Official AMD Barcelona Thread

[savantu]

he-he
AMD claims Barcelona is 50 per cent faster than Clovertown in server apps
So 2GHz of K10 will be enough for all =)

sure Barcelona at 2 GHz will be very competitive

What a fine example of blind faith...

You should remember that even in AMDs simulated scores , they were comparing a 2.6GHz K10 with a 2.66GHz Clovertown.
Since then , K10 went to 2GHz and Clovertown to 3GHz.

Mark my words : K10 won't even get near to the 2.66GHz Clovertown , not to mention the 3GHz 5365.

[terrace215]

can you please show us the secret info you got about K10's IPC?
i ask because the last official infos say K10 has 20-50-70% IPC boost.

No "infos say K10 has 20-50-70% IPC boost".

Now all AMD will say is that FOUR K10 cores @ 2GHz will perform 40-70% faster than TWO top-bin K8 cores, in certain cherry-picked benchmarks.

That's not a statement about K10 core IPC vs K8 core IPC.

That figure is likely 10-15% in most cases, putting Barcelona IPC roughly on par with Core 2 Duo IPC.

In applications that max out main memory bandwidth, the K10 will do relatively better. In apps with a large cache footprint, C2D will do relatively better.

But enough is known about the K10 design, plus AMD's simulated performance on a few benchmarks, to pin down its IPC.

I can understand any confusion if you only looked at AMD press releases, since, when they issue new claims, they fail to mention that the old claims no longer apply. And they've changed their story at least 3 times at this point:

First: Barcelona will outperform Intel's fastest Clovertown by X & Y%.

(AMD begins to have clockspeed concerns, so in April: )

Second: Barcelona will outperform Intel's Clovertown clock-for-clock by X and Y% on two 2P spec_rate benchmarks.

(Intel improves its compiler and tuning on spec_rate, and Barcelona clock is still dreadfully low, so change to: )

Third: Barcelona will outperform AMD's own DC K8, per socket (not core) by X and Y% on two types of applications.

and also:

Third(b): Barcelona will have the best performance PER WATT.

(But now Intel will launch 50W 2GHz Clovertown w/1333FSB in August, so AMD will be abandoning 3b shortly. )

[nn_step]

God people, if you don't believe that K10 will not outperform K8 on the core level. Then I will be forced to wonder how the do you remember to keep breathing.
The Short way to see K10:
It performs better than K8 by a significant margin in many levels.

[red]

It performs better than K8 by a significant margin in many levels.

Not according to my calculations! http://www.xtremesystems.org/forums/...&postcount=602

[savantu]

God people, if you don't believe that K10 will not outperform K8 on the core level. Then I will be forced to wonder how the do you remember to keep breathing.
The Short way to see K10:
It performs better than K8 by a significant margin in many levels.

You're in for a big surprise in a few tests....

[alayashu]

terrace, they said something that can be interpreted in many extreme ways.
but your interpretation, that K10 is as fast or slower than K8 is very improbable.
IMO their 50-70% was at the core level / per clock in multithreaded environment
on FP/SSE. Though, taking the same discussion from head to tail every month is
useless, there are 2 more months until the server chips are out, and 1 month till
the long awaited benches pop.

You're in for a big surprise in a few tests....

allrite then, show us those surprise benches. I hope you weren't referring to rendering eh

[Shintai]

he-he
AMD claims Barcelona is 50 per cent faster than Clovertown in server apps
So 2GHz of K10 will be enough for all =)

sure Barcelona at 2 GHz will be very competitive

Where do you see server apps. All they showed is a specfp_rate to be 50% faster. And they even in the press release stays at estimated performance and only "certain applications".

[BrowncoatGR]

"Now it is: 4 Barc cores will outperform 2 K8 cores by up to 40 to 70%. Big deal."

Those are handpicked benches. One is the outdated SPECfp_rate2000
http://i17.tinypic.com/2eq51k5.jpg and the other "OLTP" (usually they should specify if this is TPC-C or H) http://i16.tinypic.com/4d6wjmo.jpg

http://online.wsj.com/article/SB1183...googlenews_wsj
http://news.moneycentral.msn.com/pro...629&ID=7108474
http://www.channelregister.co.uk/200...arcelona_2ghz/
"Mr. Allen estimated that the first versions of Barcelona still will be 40% to 50% faster than existing Opteron chips, which have two processors."
"We will be seeing a performance boost of 40-50 per cent above our highest frequency dual-core products that are available today," AMD VP Randy Allen told us.

