Clovertown vs. Harpertown and oh yeah Barcelona too...

[qurious69ss]

This is more for a person looking at servers.

http://www.hardwarezone.com/articles/view.php?cid=2&id=2411

 

[yomamafor1]

So this pretty much confirms that Barcelona's single threaded performance is not up to par with Core 2's performance, even in SPEC_fp test. However, due to its native quad core architecture, Barcelona can catch up to Clovertown in rated tests, notably in SPEC_fp rate test.

So making their quad core "native" was, IMO, not an option, but a necessity. As you can see above, core-for-core performance speaking, Barcelona is significantly lagging behind Core 2. But with superior scaling, AMD can still remain competitive in this field. Now all they need to do is to bump up the frequency.

Thanks qurious, very nice find.

 

[qurious69ss]

So this pretty much confirms that Barcelona's single threaded performance is not up to par with Core 2's performance, even in SPEC_fp test. However, due to its native quad core architecture, Barcelona can catch up to Clovertown in rated tests, notably in SPEC_fp rate test.

http://www.hardwarezone.com/img/data/articles/2007/2411/spec-fp.gif
http://www.hardwarezone.com/img/data/articles/2007/2411/spec-fp-rate.gif

So making their quad core "native" was, IMO, not an option, but a necessity. As you can see above, core-for-core performance speaking, Barcelona is significantly lagging behind Core 2. But with superior scaling, AMD can still remain competitive in this field. Now all they need to do is to bump up the frequency.

Thanks qurious, very nice find.

yes, superior scaling is the one strength that AMD still has over Intel, heck even with K8 AMD is more then competitive with memory intensive apps due to IMC and HT. AMD will continue to have a lead there, only problem is that this market is relatively small. AMD should of done more work with the core.

 

[pete4r]

So this pretty much confirms that Barcelona's single threaded performance is not up to par with Core 2's performance, even in SPEC_fp test. However, due to its native quad core architecture, Barcelona can catch up to Clovertown in rated tests, notably in SPEC_fp rate test.

So making their quad core "native" was, IMO, not an option, but a necessity. As you can see above, core-for-core performance speaking, Barcelona is significantly lagging behind Core 2. But with superior scaling, AMD can still remain competitive in this field. Now all they need to do is to bump up the frequency.

Thanks qurious, very nice find.

If you are going to get a Quad Core chip for either Server or Workstation, look at the full potential of that chip, if you are only interested in single core vs core performance, get yourself a single core processor.

Who cares if Barcelona's per core performance is crap, as long as the chip works well with all 4 cores together thats "how it ment to work"

Same as people giving out about Intel's dual or quad core approch, who gives a crap its a "native" or not, as long as it works well with all cores function.

I dont think u'd be a happy man buying a quad core and only 1 core in there that works. or maybe in your case you can moan about per core performance.

 

[accord99]

So this pretty much confirms that Barcelona's single threaded performance is not up to par with Core 2's performance, even in SPEC_fp test. However, due to its native quad core architecture, Barcelona can catch up to Clovertown in rated tests, notably in SPEC_fp rate test.

It's not the native quad-core architecture, it's the superior memory bandwidth available on the Socket F platform that enables Barcelona to perform better in the extremely bandwidth dependent SPECfp_rate benchmark. It's the same reason two FX-74s in a QuadFX setup would outscore a QX6700 by >40% in the same test despite losing nearly every desktop benchmark.

 

[yomamafor1]

If you are going to get a Quad Core chip for either Server or Workstation, look at the full potential of that chip, if you are only interested in single core vs core performance, get yourself a single core processor.

Who cares if Barcelona's per core performance is crap, as long as the chip works well with all 4 cores together thats "how it ment to work"

Same as people giving out about Intel's dual or quad core approch, who gives a crap its a "native" or not, as long as it works well with all cores function.

I dont think u'd be a happy man buying a quad core and only 1 core in there that works. or maybe in your case you can moan about per core performance.

The reason I compare core for core performance is because its a lot more relevant to desktop applications, more relevant to us as enthusiasts. Some games today don't take full potential of quad cores, and most of them are still stuck with single or dual threaded programming. Having a faster throughput doesn't mean your windows boot up time will be faster; it doesn't make you multi-task faster, as programs themselves are still mostly single threaded.

For server, its whole another story. If Barcelona can get its clockspeed up, it'll be a formidable competition for Intel. But for desktop, its inherit IPC is falling short of Core's IPC. Therefore Phenom can only compete with price.

 

[yomamafor1]

It's not the native quad-core architecture, it's the superior memory bandwidth available on the Socket F platform that enables Barcelona to perform better in the extremely bandwidth dependent SPECfp_rate benchmark. It's the same reason two FX-74s in a QuadFX setup would outscore a QX6700 by >40% in the same test despite losing nearly every desktop benchmark.

But because of its native quad core architecture, Barcelona can enjoy such superior memory bandwidth per core without bottlenecking. I guess you're still right though, that its memory bandwidth and fast interconnect that gave AMD the SPEC_fp rate crown.

 

[computertech82]

Wow. I could have sworn I saw benchies that had the Barc running better as a server type (must have been simulated only). AMD got it's butt whipped pretty bad in those test.
I think the Barc performance will be the mark of how the Phenom will be. Not looking good.

 

[ryman554]

Not quite.

Barcelona is going to perform *better* in server workloads vs. penryn than Phenom will on the desktop.

Make no mistake, barcy excels at those memory hungry applications. Just look at the _rate_ scores. Too bad there isn't anything on the desktop that remotely needs such bandwidth...