First *real-world* benches of Barcelona?

[gOJDO]

@Gary Key

When you are talking about K10 performance and comparing it to Clovertown, please be more specific about the kind of software. How does Bareclona's AL performance competes against Clovertown?

Thank you

 

[Amaroque]

He is under NDA still, with regards to both processors...

Thanks for sharing what you can Gary. :beer:

 

[Idontcare]

Originally posted by: Amaroque
He is under NDA still, with regards to both processors...

Thanks for sharing what you can Gary. :beer:

Seconded. Many thanks Gary for poking your head in here and giving us some hints. Hopefully the crowd doesn't skewer you this time in exchange for your niceness.

 

[HopJokey]

Originally posted by: gOJDO
@Gary Key

When you are talking about K10 performance and comparing it to Clovertown, please be more specific about the kind of software. How does Bareclona's AL performance competes against Clovertown?

Thank you

Patience my friend, just four more days until NDA is lifted. Gary has been quite generous with the info thus far considering he is under NDA.

 

[lopri]

Originally posted by: Amaroque
He is under NDA still, with regards to both processors...

Thanks for sharing what you can Gary. :beer:

3rd'ed. I am so glad that we have him here.

 

[redpriest_]

Michael Dell's quote got changed to:

""If you look at floating point instructions, Barcelona is about 30 percent faster than Clovertown. However, if you look at integer instructions, Clovertown is about 30 percent faster than Barcelona," Dell said."

 

[Idontcare]

Originally posted by: redpriest_
Michael Dell's quote got changed to:

""If you look at floating point instructions, Barcelona is about 30 percent faster than Clovertown. However, if you look at integer instructions, Clovertown is about 30 percent faster than Barcelona," Dell said."

I thought AMD made a big deal out of their simulations showing a hypothetical 2.6GHz K10 spanking 2.93GHz Clovertown by 40% in Spec_int?

Did I miss something? Is Dell referring to 2.0 GHz H10 versus 3.0 GHz Clovertown?

 

[Accord99]

That was a 2 socket, 2GHz Barcelona vs a 2.33GHz Clovertown in SPECfp_rate2006.

 

[gOJDO]

Originally posted by: IdontcareIs Dell referring to 2.0 GHz H10 versus 3.0 GHz Clovertown?

You must be kidding.
A 2GHz Barcy being 30% faster then 3GHz Clover would mean Barcy does +95% more IPC then Clover. That is impossible.

 

[bryanW1995]

not impossible, but very unlikely.

Gary has the goods in his hands and is probably busy running benchmarks as we speak. Gary, you have one of the best jobs in the world. Thanks!

It's good to see that barc won't be a complete flop, now it's just up to amd to get their clock speeds up to competitive levels in a hurry.

 

[gOJDO]

@bryanW1995

Please point at such case where K10 can actually do +95% more IPC then Clovertown?

 

[Nickel020]

Well if a C2D can easily do 95% more than a Pentium D, why shouldn't a new CPU be able to be this much better than the C2D? Of course we know that Barcelona is not that fast, but it's certainly not impossible.

 

[Reynod]

http://forum.coolaler.com/show...d.php?t=161127&page=34

http://img296.imageshack.us/my...mage=barcelona1ph6.jpg

heartfelt's post at coolaler

 

[Nickel020]

Links don't work...

 

[gOJDO]

Originally posted by: Nickel020
Well if a C2D can easily do 95% more than a Pentium D, why shouldn't a new CPU be able to be this much better than the C2D? Of course we know that Barcelona is not that fast, but it's certainly not impossible.

According to your comment I assume you do not understand Pentium D, Core2, K8 and K10 architectures. If this is the case it would be very difficult for me to explain to you why Core2 is so much more efficient then Pentium D and why K10 has no theoretical potential to outperform Core2 by that %, even in the best scenario when using a FP and memory bandwidth hungry software optimized specially for K10.

 

[Nickel020]

You must have some insider information and some expert knowledge of how a CPU works, or you would not make the claim that it's impossible for K10 to be 95% faster than a C2D. Sure it's possible, for all we know there may be optimizations that make it a lot faster in some very specific applications. And we haven't seen any tests whatsoever of the final silicon. There can be a huge difference between engineering samples and final silicon where lots and lots of the errors are fixed.
And please share some of you knowledge of microprocessor architecture with me, so I can begin to understand why it's not possible :S

 

[Arkaign]

Originally posted by: Nickel020
You must have some insider information and some expert knowledge of how a CPU works, or you would not make the claim that it's impossible for K10 to be 95% faster than a C2D. Sure it's possible, for all we know there may be optimizations that make it a lot faster in some very specific applications. And we haven't seen any tests whatsoever of the final silicon. There can be a huge difference between engineering samples and final silicon where lots and lots of the errors are fixed.
And please share some of you knowledge of microprocessor architecture with me, so I can begin to understand why it's not possible :S

# of FPU/ALU units, pipeline design, cache size, bus architecture, all point to a 30% clock for clock advantage AT BEST over C2D design. Google it, I don't feel like typing a bunch of crap here that's widely available on the net.

