Intel Quad VS AMD Phenom

[myrlar]

Hi,
I'm not trying to flame here, but I am reading a lot of posts (off this forum) about how the new AMD quads (Phenom/Barcelona) will walk over the Intel quads.

Is this true? I haven't seen any benchmarks anywhere... Am I blind?

To tell the truth, all I've read is articles on AMD boasting about their new chips... No prices, no figures...

Any comments?

Is it worth waiting for the Phenom to shows its fangs before I jump and buy a new Intel quad?

 

[Sable]

Crivens.

No one has seen any real numbers for the Phenom/Barca so it's ALL just guess work at the moment.

I've seen people say that clock for clock the Barca is slightly faster. BUT AMD won't be releasing any high clocked (ie matching Intels clocks) before December. They're releasing at 2.1ghz and they need 2.6ghz to be competative.

AND Intel also have Penryn to roll out which should/could/might give them back the performance crown. IF AMD manage to take it.

NO ONE KNOWS!!!!!!!!!!!!!!!! Buy now if you need it, wait for benchmarks if you don't.

 

[Calin]

Until now, AMD says its processors are faster clock for clock than Intel's. This is only a rumor - until now, there are no hard results to make a comparison. Judging by recent history, AMD is not where he says it is, and unfortunately, even with a better processor AMD is forced to wage a price war against Intel.
The only thing which a very good AMD processor could bring would be some lower prices in Intel's camp. So, if you find current quad core prices bearable, and need (or can use) a quad core, you could buy one. If not, you could wait for them to become more common, and decrease in price. I won't hold my breath until AMD builds a better processor than Intel

 

[LightningRider]

Yeah, AMD is just blowing hot air for all we know. There have been no real concrete benchmarks so no one can know. They will probably be a little faster or at least just as fast as the current Core 2's when they finally release but I'd be surprised if they surpassed the Penryn as well.

After the 22nd it will be a great time to buy a new Quad if you feel the need with the prices dropping and all. They are powerful chips and I'd say if you're in the market for one, go for it. You can always wait for the new CPU's though. Penryn (Intel's new CPU line to be released) should be coming around December. AMD's new offering (Phenom) will probably be around the same time.

Personally I'm waiting for Penryn, I think Intel is still going to have the performance crown with Penryn and Phenom compared side by side. But no one can know for sure at this point.

 

[Miramonti]

Claims about them are bold...yet they are delayed...do the math to this point...

 

[gobucks]

I really hope AMD gets competitive soon, because we all know how lazy intel can get when AMD isn't breathing down its neck.

Unfortunately, the more I look at the numbers, I really just don't see it happening, at least short term. Right now, K8 is 20% slower clock-for-clock than Conroe. From the benches I've seen, it looks like Penryn will be another 10% faster than Conroe, meaning that Barcelona needs to have at least a 30% clock-for-clock advantage over K8 just to be competitive clock-for-clock with Penryn. Then, on top of that, Intel looks more than capable of producing chips clocked almost twice as high (3.33GHz) as AMD (1.6GHz so far). If Conroe is any indication, Penryn probably has a similar cushion of frequency headroom, meaning that there will probably be a decent number of Penryns that will be capable of 4GHz, AKA 2x AMD's targeted (and so far unable to produce) launch speed of 2GHz. Maybe AMD will be able to get to a 45nm die shrink sooner than we think, and this will allow barcelona to hit really high speeds, but the results so far are not encouraging.

Then, looking at the Barcelona architecture makes me even more depressed. The caches are ridiculous - did anybody else notice that they actually REDUCED the size of the L1 cache from 128K to 64K, AKA the same as Conroe's? The large on-die cache was one of AMD's biggest advantages in design, and they just threw that advantage out the window. On top of that, the L2 is still only 512K, 1/6 or 1/12 the size of those on Penryn. As for the L3, it is both small (2MB) and slow (friggin level three!!!!!) so I don't see why they decided that that shared L3 is going to answer all their problems. Additionally, AMD seriously needs to step it up with memory bandwidth - socket 939 used almost every bit of its DDR bandwidth, but AM2's DDR2 bandwidth efficiency was piss poor compared to any of Intel's chips. The only advantage I see is the native Quad-core design that doesn't rely on the bus for communication between chips, but benchmarks of current Kentsfields don't really show this to be a huge problem - the bus is fast enough to accommodate this, resulting in maybe a 2-3% performance hit. Of course, this is only an issue with heavily multithreaded apps, and very few programs currently use more than 2 cores simultaneously.

 

[bryanW1995]

wow, where did you get all of this info? AMD doesn't need as much cache as intel b/c they're not using by fsb. Can't wait to see what nehalem can do without this bottleneck. According to AT, amd is currently producing 2ghz barcelonas. They are still getting killed by intel, but at least get your facts straight.

