Phenom/Barcelona Reviews/Previews and Pre-release Discussion

[hUMANbEATbOX]

I watched the "yawn" intro seminar running tonight and posted it here but they did show some benches with impressing "looking" results. I think by weeks end we will have an idea what the platform will do. Hope S7 can get his board cooperating.

I hate those types of PR events. Same with Apple and the new ipods. ZZZZZ..

All I have been waiting for is to see someone with some skills, equipment and some time to tear these new cpu's a new one. If they look like they are fun to OC (aside from the performance "crown"), then I may try one in the future (after an Intel upgrade mind u ).

Phenom will be the true testing ground in terms of OC. Definitive results won't be common knowledge until these have been around for awhile and there are some mature boards/BIOS out. Sucks to be us.

Unfortunately, the one thing that is immediately apparent is that AMD is not going to have any reprieve when it comes to pricing wars. Margins are getting better for Intel and worse for AMD (excuse my perhaps limited knowledge in the realm of finance, just seems to make sense to me).

 

[inkfx]

Yeah I can't wait to see the Phenom results. I'd be so excited going back to AMD. That is where all the magic originated for me.

 

[Dan0512]

I personally want to see a single socket Phenom system running in its HT3 greatness.

dan

 

[MadMan007]

It's server chip and it is much improved over its opteron predecessor. Why so much fuss?
I don't know why anandtech did not use the Asus 4x4 board instead, since that board is the high end gaming platform. It would have made more sense as far as performance in desktop is concerned. Does anyone know why that board was not used?

My guess would be BIOS compatability.

 

[savageseb]

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.

niiiceeee!!

 

[Zerix01]

This is the kind of shoddy comparisons that makes me laugh. Comparing an 8 core Barcelona system (look at the screenshot from your link with taskmanager clearly showing 8 cores) vs a dual core 5150, that they did not even provide a taskmanager screenshot for so it was probably running with just one dual core. So an 8 core vs 2 core hahahahahah, get some real credible benches.

Okay Ha, ha, I missed that they were running in dual socket, and I thought that intel processor was a quad. But on the other hand they do not mention how many sockets they have the 5150 running in and I'm wondering why the auther was surprised to see the Barc, beat the Xeon. I'm thinking at least dual socket, no one would be dumb enough to compare 2 cores vs 8 (other than me not RTFA, lol). But even 4 vs 8 is a kinda strange, can this Xeon run in a quad socket setup?

 

[Douken]

The only thing that I'm concerned about is how the Phenom will OC. C2D can easily OC 40-60% over its default clock, I would like to see the same or at least something close with the Phenom. Just Imagine some 3.5GHz Phenom folding for you.

 

[Kuroimaho]

I do not expect them to oc better than K8, but if penryn can go to 4G according to Intel AMD can do that to at 45nm hopefully. Their steppings get better and better over the time check final windsors but to wait till they come out gonna be a hard task.

 

[AlabamaCajun]

I do not expect them to oc better than K8, but if penryn can go to 4G according to Intel AMD can do that to at 45nm hopefully. Their steppings get better and better over the time check final windsors but to wait till they come out gonna be a hard task.

I'm with you on the speedup occuring at 45nm. AMD will now focus on productivity. I think engineering will continue on two fronts in the CPU area. One team going at the 45 nm migration, the other working on revisions to the 65nm K10 chip family. From what I've read, 45 nm will be more than a shrink with new metals/elements used in the diffusion process. I also think a new masking technology is going to be used as light based lithography is having a problem with diffusion at the edges of the shadow mask scattering the light particles where is passes through shrinking transparent regions.

 

[AlabamaCajun]

No doubt that AMD has to keep the throttle balls to the wall in order to keep up. I've seen some impressive results and it matches the engineering specs for Barc. Sure there was hype and the ever present intel compiled programs running around that will give chipzilla an edge. AMD has big support in the market place and US and International companies realize this adding that the German government kicked in some spare d'marks for economic development. Barcelona is an Innovative product that is going to change the way we work and play. Now we just need to see the Phenoms and focus on what comes our way.

 

[Super Nade]

He complains despite being Numero uno fansite they didn't get a bunch of servers to test.
Btw do we really need new bench threads to be opened by everyone ?

Correct. No need for a thousand threads on the same topic. I'm closing this one. Feel free to re-post your thoughts here-->

http://www.ocforums.com/showthread.php?t=526901

Content is not lost as I merged this thread with the one titled "Barcelona Reviews".

 

[savageseb]

http://anandtech.com/IT/showdoc.aspx?i=3091&p=2

 

[]-[itman]

The Q6600->QX6800 scaling is more like 85% in Lightwave, performance increases by 21% with a 25% change in clockspeed. And like I said before, the K10 example used a much lower resolution for the games.

It's 70.2% efficiency; 17.55% performance with 25% clock speed gain:
[(129.9-107.1)/129.9] x 100= 17.55%; Then to check, .1755 * 129.9 = 22.8, 129.9 -
22.8 = 107.1

It's not scheduling, it's the fact that memory latency is slower with the second socket. No OS can fix it.

No OS can fix it, but like I said, with better software, you can have less impact. That's exactly how they get so much power out of game consoles that are basically PC's, better software for the same hardware. You may be right, the added latency may bring the results closer together, but when Anandtech states that these results should represent a minimum increase, when they are full aware of the added latency, I tend to trust the results as a good baseline.