But yes, 95% faster than C2D is impossible for the K10 design.

 

[Arkaign]

If you're too lazy to look :

http://en.wikipedia.org/wiki/A..._the_microarchitecture

http://www.tomshardware.com/fo...hitecture-architecture

 

[nj2112]

Originally posted by: Reynod
http://forum.coolaler.com/show...d.php?t=161127&page=34

http://img296.imageshack.us/my...mage=barcelona1ph6.jpg

heartfelt's post at coolaler

This fake seems to be being posted all over the place.

The original source is here.

http://www.xtremesystems.org/f...=2411989&postcount=504

Quote from original post:-
"Clearly the above is fake, just doing it to show you how wary you should be of benchmarks these next couple of days."

 

[bryanW1995]

I wasn't trying to make any kind of statement, I just meant that 95 % is POSSIBLE. Arkaign makes a very convincing argument, and after reading that info I am even more convinced that it won't happen, but we could be wrong.

 

[Jeff007245]

Originally posted by: Gary Key

Originally posted by: SickBeast
I wonder why the chip would scale so well beyond 2.4ghz...it must have something to do with the memory controller running at a higher frequency.

So perhaps Gary Key is right and these benchmarks are from crippled Barcelonas.

Also, the first RD790 boards we have will undergo another spin so any Phenom results with those boards are subject to interpretation depending on whether you like AMD or not.

From that I take it that the chip did not perform to expectations, however it's very possible that it's due to a problem with the platform. Why not just run the chip on a current AM2/AM2+ board?

1. Around 2.4GHz and higher you will want to run CAS4 1066 and at 3GHz+, we expect/estimate that 1333 CAS5 will come in handy. AMD is working very closely with the memory suppliers at this time to get low latency DDR2-1066 ready quickly and to start looking at DDR2-1333 next year before they worry about the switch to DDR3. Memory latencies are going to be a key with this CPU and the performance oriented consumer chipsets.

2. The current AM2+ boards are still immature from a driver/chipset viewpoint, at least to the point of not providing benchmarks yet, once they get closer, expect some numbers.

3. The lower speed Barcelonas on the server chipsets are not going to shine that well in a lot of consumer applications (against higher clocked Yorkfields, but that is not the target market right now), so AMD desperately needs to get the speeds up for this chip to show its true potential. Right now, its doing a lot better than what we saw at Computex and we understand the latest silicon is a marked improvement (several of the board guys were extremely pleased with the last samples) over the last spin we tested. The numbers will be out in a couple of weeks, some will be very happy, some will not, but at least the damn thing will finally be shipping.

p.s. Not trying to be vague, just until the final CPU samples are in and the green light is given by the board guys, no real point in guesstimating.

Sorry for the super late reply towards what was said here. I wanted to ask this question a while back, but when I registered on these forums, the activation email was sent to my junk mail and only found it now.

First of all, I bought a 2 sets of Mushkin's 996535 DDR2 XP2-8500 5-5-4-12 2x1GB memory. I plan on building an AMD Phenom system come december and went ahead and bought memory(2 months ago) because supposedly DDR2 prices was going to increase, and sure enough they did. I bought 2 sets of this memory because the price was great and was rated at 1066mhz, which is supposedly the 1:1 speed of upcomming AMD Procs. Now... The question is was it a good move to buy this memory already? I ask because I am now confused after what Gary Key says in the quote above.

I already knew before buying this ram that Phenom will be compatible with 1066 memory. Going by what Gary Key said above, Case 4 at 1066 would be ideal whenever Phenom is launched. And Phenom at 3.0ghz+ would require DDR2-1333mhz at cas 5. I'm taking what is said with a grain of salt, but again he does most likely have early revisions of the chips, and knows more than I do.

Now I am stuck because I don't know whether or not I made the right decision on purchasing this ram already. I thought DDR2 was already done, and that its full potential has already been reached with the Micron D9 chips. Also, since DDR3 is emerging, I thought memory manufacturers would concentrate on that instead, because full line-ups of DDR2 are already out there. SO are there really going to be faster DDR2 chips produced? DDR2-1066 cas 4? DDR2-1333 cas5? Just really want to know, because I will be quite unpleasant to finding out the memory I bought will be obsolete once Phenom actually launches.

