Phenom/Barcelona Reviews/Previews and Pre-release Discussion

ryanmartini · Sep 10, 2007

there is a guy over @ xs that is going to run a wide series of benchmarks on some barcelonas.

When it starts, maybe I can see if I can repost it with credits to him.

GTengineer · Sep 10, 2007

more benches straight from AMD.....
http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_8796_8800,00.html

Kuroimaho · Sep 10, 2007

TG daily forms their opinion wonder if a TG review will appear or based on other reviews. LINK

On most benchmarks, existing Intel products (65 nm products) are able to exceed Barcelona's performance with very few benchmarks showing up where Barcelona wins.

savageseb · Sep 10, 2007

Kuroimaho said:
TG daily forms their opinion wonder if a TG review will appear or based on other reviews. LINK

that depends on the benching being applied , overall real world benching has already been established.

i would think...

ryboto · Sep 10, 2007

GTengineer said:
more benches straight from AMD.....
http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_8796_8800,00.html

those paint a slightly different picture...

Maverick0984 · Sep 10, 2007

Um, to be fair, you also claimed a whole heap of other things not qualified yet by any review.

]-[itman · Sep 10, 2007

Kuroimaho said:
Well these are all server boards they did their best to give you a hint how it performs.

21% improvement in HL2 over K8 when a 2.4G C2D beat the 3G K8. LINK
Might be enough to catch up but with the clockspeed difference between the two P.

If the anandtech phenom preview holds true, that would put a 2ghz phenom equal to a 2.13ghz Core2; atleast in HL2

AlabamaCajun · Sep 10, 2007

Looks like the Opterpons rip the Xeons a new one on Vitualization and FP but gets kicked in the u-nads with the Int.
"AMD, you have about 6 months to fix the core instruction processor(s).

Edit: Wait

This gets better, lets level the playing field

eats words

http://www.theinquirer.net/?article=42257

Dan0512 · Sep 10, 2007

I found a German site that reviewed a 2Ghz Barcelona chi:

http://www.tecchannel.de/server/prozessoren/1729224/

Dan0512 · Sep 10, 2007

9879DFE550B98A6AC86D27EF680DD1B7_1000x700.jpg.

th3 · Sep 10, 2007

Maverick0984 said:
Um, to be fair, you also claimed a whole heap of other things not qualified yet by any review.

And to be very fair, he also claimed world economics as we know it would break down today (because of the "political" NDA, K10 was so fast that government had to put restrictions on releasing benchmarks or something). Also, Intels HQ should be a smoking ruin by now, they should have surrendered to AMDs superiority and closed down their fabs by now.

Yeah, he really told us so.

Maverick0984 · Sep 10, 2007

th3 said:
And to be very fair, he also claimed world economics as we know it would break down today (because of the "political" NDA, K10 was so fast that government had to put restrictions on releasing benchmarks or something). Also, Intels HQ should be a smoking ruin by now, they should have surrendered to AMDs superiority and closed down their fabs by now.

Yeah, he really told us so.

Yea, that too.

gone_fishin · Sep 10, 2007

Dan0512 said:

Why is anyone excited about this, it is clearly lacking (sorry fanboys)?
This is a server chip that needs an expensive server motherboard that only supports expensive ecc memory. Not in the desktop enduser category.
Besides the reviews make it clear, the OLD Intel design beats this thing and the new Penryn based design is almost out the door.

They need to use this as a stepping stone and release the next generation asap to compete.

AlabamaCajun · Sep 10, 2007

Whatever, Intel is just history I guess

Great benches Dan, the Spec int needs to be redone because it was compiled for intel procs which is something that both AMD and Intel do.

]-[itman · Sep 10, 2007

gone_fishin said:
Why is anyone excited about this, it is clearly lacking (sorry fanboys)?
This is a server chip that needs an expensive server motherboard that only supports expensive ecc memory. Not in the desktop enduser category.
Besides the reviews make it clear, the OLD Intel design beats this thing and the new Penryn based design is almost out the door.

They need to use this as a stepping stone and release the next generation asap to compete.

I'm not quite sure what you're getting at. The reviews we have show the K10 architecture in servers because AMD is releasing their Opteron line today, not the Phenom (desktop). When Phenom is released, we'll see desktop boards with non-ecc RAM and then we'll see how K10 does on desktop, until then the best we have is approximations.