Some calculating, the fastest 2P K8 can deliver 4*3GHz=12GHz. 45% (40+50/2) faster than that would be 17.4Ghz. Divided by 8 K10 cores=2.175GHz. 2.175/2=8.75% IPC K10 advantage?

You cant just multiply frequency for every core as that doesnt take into account scaling, intercore communication etc. You're results are completely invalid.

[BrowncoatGR]

You need to note a few facts we have now:
1. R&D budgets are declining.

You are wrong R&D expenses are constantly increasing. Particularly node shrinks.

2. People will not upgrade unless there is something better.

That is not true. Most ppl treat their computers like every other electrical appliance in their houe. It breaks or a new gadget comes out and they'll gat a new computer

3. Vista is NOT late due to lack of competition. Do some research. Vista was basicly abandoned and started from scratch again.

Actually no it wasnt restarted from scratch, they just dropped all the innovative features in order to not delay the release even further as it was quite clear they couldnt be sure they could finalize the new features in a timely fashion.

4. x86 compability is the ABSOLUTE reason we still got x86 CPUs. There is no technical reason for us to stay with it besides that. And nobody dares to change it because the competitor would run of with it all. IA64 was a bold move, now its isolated to bigtin because AMD just added x64 and gave us 20 more years with x86. Thats not development, thats stagnation.

Compatibility is not the only reason IA64 failed to become mainstream. Performance just wasn.t there. It took Intel ages to get the compiler to work as they wanted. You have to understand that to push for a new arch to get accepted Intel needed to overcome the resistance the lack of compatibility would create by offering substantial performance improvements. That never happened , so IA64 failed. AMD just saw the opportunity and grabbed it.

And remember, I didnt say a monopoly would be beneficial forever.

And if I can add the broadband part again. Monopolies gave us fiber, not competition.

There can be no temporary monopoly in the x86 market. If AMD goes out of business, Intel will be all there will be and im quite sure they'd start pushing IA64 again. The barrier for entry is just too high for any company to enter the market.

[Jimmer411]

Seriously the only people that I hear freaking out over the performance differences between C2D and A64 are enthusiasts. Our share of the market is tiny.

Lets say K10 matches C2D clock for clock, and only overclocks to 3.2ghz average. But at the same time draws much less power and runs cooler. Is that not so much of a bad thing? This is like Ford releasing next years Mustang with the same exact performance, but with 5-10MPG more fuel economy.

Seeing how little of a difference CPU makes at higher resolutions in gaming, that unless you do something like encoding/decoding and CPU intensive applications, that there is no difference. Joe six pack is going to see the 5600+ and E6600 performing roughly the same at stock, and for Joe sixpack overclocking is taboo, he dont want to fry his expensive computer.

For alot of people out there, upgrading to a quadcore also involves upgrading their PSU. And alot of us really only want to upgrade our computers, not rebuild a whole new one. Which is why alot of people didnt switch from 939 or 754 to AM2.

[BrowncoatGR]

No "infos say K10 has 20-50-70% IPC boost".

Now all AMD will say is that FOUR K10 cores @ 2GHz will perform 40-70% faster than TWO top-bin K8 cores, in certain cherry-picked benchmarks.

That's not a statement about K10 core IPC vs K8 core IPC.

That figure is likely 10-15% in most cases, putting Barcelona IPC roughly on par with Core 2 Duo IPC.

In applications that max out main memory bandwidth, the K10 will do relatively better. In apps with a large cache footprint, C2D will do relatively better.

But enough is known about the K10 design, plus AMD's simulated performance on a few benchmarks, to pin down its IPC.

I can understand any confusion if you only looked at AMD press releases, since, when they issue new claims, they fail to mention that the old claims no longer apply. And they've changed their story at least 3 times at this point:

First: Barcelona will outperform Intel's fastest Clovertown by X & Y%.

(AMD begins to have clockspeed concerns, so in April: )

Second: Barcelona will outperform Intel's Clovertown clock-for-clock by X and Y% on two 2P spec_rate benchmarks.

(Intel improves its compiler and tuning on spec_rate, and Barcelona clock is still dreadfully low, so change to: )

Third: Barcelona will outperform AMD's own DC K8, per socket (not core) by X and Y% on two types of applications.

and also:

Third(b): Barcelona will have the best performance PER WATT.

