A First Look at AMD's Triple Core Phenom

[epsilon84]

I don't know if anyone actually realises this, but many of the tri core benchmark results are lower than that of C2D or even A64 X2.

Compare http://www.tomshardware.com/2008/02/28/a_first_look_at_amd_triple_core_phenom/page6.html to http://www.tomshardware.com/2008/02/13/wolfdale_shrinks_transistors/page7.html for example.

No surprise then that C2D and X2 numbers are not available for comparison. 😛

To be honest, I'm not quite sure what market tri-cores will appeal to. In most cases, it doesn't provide noticeably better multithreaded performance than higher clocked dual cores, and is slower in single/dual threaded software.

The price gap between dual and quads is already very tight, it doesn't seem like there is much room for tri core pricing. It will be interesting to see how it all pans out, but I'd be very surprised if it actually ends up a hit on enthusiasts sites such as THG.

 

[Loser777]

Well, no real big surprises here. Kind of an unnecessary article if you ask me. I'm just interested in the B3s and the 45nm ones. These benchmarks don't real much now since these B2 cores will be tossed quickly.

 

[yipsl]

Here are some bigger questions IMO:
1) Cache. Will the Tri core have access to the same cache as the Quad? Thus in applications which require a large cache, will the increased cache hit rate make up for the loss of a core?

2) Overclocking! Let's face it if you don't overclock or are on a tight budget, AMD is likely your choice. Yeah it gives up some 200-600MHz to the C2D but for a given amount of cash, AMD has a solid position. Once you begin overclocking this all changes. The C2D simply blows AMD away (Or if you have several hundred to spend on the CPU). For example, I bought X2 5600 w/ 1Meg cache. $100 bucks delivered back before the abundance of C2D options. OC'd it to 2.95G. 690 chipset, AM2 socket. Current quads are a bad upgrade choice. Well this tri may be perfect! One less core, less heat, better OCing, yes yes?
*note* I did read the article, heat production was still up indicating that the core was likely running to some degree. However you didn't even attempt OCing!

Anyway, Tom's is no longer my first stop for "solid" reviews because these reviews simply don't answer the important questions.

Simply Disappointed.

The L3 cache question is a good one, but you miss the point of the article regarding overclocking and heat. This is not a B3 triple core Phenom 8000. It is a quad core with one core disabled in bios to simulate scaling between cores. When we see actual retail B3 cores, both quad and triple, then we'll see whether they overclock better. I don't expect Phenom's to overclock until 45nm Shanghai on AM3 this time next year.

As for Phenom B3 triple vs. C2D, that's not the point of the article and won't be addressed until the CPU releases. Few buying an enthusiast system will go for Phenom quads or triples. It's a good upgrade choice for an existing AM2 board that has an available bios. It's a good upgrade choice for someone wanting to position themselves for 45nm Phenom on a 790 Crossfire board.

The heat issues in the article merely show that heat doesn't change with one core disabled in the bios. We'll see how thermals go with genuine triple cores. Like Tom's Phenom core vs. Athlon X2 core, the article addresses real issues of how the various flavors of AMD CPU's relate to each other. It's time we had more than just the "OMG, Phenom suxs! It's slower than Q6600" sort of posts based on benchmarks that are only about overclockers and the enthusiast's who overclock.

Just as someone building an Intel system needs information to decide between various Intel flavors, so too do those building or upgrading AMD systems. That's what the Tom's Phenom articles have been about.

 

[Shadow703793]

^Agreed.

 

[Mathos]

I don't know if anyone actually realises this, but many of the tri core benchmark results are lower than that of C2D or even A64 X2.

Compare http://www.tomshardware.com/2008/02/28/a_first_look_at_amd_triple_core_phenom/page6.html to http://www.tomshardware.com/2008/02/13/wolfdale_shrinks_transistors/page7.html for example.

No surprise then that C2D and X2 numbers are not available for comparison. 😛

To be honest, I'm not quite sure what market tri-cores will appeal to. In most cases, it doesn't provide noticeably better multithreaded performance than higher clocked dual cores, and is slower in single/dual threaded software.