BTW, atleast in the server environment, K10 looks pretty impressive clock for clock, on average beating a quad-core xeon clocked 16.5% higher.

gone_fishin · Sep 10, 2007

]-[itman said:
I'm not quite sure what you're getting at. The reviews we have show the K10 architecture in servers because AMD is releasing their Opteron line today, not the Phenom (desktop). When Phenom is released, we'll see desktop boards with non-ecc RAM and then we'll see how K10 does on desktop, until then the best we have is approximations.

BTW, atleast in the server environment, K10 looks pretty impressive clock for clock, on average beating a quad-core xeon clocked 16.5% higher.

I recall reading in the real world benchmark of aerodymics calculations getting smoked handily by lower clocked xeons. Also somewhere in there they were suggesting that the L3 cache is poorly implemented and takes away from its on die memory controllers latency advantages.

]-[itman · Sep 10, 2007

gone_fishin said:
I recall reading in the real world benchmark of aerodymics calculations getting smoked handily by lower clocked xeons.

You can't base your conclusion of an architecture with the results of one benchmark. I could do the same with WinRAR and say that an Opteron kicks Xeon butt needing only 2ghz to give the same performance as intel's flagship quad core at 2.93ghz.

Also somewhere in there they were suggesting that the L3 cache is poorly implemented and takes away from its on die memory controllers latency advantages.

It's not that it's poorly implemented, that's just the reality of L3 cache, it's a trade-off. Yes you will increase your latency when going to main memory, but most of the time you should be able to find much of the info in L3 cache, so you don't need to go to main memory thereby increasing performance. In some situations it will help significantly, in others, it may hurt a little. Obviously AMD felt that it's testing showed the L3 cache tends to help more than it hurts.

gone_fishin · Sep 10, 2007

]-[itman said:
You can't base your conclusion of an architecture with the results of one benchmark. I could do the same with WinRAR and say that an Opteron kicks Xeon butt needing only 2ghz to give the same performance as intel's flagship quad core at 2.93ghz.

It's not that it's poorly implemented, that's just the reality of L3 cache, it's a trade-off. Yes you will increase your latency when going to main memory, but most of the time you should be able to find much of the info in L3 cache, so you don't need to go to main memory thereby increasing performance. In some situations it will help significantly, in others, it may hurt a little. Obviously AMD felt that it's testing showed the L3 cache tends to help more than it hurts.

I like how the editorial on the front page of this site from today supports my view that this is just a stepping stone and is nowhere near dethroning Intel.

The high voltage he points out to reach 2.5Ghz is an eye opener. Also the die size is more than the clovertown and the Penryn based tigerton is to be even smaller, so no price per chip advantage. I see no competition here.

AlabamaCajun · Sep 10, 2007

]-[itman said:
You can't base your conclusion of an architecture with the results of one benchmark. I could do the same with WinRAR and say that an Opteron kicks Xeon butt needing only 2ghz to give the same performance as intel's flagship quad core at 2.93ghz.

It's not that it's poorly implemented, that's just the reality of L3 cache, it's a trade-off. Yes you will increase your latency when going to main memory, but most of the time you should be able to find much of the info in L3 cache, so you don't need to go to main memory thereby increasing performance. In some situations it will help significantly, in others, it may hurt a little. Obviously AMD felt that it's testing showed the L3 cache tends to help more than it hurts.

Agreed, a few benches don't make the difference, it's the overall performance where it will show. I'm sure Intel is going to hold some areas and niche markets but that is to be expected.

L3 is not a stop off on the way to L2, it's only used when needed.

AlabamaCajun · Sep 10, 2007

gone_fishin said:
I like how the editorial on the front page of this site from today supports my view that this is just a stepping stone and is nowhere near dethroning Intel.

The high voltage he points out to reach 2.5Ghz is an eye opener. Also the die size is more than the clovertown and the Penryn based tigerton is to be even smaller, so no price per chip advantage. I see no competition here.

The high voltage more than likely is because he is running on an ES and is expected to run it up to test results. As for chip are you including both chips or one and also factor in the weld process, thats 775 plus interconnects and power to the second chip.

Phenom/Barcelona Reviews/Previews and Pre-release Discussion

ryanmartini

Member

GTengineer

Member

Kuroimaho

Member

savageseb

Member

ryboto

Member

Maverick0984

Member

]-[itman

AlabamaCajun

Member

Dan0512

Member

Dan0512

Member

th3

Member

Maverick0984

Member

gone_fishin

BandSaw King

AlabamaCajun

Member

]-[itman

gone_fishin

BandSaw King

]-[itman

gone_fishin

BandSaw King

AlabamaCajun

Member

AlabamaCajun

Member

Similar threads