(But now Intel will launch 50W 2GHz Clovertown w/1333FSB in August, so AMD will be abandoning 3b shortly. )

You are speculating and presenting it as Gospel. None of us knows the IPC of K10. It is impossible to get a clear picture from AMD's statements. You want to speculate thats fine, just dont present your speculations as fact. As for compiler tuning,why would you run a K10 benchmark on code produced by the Intel compiler? If you meant that you would use Intels compiler for Clovertown and something else for K10 thats a really bad benchmarking practice as the scores are not directly comparable.

[Shintai]

You are wrong R&D expenses are constantly increasing. Particularly node shrinks.

Oh really? Can you tell me why AMD reduced its R&D with 500million and Intel have reduced it with 300million, and might reduce it further?

Dont mix factory cost with R&D.

That is not true. Most ppl treat their computers like every other electrical appliance in their houe. It breaks or a new gadget comes out and they'll gat a new computer

There is no reason a computer cant last 5years or more. Today it usually last 2-3 years due to people wanting the newer and faster.

Actually no it wasnt restarted from scratch, they just dropped all the innovative features in order to not delay the release even further as it was quite clear they couldnt be sure they could finalize the new features in a timely fashion.

No, in the middle of the Vista development security got alot more focus. Alot of people was moved to the XP Sp2 project aswell. And they had to restart most part of the Vista project with security in mind after reeducation of the staff.

Compatibility is not the only reason IA64 failed to become mainstream. Performance just wasn.t there. It took Intel ages to get the compiler to work as they wanted. You have to understand that to push for a new arch to get accepted Intel needed to overcome the resistance the lack of compatibility would create by offering substantial performance improvements. That never happened , so IA64 failed. AMD just saw the opportunity and grabbed it.

So you admit that competition is the issue. You dont develop it overnight. IA64 performance in itself is good. Compability performance sucked. AMD did as I described. Added a cheap touch and we sit in the same garbage for the next 20 years. If over 10 billion$ cant change architecture in a competitive environment. Then what is needed?

There can be no temporary monopoly in the x86 market. If AMD goes out of business, Intel will be all there will be and im quite sure they'd start pushing IA64 again. The barrier for entry is just too high for any company to enter the market.

It actually worked fine for transmeta. Dont tell me nobody will enter the market. If the profit is good enough there to be made, then you will get competition. Its no more different on how Via, Cyrix, Transmeta and Nexgen started.

The only barrier there is, is the direct foundry advantage a company can have.

[Zytek_Fan]

Oh really? Can you tell me why AMD reduced its R&D with 500million and Intel have reduced it with 300million, and might reduce it further?

You do realise that Intel at any moment can substantially increase R&D spending?

[BrowncoatGR]

This is getting totally offtopic.

Oh really? Can you tell me why AMD reduced its R&D with 500million and Intel have reduced it with 300million, and might reduce it further?

Dont mix factory cost with R&D.

AMD intends to cut 500mil by slowing down the upgrade of FAB30 -> FAB38 not cuttint R&D Dunno about Intel

There is no reason a computer cant last 5years or more. Today it usually last 2-3 years due to people wanting the newer and faster.

It is possible to skew the perception ofthe public you know. In fact Intel were and still are experts at it.

So you admit that competition is the issue. You dont develop it overnight. IA64 performance in itself is good. Compability performance sucked. AMD did as I described. Added a cheap touch and we sit in the same garbage for the next 20 years. If over 10 billion$ cant change architecture in a competitive environment. Then what is needed?

NO i dont admit anything like that. Dont put words in my mouth. Any new technology has to offer significant improvements in order to overcome the early adoption barrier all new technologies have to face. If it does not deliver it does not get adopted. IA64 never delivered on Intels promises and expectations. AMD on the other hand made significant improvements to the x86 arch. The IMC, HT, C'N'Q, AMD64 and soon split power planes, independent core throttling, on the fly gpu switching(for mobile platforms), CTM, heterogenous computing, high IPC focus,native dualcore/quadcore etc. Basically AMD made the dying x86 competitive again. BTW same crap? Where does that leave C2D...

It actually worked fine for transmeta. Dont tell me nobody will enter the market. If the profit is good enough there to be made, then you will get competition. Its no more different on how Via, Cyrix, Transmeta and Nexgen started.