We have AMD server platform vs Intel server platform on Techreport with desktop and workstation like applications.

Again, in the Anandtech preview they state that the K10 responds very well to memory bandwidth, moreso than any previous chip, so you cannot say that AMD loses when we only have an estimate as to how it will perform in those applications. Besides that, the Techreport review didn't show a landslide for intel as you suggest. I count AMD with 3 wins (4 if you include the energy results) versus intel's 5. Folding and WME I have as ties. In WME, the 2ghz K10 loses slightly to the Xeon, but the K10 scales much better in WME and the 2.5ghz K10 would beat a 2.5ghz Xeon, even if the Xeon's scaling was 100% between 2.33ghz and 2.5ghz. I throw out the Sandra results as this is not a workstation benchmark, it is purely synthetic and obviously does not represent real world results. I also threw out the 2nd pov-ray benchmark as they said that the Opteron server's NUMA system was causing problems and a dekstop/workstation platform would not use this system.

I'm not saying that a Phenom is going to destroy a Core2, but I'm saying it's not a clear win either way and we should wait before drawing any firm conclusions.

 

[juane414]

I don't care about the "i'm better than you mumbo jumbo"... I just want to see some 1M superpi results

 

[SLee]

It's 70.2% efficiency; 17.55% performance with 25% clock speed gain:
[(129.9-107.1)/129.9] x 100= 17.55%; Then to check, .1755 * 129.9 = 22.8, 129.9 -
22.8 = 107.1

That's not the correct way to calculate efficiency with benchmarks in seconds. Imagine if the time was 65s for a theoretical 4.8GHz, this equation would calculate the efficiency as [(129.9-65)/129.9] = 0.5. Divided by 100% clock speed gain, this only gives an efficiency of 50%, when the actual efficiency is 100%.

Besides that, the Techreport review didn't show a landslide for intel as you suggest. I count AMD with 3 wins (4 if you include the energy results) versus intel's 5. Folding and WME I have as ties. In WME, the 2ghz K10 loses slightly to the Xeon, but the K10 scales much better in WME and the 2.5ghz K10 would beat a 2.5ghz Xeon,

Or maybe they had some testing problems, considering there's no reason the 2.33GHz should score the same as a 2GHz, when the 3GHz shows that there is no bandwidth restriction. WME is clearly a win for Xeon, for Folding the Xeon wins in the most important gromacs core.

I also threw out the 2nd pov-ray benchmark as they said that the Opteron server's NUMA system was causing problems and a dekstop/workstation platform would not use this system.

Then the 1st Pov-ray test is also irrelevant. The 2nd test is the official benchmark.

 

[savageseb]

this is simple,

you put 2 quad opteron barcelonas @ 2.0 ghz
vs
2x E5300 2.33 clovertown quad xeons.

what do you get?

price wise and performance wise, a kick in the nuts for intel.

http://anandtech.com/IT/showdoc.aspx?i=3091&p=11

http://anandtech.com/IT/showdoc.aspx?i=3091&p=10

and given the "how" of AMD´s native quad it is sillier to bench on super pi 1m for drawing out conclusions.

and we have yet to see what an Overclocked opteron 2200 or 2300 can do.

 

[MadMan007]

this is simple,

you put 2 quad opteron barcelonas @ 2.0 ghz
vs
2x E5300 2.33 clovertown quad xeons.

what do you get?

price wise and performance wise, a kick in the nuts for intel.

http://anandtech.com/IT/showdoc.aspx?i=3091&p=11

http://anandtech.com/IT/showdoc.aspx?i=3091&p=10

and given the "how" of AMD´s native quad it is sillier to bench on super pi 1m for drawing out conclusions.

and we have yet to see what an Overclocked opteron 2200 or 2300 can do.

At least the second one is an appropriate 'cherry-pick one benchmark as an example' benchmark. But in the first link the dual Xeons crush the Opteron, both in absolute terms and in per-clock terms. Way to link the exact opposite of what you wish to be true savage!

What this does show is that each app is different on each CPU and having a clear-cut all around better CPU may just not happen this round for Phenom vs Penryn.

 

[savageseb]

At least the second one is an appropriate 'cherry-pick one benchmark as an example' benchmark. But in the first link the dual Xeons crush the Opteron, both in absolute terms and in per-clock terms. Way to link the exact opposite of what you wish to be true savage!

What this does show is that each app is different on each CPU and having a clear-cut all around better CPU may just not happen this round for Phenom vs Penryn.

way to kick my good will out the door madman.

it seems you missed the part where dual socket opteron @ 2ghz was only 5 points away dual socket xeon quad at 2.33ghz.

but i think we all know where it is that amd had a 30% disadvantage.

 

[MadMan007]

Actually by showing those two benchmarks one would think you'd be trying to show a balanced viewpoint. Were you? A 'kick in the nuts' for Intel would make one think not, and it's very unlike pretty much all your Barcelona FTW! posts.

 

[Neuromancer]

AMD is touting the new opterons as the performance per watt kings. Something IT departments are very keen on, so that growth in the server market they grabbed over hte last couple of years looks to be on steady ground.

I do hope the phenoms come out at least equal to the conroes in benchmarking ability. I know they will outperform it in general repsosiveness of hte desktop... heck my old stock 4600 x2 FEELS snappier then my 3.2GHz conroe... it just gets totally decimated in every benchmark but memory based ones.