I agree with you that we should be seeing lots of penryns at 4ghz in dec. AMD will probably be ok on a clock for clock basis with phenom vs penryn, but the problem is this: if they can't reliably make a barcelona over 2.0 ghz, how much overclocking headroom will there be? In the server market that's fine because nobody cares about overclocking, but phenom is going to get killed without a serious clock for clock advantage over penryn. Let's just assume that AMD pulls a rabbit out of their hat and introduces phenom in nov at 2.6 ghz. Are they really going to have a 28% clock for clock advantage over penryn? I think that we all know the answer to that one.

 

[Phynaz]

During the AMD Q2 conference call, AMD pretty much admitted they will not have the top performing processors.

 

[SickBeast]

By the time Phenom is in mass-production and widely available it will be 2008. I can't see it beating out what intel has at that point, but I suppose time will tell.

It's not really a fair comparison seeing as the intel quads have been on the market for a long time.

If you want to compare features:

- Barcelona will be able to independantly adjust clockspeeds/voltages for each core. This could allow for higher overclocks.

- Barcelona is a true 'quad core' implementation whereas intel has simply stuck two dual-cores together

- Barcelona has a little more power at a given frequency

 

[bryanW1995]

so you're saying that you might be able to get a 2.6 phenom and overclock core 0 to 3.3, core1 to 3.0, and leave core2 & 3 at 2.6? Will that be in the bios? Wow, that could make overclocking even more complicated than it already is!

 

[AllGamer]

well... until Intel came out with their "Core" series or CPUs AMD always had the Lead in the CPU battle through the P4 series

clock by clock intel was a no fight.

but now that the Core series are out, they are sort of at par, but Intel has the lead.

as of this moment current Dual Opteron Duo or Quad, does perform a little better at games and business applications, than the current line of Core Duo Xeon or Xeon Quads setups.

the only lead Intel has againts AMD is on the number crunching and video encoding area.

That being said... if the new AMD CPUs really delivers, it'll definitely leave the Intel Core series in the Dust.

and i do hope that becomes real, because that's the only way to keep both Intel, and AMD in the fight

is like going up the ladder, if both give up, or if either one fell off the ladder, then it becomes a Monopoly, both in market and in Innovation.

and everyone knows having a Stagnated CPU industry is bad for science.

If we want to see more sophisticated AI, machines, CPU, etc, you better hope each company keeps wining one round after another, if either of them wins too many rounds in a row.... that's a unhealthy sign

 

[Phynaz]

AllGamer,

AMD did not always have the performace lead over the P4. During the last year and a half of the P4's life they did. Otherwise is was see-saw thing. For example, the Northwood P4 left the Athlon in the dust.

C2D's are faster than Opterons just about everything. Just look at the benchmarks here.

Don't worry about AMD delivering. At 2Ghz they will be slapped silly by 3.33 Ghz Penryn's.

 

[bryanW1995]

sadly, you are correct. they will lose at 2.6, but at 2.0 it will be embarrassing. Intel just announced that the first 45nm extreme edition will be a 3.33 ghz penryn in q4. I'm guessing that it will be out about the same time the quad cores came out last year.

 

[zsdersw]

Originally posted by: SickBeast
- Barcelona is a true 'quad core' implementation

Nonsense. Intel's quad-core processors aren't "false quad core". Using "true" to describe AMD's current approach to quad-core CPUs is wrong.

"True" quad-core processors are all processors that have 4 cores in one chip package. That is not an AMD-only thing.

If you want to refer to the differences between Intel's current approach to quad-core and AMD's upcoming approach to quad-core, that's fine (and appropriate).. but Intel's quad-core products are no less "true quad-core" than AMD's.

 

[amenx]

It would look pretty embarassing for AMD if their "True Quads" didnt show significant performance advantages over Intels "false quads" when they come out. I think they know that.. big question is if this has anything to do with the delays in bringing these fabulous quads out.

 

[covert24]

Originally posted by: zsdersw

Originally posted by: SickBeast
- Barcelona is a true 'quad core' implementation

Nonsense. Intel's quad-core processors aren't "false quad core". Using "true" to describe AMD's current approach to quad-core CPUs is wrong.

"True" quad-core processors are all processors that have 4 cores in one chip package. That is not an AMD-only thing.

If you want to refer to the differences between Intel's current approach to quad-core and AMD's upcoming approach to quad-core, that's fine (and appropriate).. but Intel's quad-core products are no less "true quad-core" than AMD's.

true in this tense means all cores on one die. Intel just combined 2 core 2 duos in order to make the quad. AMD is putting a bit more research and engineering into it and putting all 4 cores on one die.

Also 4 cores on one die will be a lot easier to cool than 2 dies apart from eachother since the heat will be coming from one centralized place rather then 2 big locations.