The only barrier there is, is the direct foundry advantage a company can have.

Notice a trend related to all those companies you mentioned? They are all out of business(in cpu manufacturing) or occupy a fringe segment of the market! It is very difficult to compete with a huge company like Intel. You need to spend huge amounts of money in R&D just to get started, then spend quite a lot of money on manufacturing, all that in a highly competitive market. No investor who has a bit of common sense would spend money on a x86 startup.

Now to get on topic. Barcelona is a server part. A market where absolute performance means very little. Improved powersaving and low TDP are going to matter a lot for datacenters and multisocket sytems. That is Barcelonas target market. If it can deliver on the IPC front it is a winner at least until Intel releases 45nm server cpus in Q4/Q1 08. Desktop parts are going to get clocked higher than 2.0ghz and there is no telling right now how good or bad a 2.6ghz k10 compares to Kentsfield.

[nn_step]

Oh really? Can you tell me why AMD reduced its R&D with 500million and Intel have reduced it with 300million, and might reduce it further?

That sounds contrary from what I've heard. Yes I agree they have trimmed down their lesser performing staff but they did significantly increase their R&D staff and Budget in the past year.

[GoThr3k]

BrowncoatGR and nn_step just made my day

[Zytek_Fan]

That sounds contrary from what I've heard. Yes I agree they have trimmed down their lesser performing staff but they did significantly increase their R&D staff and Budget in the past year.

They cut a few thousand workers from Hillsboro, then they will hire any graphics engineer that walks in their doors

BrowncoatGR and nn_step just made my day

I agree.

[terrace215]

You are speculating and presenting it as Gospel. None of us knows the IPC of K10. It is impossible to get a clear picture from AMD's statements. You want to speculate thats fine, just dont present your speculations as fact.

Do you have reading comprehension difficulties? I explained how I came to my conclusions. Of course it is "speculation". Do you really want every assertion every poster makes to be preceded with "IMO"??

And just because you haven't yet been spoon-fed 10 different reviews of Barcelona doesn't mean that the rest of us cannot figure out a fairly small window for K10 IPC based on what AMD has revealed in terms of its design and selected benchmarks. It's really not that hard.

As for compiler tuning,why would you run a K10 benchmark on code produced by the Intel compiler? If you meant that you would use Intels compiler for Clovertown and something else for K10 thats a really bad benchmarking practice as the scores are not directly comparable.

A K10 benchmark? It's the SPEC 2006 CPU integer throughput benchmark! Your last sentence is nonsense. SPEC is designed to compare the performance of different CPUs, not all of them even x86. Of course submissions use different compilers! Sheesh.

[californian7856]

i wonder why AMD didnt setup 2 pcs with kentsfield QX6700 @ 3.7GHz vs K10 @ 2+ to 3GHz like intel did 4moths b4 conroe launch??

[quake6]

I have one question.This http://www.techpowerup.com/?33924
That is for the server,right?Can it be used in am2/am2+ socket or it is for 1207 or what so ever socket?

[Shintai]

I have one question.This http://www.techpowerup.com/?33924
That is for the server,right?Can it be used in am2/am2+ socket or it is for 1207 or what so ever socket?

Only server. The single socket versions wont ship there.

[Shintai]

That sounds contrary from what I've heard. Yes I agree they have trimmed down their lesser performing staff but they did significantly increase their R&D staff and Budget in the past year.

You missed the point completely. It was a monopol vs competition issue. About competition aint equal to higher R&D.

In 2007 they both reduced heavily on R&D.

On a funny sidenote, MS R&D budget is higher than Intel+AMD.

[madcho]

Full bench of Barcelona from xbit labs wanted here :p

[quake6]

Only server. The single socket versions wont ship there.

When it will be for desktop?

[dexman]

On a funny sidenote, MS R&D budget is higher than Intel+AMD.

Ever wondered why is that? Because of higher competition. MS is not into OS only, they do all kinds of things. I´m not saying Intel is just about processors but they are pretty much about hardware. And setting up a software and hardware(I know, MS does hardware too but main revenue comes from software) bussiness is heck of a difference.

You don´t see any open source supporters manufacturing processors in their garages, do you? Nor you see anyone freely distributing hardware. Hence the competition in software market is higher. You gotta keep pushing if you price your product as MS do while there are free OS that do pretty much the same, except not having support from game developers.

Basically you are comparing apples with oranges.