The Phenom vs. Athlon Core Shootout

[gpippas]

http://www.tomshardware.com/2007/12/19/amd_phenom_athlon_64_x2/

Yes there is a small improvement over K8 cpu's but not only is it small its clock for clock. The highest available clocked Phenom is 2.3ghz. The X2 comparison would be a 4400X2 running at 2.3ghz. If you upgrade from that you will get a 15-20% increase in performance. However if you have any model clocked higher or overclocked higher you are essentially downgrading to a Phenom. Anybody with a decent Core or Quad would also be downgrading.

On another note AMD has now officially released a Phenom 9600 Black Edition. It will cost the same amount as a regular 9600 just with an unlocked multiplier. Judgeing from the above review and other sites the unlocked versions should at least hit 2.6ghz. So they are essentially releaseing a Phenom 9900 but absolving themselves of all responsibility if it doesn't work.

AMD included the warning, "AMD is not responsible for damages caused by overclocking (even when overclocking is enabled with AMD OverDrive(tm) software)".

Here's the link
http://www.tomshardware.com/2007/12/19/amd_releases_unlocked_9600_black_edition/

 

[zenmaster]

Seems like somebody at THG really has a thing for this Phenom.

Hey, that's nice that it is slightly faster clock for clock, but I don't even trust that. I would be really curious about some of the settings they tweaked such as the NB which AMD reduced to 1.8 from 2.0 in the ES samples. Did THG tweak that back up? From what I read that really knows a few % off the Phenom speed. If suspect they did.

On top of that, why no comparisons vs the faster X2 Chips at full speed?


The Phenom has a shared L3 Cache so the single core got the whole thing. As you add cores, the relative gain from that L3 probably goes down as more cores are competing for that cache. The result would be sort of like taking the L3 cache and splitting it in half.

In fact, the whole Single Core thing was Silly.
Test Using 4cores and 2cores that is how they will be used IRL.
Test at Stock. Test @2.6. Test the X2 @ Stock. Test the X2 @3.4

Why Only Compare on OC'd chip vs an UnderClocked Chip.
I get the point of Clock per Clock per core, but I suspect the much cheaper X2 will win most test of Stock vs Stock and OC vs OC.
The Question is by how much?

 

[runswindows95]

I agree, zen. I would love to see someone compare them as is.

 

[gpippas]

I agree with you Zen. I don't see how its achieved anything. And throughout the article they mention that AMD said it was faster the X2 by 20%. In the conclusion they said its not as high as that but didn't table or graph anything to show the actual % difference.

 

[turboflame]

The Phenom has a shared L3 Cache so the single core got the whole thing. As you add cores, the relative gain from that L3 probably goes down as more cores are competing for that cache. The result would be sort of like taking the L3 cache and splitting it in half.

That was the first thing that came to my mind

I don't mind THG and other reviewers backing AMD to help them out but this was blatantly deceptive

 

[shabodah]

I don't think we really need more articles about the fact that phenom is not as fast clock-for-clock as expected, but like this article, we DO need more helping to figure out WHY it is. Personally, I'd like to know what is going on with the northbridge speeds. If the AM2+ boards are designed to lock the IMC at 1.8ghz, even if the processor is running at 2.2ghz or 2.3ghz, what happens in a regular AM2 board? Does it actually run at processor speed like regular x2's do? The AM2 Windsor chips were known to run slower than 939 chips when the IMC was below 2.4ghz. This seems very similiar, and yet, somehow, no one is talking about it.

It is also good to hear that SOMETHING is going on with Z-Ram. I really expected to see it implemented in some way by AMD by now, especially since they could now use it effectively in GPU's.

 

[dragonsprayer]

MORE BIAS!

The socket am2 came way after amd dual cores and after socket 775 - THG says a seamless transition from single core to quad?????

This is total bs!

amd has an advantage over intel? once again THG making stuff up to help even the playing field!


you guys are really losing credibility with statements like this:

"In fact, AMD has a pretty significant advantage over Intel when it comes to upgrading existing systems with a quad core processor."

 

[dragonsprayer]

you can drop a celron in a quad core mobo too - lol - if you like.
most single core users are on socket 939 - anyone with a decent system with am2 used a dual core - com on thg!

"As a consequence, it is technically possible to deploy a quad core Phenom processor into a Socket AM2 motherboard that has been running an Athlon 64 or Athlon 64 X2 - all you need is a BIOS update"



this is my last word on this subject:

advantage? - its called ddr3! try dropping that in your am2 mobo!

 

[gpippas]

"In fact, AMD has a pretty significant advantage over Intel when it comes to upgrading existing systems with a quad core processor."

Thats what really got me aswel. If it wasn't for that I wouldn't have started the thread. Thats why I said what I said. And this is coming from an AMD fan. I just can't stand for misinformation like that regardless of my bias. The only people who should consider an upgrade to current Phenoms are either running Pentium 4's, single core Athlons or have a 3800X2 which they havn't overclocked and should have.