The price gap between dual and quads is already very tight, it doesn't seem like there is much room for tri core pricing. It will be interesting to see how it all pans out, but I'd be very surprised if it actually ends up a hit on enthusiasts sites such as THG.

Yeah this is true, I'd like to see a C2D running at around 2.6ghz and an athlon X2 running at 2.6ghz as well just for comparison. Should beat X2, but c2d/Penryn would still probably win, except maybe on the thread optimized apps. None of this stuff pitting a 3.2ghz cpu against a 2.6ghz cpu.

 

[thefumigator]

Some of this is true. But Asus usually will update if they said it will support a CPU. My mobo is a P5K-E with the P35 chipset and it stated on the box it will support Intels 45nm CPUs. And so far the last BIOS added support for every Wolfdale dual core and every Yorkfeild quad core.

They could have not added it and forced us to buy a X38/X48 mobo but Asus is one of the better companies out there. Hell my old P4P800 supports every PGA478 Pentium 4 CPU out there. It was the LGA775's when BIOS updates stopped.

Now to compare AM2 vs AM2+ isn't that hard even with thought. The HT 3.0 vs HT 2.0 will make the major difference in memory sensitive applications as the HT link is mainly a memory bandwidth booster hence why HT gets better memory bandwidth than FSB does.

I doubt the differemce will be big enough in gaming and other big user apps to really upgrade to a AM2+. But once apps and games start to take advantage of the extra memory bandwidth it might prove to be worthwhile. But I don't expect that to happen really until the time of Nehalem and AM3.

But yea I would love to see a comparison of the two sockets just to see what the difference is.

I agree that asus is one of the greatest, actually its number 1 to me, followed by Gigabyte.
We both agree then, that "we were disappointed to see that Phenom support was virtually nonexistent" statement is extremist, as there were many mobos for testing it. Even checking asus site there are many very interesting ones, but I really don't know how popular they are... and you mention popularity. Still support doesn't seem "virtually nonexistant" to me.

I'm still having doubts about performance impact. To me, just a guess, could be noticed in higher frequency phenoms, when they finally arrive. But I'm pretty sure that today "AM2 classic" has very little (to zero) performance impact on actual available phenoms...

But lets say the truth, the AM2 platform rocks, what may suck if anyone wants to know, may be the CPU. I still think that the platform itself hasn't been limiting performance with actual AM2 cpus.
But yeah I agree its always good to improve and thats AM2+ all about, don't know if performance wise but features are cool. I wish there was a really fast CPU (I really mean fast) to put AM2+ on its pants, as it shouts for a challenge...

 

[epsilon84]

Yeah this is true, I'd like to see a C2D running at around 2.6ghz and an athlon X2 running at 2.6ghz as well just for comparison. Should beat X2, but c2d/Penryn would still probably win, except maybe on the thread optimized apps. None of this stuff pitting a 3.2ghz cpu against a 2.6ghz cpu.

I actually don't mind non clock for clock comparisons as long as the products being compared are at equivalent price levels. I don't see why a $150 tri core should not be compared to a $150 dual core.

 

[Mathos]

I actually don't mind non clock for clock comparisons as long as the products being compared are at equivalent price levels. I don't see why a $150 tri core should not be compared to a $150 dual core.

Well, I do mind some times, since you end up with a bunch of dimwits going the old X2 outperforms the tri core, not taking into account the clock difference. Granted we won't know how high they'll clock till they come out, and whether or not they'll be off of the b2 or b3 revision. Though, there is the chance that with a core disabled even b2 may be able to push 3ghz on some processors. Just need them to up the IMC to the core speed, or make it be able to be set within 200Mhz like it's suppose to be.

 

[epsilon84]

Well, I do mind some times, since you end up with a bunch of dimwits going the old X2 outperforms the tri core, not taking into account the clock difference. Granted we won't know how high they'll clock till they come out, and whether or not they'll be off of the b2 or b3 revision. Though, there is the chance that with a core disabled even b2 may be able to push 3ghz on some processors. Just need them to up the IMC to the core speed, or make it be able to be set within 200Mhz like it's suppose to be.

