weren't that significant over 2MB cache on Conroe. Caching algorithms may have also been tweaked. The big reason though could be that Intel has finally figured out how to optimize Core 2 in using the 1067Mhz FSB. Core 2 obviously has more memory bandwidth than 800Mhz Netbursts, but it also has poorer FSB utilitization than 1067MHz Preslers. If Kentsfield is even just using the 1067GHz FSB to Presler levels, they'd probably already have an extra 10% bandwidth over Conroe using the same FSB.

http://www.techreport.com/reviews/2006q3/e6300-vs-sff/index.x?pg=3

Obviously those SiSoft reading may not be accurate, but it is sensible. Conroe relies on it's low latency, large L2 cache, while Presler relied on raw memory bandwidth so it's quite possible that Presler had a better optimized FSB for that purpose. It's also possible that the Oregon's design team had more experience with high FSBs than Haifa which up to now have only used 667MHz FSB optimized for power usage rather than raw performance.
 
The proverbial proof is in the proverbial pudding.

Jack

Just bookmarking.

Anyway, I found this interesting (and so did the author, it seems):

When talking to both Phil and Pat, we got some interesting answers about advancements in Instructions Per Clock with the Barcelona Pat Conway said that we will see a “performance improvement of 150% over the next two years that will be within the same power envelope.” If you take into account that you can back off frequency and add more cores, some of this can be achieved now if you are counting IPC across all cores. (See below how simply pulling back on frequency by as little as 16% can have a huge impact on power draw.) Intel is bound by the same principles as well.

On IPC and AMD’s Barcelona Phil Hester had this to say: “We want to stay focused on upgrade compatibility. We must realize more IPC per watt, and our next-generation architecture will show a 50% improvement of IPC per watt.”
(http://enthusiast.hardocp.com/article.html?art=MTE3NCwxLCxoZW50aHVzaWFzdA==)

(Just haven't got the time to go through C2D vs Barcelona IPC/Watt; it should be interesting, though, supposing Barcelona's 50% IPC/Watt improvement over the K8...).


Cheers!
 
deffinately an interesting topic.i sincerely hope that amd comes up with something substantial to kick intels ass with.

Dahak

EVGA NF4 SLI MB
X2 4400+@2.4 S-939
2 7800GT'S IN SLI MODE
2X1GIG DDR400 MEMORY IN DC MODE
WD300GIG HD
520WATT PSU
EXTREME 19IN.MONITOR
 
Why am i not surprised. Hector saying "We can´t match Intels speed" would´ve made my day. It would´ve given them a better media coverage for sure... maybe someone should mention that to the amd PR team. :lol:
 
There will be no clock speed war. It is long over ever since netburst died. Intel can only clock the netburst architecture to 3.8G, not Core architecture. Remember to clock higher, your pipeline depth should be longer (P4 pres 31stage, Core 14stage).
 
What would you expect AMD to say?

If they said "K8L will be slower than Kentsfield", do you think the typical person will go out and buy K8L?

You should also update the title to say Kentsfield or Core 2 Quadro, not Core 2 Duo. If the quad core K8L cannot beat a dual core Core 2 Duo, then that would certainly be an embarassment.

Exactly, AMD is saying exactly what people wanna hear. Kinda like planting the seed. I like Barcelona. cool name. 8)

Just like the conroe build up, maybe they can provide on what they are saying
 
One thing is certain, I have completely been able to blow the Xbitlabs memory part 1 analysis out of the water.

Your work would be for naught if it had little impact. Xbitlabs, though incorrect, has a lot of publicity. All you need now is a high-volume place to publish your review.
 
Okay after reading through this post it seems there are 2 points of view they are:
Intel fanboys - Intel pwns AMD cant beat them with K8L and
AMD fanboys - Haha intel gonna loose the lead again

All this aside it annoys me when people say "AMD can't come up with a conrow counter in the near future" well unless you work for AMD tech dept. I dont think you are qualified to say that, anyway my $0.02 worth
 
this isn't in direct reply to jumpingjack, cause he is one of the people here who posts more info than i know...

my question is why does this amd vs. intel crap always turn into a pissing contest??? is this to make up for all the times that most of you got beat up on the playground? ffs??? bringing supercars into the mix? what the hell does that prove? everyday mechanics build cars that will smoke any supercar on a dragstrip...and bloody hell, the lotus elise can pull more g's than any supercar...

what the hell is this??? the geek olympics???

ffs...so some other geek has a spectra fiber pocker protector... my programs all seem to run fine(minus AAO2.7 but we don't need to get into that)... can't you asshats live with the fact that better products may not have your favorite brand name?!?!?!?!?!?
 
Everyone knows K8L is gonna smoke Intels quads. Its native quad, Intels are only 2 core 2's on die.

Surely first Kentsfields will be 2x core 2 duo in one package but next steppings will probably be native and I bet that native ones will come out before K8L even goes to mass production.

Btw, most of the current applications save benchmarks and 3D rendering are hacked to support 2 CPUs and they will have equally poor performance on both quad-cores. Bad software -- that is something we should all worry about.

@Jack:
If you want to measure bus utilization precisely get VTune Performance Analyzer (version 8 + patches for c2d) from Intel developer site. There is 30 day trial version free for download.
 
Can anyone help me on this one? The only tangible information on the article on which this topic was based on is the alleged “50% improvement of IPC per watt”.

Now, how would that work exactly? Instructions per Cycle per watt… :? :roll:

You know what? I don’t care.

I didn’t even find anything on the article that would remotely suggest AMD or its representatives stated K8L would be faster than C2D (so much to say about the title of this post).


I really hope AMD comes up with something great. But I have to say this post can only be for the joy of fanboys (from either side).