 

[nightscope]

"In fact, AMD has a pretty significant advantage over Intel when it comes to upgrading existing systems with a quad core processor."

Thats what really got me aswel. If it wasn't for that I wouldn't have started the thread. Thats why I said what I said. And this is coming from an AMD fan. I just can't stand for misinformation like that regardless of my bias. The only people who should consider an upgrade to current Phenoms are either running Pentium 4's, single core Athlons or have a 3800X2 which they havn't overclocked and should have.

No, everyone will see an upgrade if they use software that optimizes all cores. And yes, AMD does have an advantage over Intel in the upgrading current platforms issue.

If you have an Am2 board, you can drop $199 and get yourself a quad core. Yes you can also do that with some socket 775 boards, but many of the new Penryns are having compatibility issues with current 775 boards. Yes, they may be fixed with a bios update, but that's not a sure thing.

 

[someguy7]

Why did they bother to not use all the cores. It doesnt make sense to me. Anybody with the quad is going to use all 4 cores. Anybody with a dual is going to use both cores.

 

[turboflame]

Why did they bother to not use all the cores. It doesnt make sense to me. Anybody with the quad is going to use all 4 cores. Anybody with a dual is going to use both cores.

This particular article was an attempt to measure performance increases in the processor's architecture. So they put a single K8 core against a single K10 core to see if the design is more efficient or not regardless of the number of cores it has.

In real world applications, you're right that it is pretty much meaningless.

 

[Ycon]

Why use an oced processor?

 

[BaronMatrix]

Well, I feel better knowing that you guys will attack ANYONE with anything positive to report about AMD. That wasn't the worst review I've seen, but it did have some mistakes.

I would have suggested both single and dual core to see what kind of scaling difference there is between the two.

A quick tabulation says K10 is about 20% faster overall, but the memory bandwidth for Sandra is troubling. Nothing should beat a Phenom with 1066 RAM.

 

[gpippas]

What I would like to see is the performance increase of an X2 proc with ddr2 1066 compared to ddr2 800. Because that might be a good upgrade. It might even give a little insight into what AM3 will be like.

 

[leo2kp]

What about using both X2 cores and two cores out of the quad showing how L3 between cores affects things?

 

[quantumsheep]

I myself do not think that Phenom is a failure. However i do believe the launch was. I believe that AMD panicked and skipped over another few weeks of final tweaking to get the product to launch Q4 07 rather than Q1 08, which was a very bad mistake. Those extra couple of months could've allowed them to fix all bugs and ramp up production at some decent clockspeeds.

So to conclude, their only mistake was releasing the chip before it was ready. However, this was one massive ****-up.

 

[gpippas]

TBH I don't think even with improved stepping Phenom's performance can increase. Obviously the clocks can be increased. But to what level. K8's were a bitch to get past 3ghz and the k10 arch is still very similar to K8's.

I think AMD knows it isn't going much higher. A 2.6ghz Phenom is called a 9900. Where do you go from there naming wise? They have already said 9X50 is B3 stepping.

 

[turboflame]

I think AMD knows it isn't going much higher. A 2.6ghz Phenom is called a 9900. Where do you go from there naming wise? They have already said 9X50 is B3 stepping.

I'm also curious about that.

I wonder if they're going to pull another "HD2X00 - HD38X0" ish naming solution and use a new stepping to justify a new naming convention. Then again there is that possibility that they know they've screwed themselves out of the possibility to get to 2.8ghz...

 

[Evilonigiri]

What about that TDP on the Q6600? With the new G0 stepping, shouldn't the TDP be 95W instead of the 105W they wrote?

And why no Intel comparison? There was an Intel article on the QX9770 and they used an 6000+ for comparison...

 

[computertech82]

The article is an extremely POOR attempt to kiss butt for the crappy cpus. They UNDERCLOCKED a x2 6000+ to about a x2 5600+ cpu. Obviously tom doesn't want you to know every singled overclocked (and non-overclocked) x2 cpu over 2.6ghz STOMPS the ever living crap out of the Phenom cpus. Basically ZERO reason to upgrade, unless you really have a application that can actually use quad cpus (naturally, the OLD Q66600 intel cpu would the the BETTER choice).

 

[someguy7]

Just use the cpu charts
http://www23.tomshardware.com/cpu_2007.html?modelx=33&model1=1074&model2=915&chart=433

 

[Slobogob]

Comparing them clock for clock with only one core has its flaws, sure, but that wasn't the point of the article. If anything they should've tested how 2 K8 cores do compared to 2 K10 cores because that will lend some insight into the scaling provided by the architecture (and mostly the L3).
It was a good read and i think AMD isn't doing a bad job with the phenom if it weren't for the delays and bugs.

 

[Evilonigiri]

Right Phenom is actually pretty good...a year ago. Because of all the delays, it's still good however it's not that great.

 

[Shadow703793]

^Agreed