So what? Just because a higher clocked X2 outperforms a lower clocked Phenom tri core (or even quad core) does that make it any less true? It's clocked higher because IT CAN BE. Do you think AMD would not clock their Phenom at 3GHz if they could? Of course they would! But the reality is that they can't, not yet anyway.

Performance per $ is the most important. If an equivalently priced dual core beats a tri core, then thats what matters. I don't need to underclock the dual core for some useless 'clock for clock' comparison to make the tri core 'look' better. It doesn't take a genius to figure out that 3x K10 2.6GHz cores would be faster than 2x K8 2.6GHz cores, it's just a pointless comparison.

 

[Mathos]

Well, the problem would be that upping the IMC does improve the performance of the cores, giving it more bang for the buck like it should. Thats why I said, it may be possible to have the tri cores clocked higher as well.

 

[epsilon84]

Well, the problem would be that upping the IMC does improve the performance of the cores, giving it more bang for the buck like it should. Thats why I said, it may be possible to have the tri cores clocked higher as well.

This is all just speculation on your part. All reports I've seen indicates tri-core will initially be clocked up to 2.5GHz only.

If nothing else, AMD wouldn't want their tri-core 'leftovers' outperforming their quads. It wouldn't make for good PR.

 

[Mathos]

This is all just speculation on your part. All reports I've seen indicates tri-core will initially be clocked up to 2.5GHz only.

If nothing else, AMD wouldn't want their tri-core 'leftovers' outperforming their quads. It wouldn't make for good PR.

Well, I can prove the IMC clock thing, but noone will be able to prove what speed the tri's will launch at till they launch.

 

[yipsl]

Well, I can prove the IMC clock thing, but noone will be able to prove what speed the tri's will launch at till they launch.

I got flamed for speculating that the triple cores will be clocked higher than the quad cores. It turns out that the B3 quads are expected to go to 2.6, but the triple cores only to 2.5. That's still higher than the actual clock of released Phenom's at the time I made the statements.

What I'd hoped for was a bad core on some Phenoms that prevented stock clocks higher than 2.6, such that when it was disabled, we might see 3.0 or 3.2 Phenom's. That's not to be with the B3's, but might be true with the 45nm.

If 45nm Phenom releases at 3.2, it will be up against Nehalem at 4.0 and above (if reports pan out). AMD's still back in the K62 days as far as market share goes, except this time around, it's the OEM's and notebooks that make them profitable. The closest thing they have to an enthusiast part is from ATI, the 3870x2.

So, should I go Wolfie 3.0 after all on a Crossfire board? Is there a board that will be Nehalem compatible in truth and not in FUD such that I'd not have to buy a new motherboard when I do go Nehalem? IMHO, the 3870x2's and 4870x2's on 24" LCD's at 1920 x 1200 really need the fastest CPU possible, and that's not AMD right now.

 

[sunangel]

I keep asking myself if most other computer peripherals have on-board hardware encoding/decoding why would I need a processor greater than 1GHz. 1GHz is more than enough to process the little extra data not processed by the hardware based peripherals. For that fact, I settled on the Celeron 420 1.6GHz. 2 Hauppage tv cards, ATI HD 2600XT, 2 7200 RPM Seagate in RAID 0, 2GB DDR2-800, and of course Vista Ultimate. My computer runs smooth as silk. I can record two tv shows at the same time in media center, watch a dvd, and surf web all at the same time. And the 420 is a single core processor. I do not get any hiccups. The celeron 420 is one bad mofo. These tri-cores ... AMD has to be marketing these towards lame enthusiasts because if the difference between a 3 core processor and a 4 core processor is only like $5-10 bucks it easy for a the average joe/jane going to tigeregg to figure out which to buy.

 

[apache_lives]

Although it will not be available before the second quarter of 2008, we decided to take an early look at the performance of AMD's upcoming triple core Phenom 8000 series. Read on for a comparison of Phenom performance at one, two, three and four cores.

http://www.tomshardware.com/2008/02/28/a_first_look_at_amd_triple_core_phenom/

Not one, not two, BUT 3 CRAPPY CORES!!!!!! 1 MORE THEN CORE'S A 2 (duo)

What next, a core 2 Trio, or a threesome, wait th...never mind.