And also the reason why the AMD quads havnt come out is because they have been doing more work and research into things. This is the main reason why AMD is lagging behind. Intel just put 2 of there already known chips together.

 

[bryanW1995]

I think that amd should have tried the intel approach and gotten to market 6 months earlier.

 

[covert24]

it would have been nice but i would rather have a AMD cpu that has been well engineered then some glued together piece.

 

[bryanW1995]

My last intel cpu was a 700 mhz PIII. I've bought a 1.2 tbird, xp 2500+, xp 3000+, A64 3700+, FX55, and opteron 180 since then, so I've been very loyal to amd over the years. I didn't want to switch back, but intel's lead is overwhelming right now. You can throw around words like "well engineered" all you want, but that won't change the fact that intel is kicking amd's a$$ right now, and it doesn't appear likely to change anytime soon.

 

[apoppin]

Originally posted by: covert24

Originally posted by: zsdersw

Originally posted by: SickBeast
- Barcelona is a true 'quad core' implementation

Nonsense. Intel's quad-core processors aren't "false quad core". Using "true" to describe AMD's current approach to quad-core CPUs is wrong.

"True" quad-core processors are all processors that have 4 cores in one chip package. That is not an AMD-only thing.

If you want to refer to the differences between Intel's current approach to quad-core and AMD's upcoming approach to quad-core, that's fine (and appropriate).. but Intel's quad-core products are no less "true quad-core" than AMD's.

true in this tense means all cores on one die. Intel just combined 2 core 2 duos in order to make the quad. AMD is putting a bit more research and engineering into it and putting all 4 cores on one die.

Also 4 cores on one die will be a lot easier to cool than 2 dies apart from eachother since the heat will be coming from one centralized place rather then 2 big locations.


And also the reason why the AMD quads havnt come out is because they have been doing more work and research into things. This is the main reason why AMD is lagging behind. Intel just put 2 of there already known chips together.

and yet it works so well for intel ... those "glued together" CPU's have the performance crown so clearly

... and while AMD is *bleeding* hundreds of millions of dollars each quarter; losing marketshare , intel is raking the bucks in - all the while working on their own native Quad-core and building market-share

Best of all ,intel's CPUs are not futureware .. you can buy them right now
[i just made that name up, AMD's intel-beating product = "futureware"]

i think intel's method is working better from a business standpoint and they will certainly be in business next year.

Maybe AMD can turn to Microsoft for a deal with the devil.

 

[zsdersw]

Originally posted by: covert24
true in this tense means all cores on one die. Intel just combined 2 core 2 duos in order to make the quad. AMD is putting a bit more research and engineering into it and putting all 4 cores on one die.

No, you don't get to determine what "true" means. The term "true quad-core" doesn't detail anything about the particular method of quad-core being implemented. The correct usage would be "single-die quad-core". When you say "Barcelona is true quad-core" you imply that there somehow aren't really 4 cores on Clovertown or Kentsfield when, in fact, there are.. and that's what's wrong with the term.

Also 4 cores on one die will be a lot easier to cool than 2 dies apart from eachother since the heat will be coming from one centralized place rather then 2 big locations.

More nonsense. Last time I checked, the heat spreader touches both dies in an Intel quad-core CPU, just like it touches the single die in Barcelona.

The difference in cooling performance between the two will be provided by the power management features of the chips themselves, not whether the heat spreader is touching two dies or one.

 

[Phynaz]

Originally posted by: bryanW1995
I think that amd should have tried the intel approach and gotten to market 6 months earlier.

AMD (Hector Ruiz) said once in an interview that he wished he had, but he didn't have the money to fund the development.

 

[zsdersw]

Originally posted by: Phynaz
AMD (Hector Ruiz) said once in an interview that he wished he had, but he didn't have the money to fund the development.

Fund the development? But.. wait a minute.. covert24 basically said that Intel didn't do any development for their quad-core chips; they just "glued together" existing chips. :roll:

 

[oscar6]

"Alright AMD .... , just take the glue, put a little hear and there. Easy as pie. *whistling*

 

[myocardia]

Originally posted by: Phynaz
AllGamer,

AMD did not always have the performace lead over the P4. During the last year and a half of the P4's life they did. Otherwise is was see-saw thing. For example, the Northwood P4 left the Athlon in the dust.

That isn't what he said. He said "well... until Intel came out with their "Core" series or CPUs AMD always had the Lead in the CPU battle through the P4 series. clock by clock intel was a no fight. " He was saying that AMD's chips have always had a clock-for-clock speed advantage. And yes, except for the last 1/3 of the PIII's lifespan, AMD has always had the clock-for-clock advantage, since the original Athlon debuted at 500 Mhz.