Thank You,
Joeffrey Orquia

 

[gOJDO]

Originally posted by: Nickel020
You must have some insider information and some expert knowledge of how a CPU works, or you would not make the claim that it's impossible for K10 to be 95% faster than a C2D. Sure it's possible, for all we know there may be optimizations that make it a lot faster in some very specific applications. And we haven't seen any tests whatsoever of the final silicon. There can be a huge difference between engineering samples and final silicon where lots and lots of the errors are fixed.
And please share some of you knowledge of microprocessor architecture with me, so I can begin to understand why it's not possible :S

You should try understanding K8 first. This is a very good article for that purpose:
http://www.chip-architect.com/...f_AMDs_64bit_Core.html

Then, you'll have to understand the K10 vs K8 improvements. You can find a lot of data about it here:
http://www.xcpus.com/forums/amd/52-barcelona-info.html

BTW, ask about specific improved part of the K10 vs the K8, so we can talk about the possible performance improvements it can bring.

And, also some basic math: Clock for clock Core2 is 20% faster then K8 in average. According to your "+95% IPC over Core2" statement K10 should be 2.34 times faster than K8 or K10 to have +134% more IPC than K8.

 

[Nemesis 1]

Originally posted by: gOJDO
@bryanW1995

Please point at such case where K10 can actually do +95% more IPC then Clovertown?

First off Thanks Gary for the heads up.

goJoDO. Only 4 days left to see the performance differances in these cpu's.

For me Penryn is the easy winner.

1). WAY more future proof. (SSE4) With performance improvements with every programm that uses it. Since were big on audio video around here the 100% improvement in the DIVX beta . Was all I needed to see. Many differant kinds of improvements can be had with differant programms.

2) I won't even argue this point. Penryn will be more efficient period . Rant and rave all ya want . Intel is brinnging DDR2-3 to the server market. Gets rid of the power consumsion numbers that AMD always rattles. So that will be out of the equasion. I mean AMD is on its 4th or 5th revision already still the power efficiency are looking bad.

3). If Penryn and K10 trade blows on benchies . Intel will be the clear winner. As Penryn can scale to much higher clocks. I don't look for much improvements in new steppings from penryn other than power efficiency. Since A1 steppings are in the wild things are looking much better in the power efficiency area.

4) It has been rumored that K10 or even bulldozer will not have SSE4 . If Amd thinks their a market leader like intel their sadly mistaken . IF intel pu pus sse5 it will die. On the other hand programms will readly optimize for sse4 as Intel is market leader with 80% market share. Some will argue that sse4 will take years to be useful. I agree this could happen . The differance however is it is these same people who waved the AMD 64 flag . Yet here we all are running 32bit. OS.

 

[Jeff007245]

Originally posted by: Nemesis 1

Originally posted by: gOJDO
@bryanW1995

Please point at such case where K10 can actually do +95% more IPC then Clovertown?

First off Thanks Gary for the heads up.

goJoDO. Only 4 days left to see the performance differances in these cpu's.

For me Penryn is the easy winner.

1). WAY more future proof. (SSE4) With performance improvements with every programm that uses it. Since were big on audio video around here the 100% improvement in the DIVX beta . Was all I needed to see. Many differant kinds of improvements can be had with differant programms.

2) I won't even argue this point. Penryn will be more efficient period . Rant and rave all ya want . Intel is brinnging DDR2-3 to the server market. Gets rid of the power consumsion numbers that AMD always rattles. So that will be out of the equasion. I mean AMD is on its 4th or 5th revision already still the power efficiency are looking bad.

3). If Penryn and K10 trade blows on benchies . Intel will be the clear winner. As Penryn can scale to much higher clocks. I don't look for much improvements in new steppings from penryn other than power efficiency. Since A1 steppings are in the wild things are looking much better in the power efficiency area.

4) It has been rumored that K10 or even bulldozer will not have SSE4 . If Amd thinks their a market leader like intel their sadly mistaken . IF intel pu pus sse5 it will die. On the other hand programms will readly optimize for sse4 as Intel is market leader with 80% market share. Some will argue that sse4 will take years to be useful. I agree this could happen . The differance however is it is these same people who waved the AMD 64 flag . Yet here we all are running 32bit. OS.

You talk alot of non-sense, and your posts are a waste of time to read. Where do you get your conclusions? You state things as if they are so true, but so far from it... You really need to start thinking before posting, because you are absolutely talking out of your A$$... I been reading around these forums for the past 1-2 weeks, and everytime I ran into your posts, it's just another bump on the road...

So please... Think harder and take things one step at a time, before even drawing up conclusions and stating it as though they are facts.
This is not allowed. This is a personal attack. And you need to edit your post. PM sent.

UPdate: PM ignored. You have a week off.

CPU moderator apoppin

 

[Zstream]

Apoppin maybe you should control the BS that is flying around here instead of the people that realize the BS and say something.

First of all, part of my duty as a moderator is to prevent personal attacks and to make sure things remain reasonably civil for a discussion. If you want to address the "BS" that is flying around, please remember to address IT - not the poster.

Also, please keep these thread-derailing comments to a PM

CPU moderator apoppin