As for the celeron 420 - its equal or better then a A64 3000+ or P4 3ghz so expect good things, and super efficent (even the E2140 does ~35w) - i wonder how easy it would run with a passive cooler...

 

[sunangel]

As for the celeron 420 - its equal or better then a A64 3000+ or P4 3ghz so expect good things, and super efficent (even the E2140 does ~35w) - i wonder how easy it would run with a passive cooler...

Very doable. I removed the fan from my Arctic Cooling Freezer 7 Pro and my temps are 34 degree celcius UNDER threshold. I do not overclock and this is inside an Antec Nine Hundred computer case with all fans set to low speed. I didn't really have to remove it because the case is already quiet with the fans on low speed ... removing the cpu fan didn't change the noise level. The only thing that this chip doesn't do well is real-time transcoding ... well it does do real-time transcoding but typically anything with a resolution greater than 320x240 doesn't playback very well. If I rip a dvd and transcode it it takes about 3 hours for a 100 minute movie. Coupled with the ATI HD 2600XT I have FULL HD 1080p glory on a Sony Bravia KDL-46XBR2. I can play Microsoft Chess Titans and Hold'em at full resolution.

 

[jimmysmitty]

Not one, not two, BUT 3 CRAPPY CORES!!!!!! 1 MORE THEN CORE'S A 2 (duo)

What next, a core 2 Trio, or a threesome, wait th...never mind.

As for the celeron 420 - its equal or better then a A64 3000+ or P4 3ghz so expect good things, and super efficent (even the E2140 does ~35w) - i wonder how easy it would run with a passive cooler...

No but Intel will have a six core, that would be a sext core maybe. That will probably push the quad cores down a bit so they will be in the tri core price range. And if people think more cores = better and a quad core cost the same tri cores will be useless.

That would be interesting.

 

[yipsl]

No but Intel will have a six core, that would be a sext core maybe. That will probably push the quad cores down a bit so they will be in the tri core price range. And if people think more cores = better and a quad core cost the same tri cores will be useless.

That would be interesting.

That's probably why Intel's doing this. Personally, I think that the Decline of Civilization can be tracked by the rise of marketing since the '50's. Too bad we can't take Douglas Adam's advice and send all the marketing departments, hedge fund managers, golden parachute CEO's and public relations reps off to colonize another solar system and leave scientists, engineers, artists, writers and theologian/philosophers to advance civilization on all levels.

 

[coret]

That's probably why Intel's doing this. Personally, I think that the Decline of Civilization can be tracked by the rise of marketing since the '50's. Too bad we can't take Douglas Adam's advice and send all the marketing departments, hedge fund managers, golden parachute CEO's and public relations reps off to colonize another solar system and leave scientists, engineers, artists, writers and theologian/philosophers to advance civilization on all levels.

Sure we could ... but we'd need the best marketing person in the world to help us sell the idea to move to the rest of his/her kind =p

 

[epsilon84]

No but Intel will have a six core, that would be a sext core maybe. That will probably push the quad cores down a bit so they will be in the tri core price range. And if people think more cores = better and a quad core cost the same tri cores will be useless.

That would be interesting.

No, it would have absolutely no affect on quads, since Dunnington is a server/workstation CPU, not a desktop CPU. It's designed to run on the Tigerton platform, which is multi-CPU. So you can effectively get 24 cores (4 sockets x 6 cores) but that has nothing to do with desktop pricing at all. 😉

 

[dengland88]

Anything new in May 08?

I am looking for a starter CPU for a 780g board. I would expect to upgrade the CPU in 12-18 months.

8x50 triple cores are now shipping with the low end 8450 (B3) for about $150. That price is about the same as a 5000+ BE with name brand heatsink and fan.

Most of my activities are web browsing and email. Occassionally, I expect to be transcoding video. I have an Opteron 165 in 939 board right now slightly overclocked.

What is the conventional wisdom right now?

Thanks