AMD REALLY needs to lower the X3's price. It shouldn't cost more than the current X2s.

Another AMD release greeted with a chorus of "Zzzzzzzzzzz..."

Dave

Something funny about <40 comments for a major product from AMD.

The cover image on the main page kindof puts me at edge, seeing those precious copper pins against eachother like that...

Wasn't there a lawsuit against Intel about when they switched from P3 to P4, and claimed better performance?

I hope the AMD marketing guys are keeping their dang mouths shut right now. ;)

Conclusions
……………
…………………
I'd go with the X4 9850 ten times out of ten. If, that is, I were somehow bound and determined to choose an AMD processor over one of Intel's current offerings.

Very Very sad statement. True but sad.

Please click the digg box on the first or last page, if you will, folks.

Also, please, erm, hose:

http://slashdot.org/firehose.pl?op=view&id=637856

Thanks!

I'm actually considering buying one of these, *only* because the wierdness of it. It's so... un-computer like. It upsets the order of things. The universe is falling apart! :-)

It might actually become a cult thing.

Call me a retard, but why are there four core windows open for a X3 review?

Poor AMD :(
These need to be dirt cheap just to keep the company alive. :(

A C2D outperforms a Phenom X3... in Cinebench. This is ridiculous.

Can I make a suggestion, Scott?

When looking at graphs such as on pages 5-8, I find myself trying to separate dual-cores from quad-cores (and triple-cores in this case). But this can only be done by reading the small labels besides each bar, which is difficult

What I'm suggesting is colours on the bars that differentiate between total numbers of cores. The "reviewed" bars could be a lighter shade (as now) or maybe some pattern fill.

This might create a garish colour scheme when we get to 6, 8 and 12 cores, but it would still help.

X4 9150e where are you.

I wonder if the third core would put these ahead of dual cores in a virtualization setting. I'd really like to see some tests run on multiple instances of virtualized XP or Vista.

AMD likely expected that these chips wouldn't win for performance. I think they're hedging on Joe Consumer to be fooled into thinking that 3 cores is better than 2, so the AMD-based system /must/ be faster. Sort of like the MHz race, only now it's the core race.

Speaking as an operating system kernel developer, there's nothing particularly special about powers of two when it comes to cores or logical processors. Each processor runs threads individually, so it's perfectly fine to have "weird" numbers of cores. Indeed, systems such as IBM's System p and System i are regularly sold with 12 quad-core processors with no ill effect. (12 is not a power of two, for the arithmetically challenged.)

The problem comes when software authors make assumptions that they don't really need to make. There's no good reason to assume that processors will always be available in quantities of powers of two, and that just smacks of bad design and programming practice.

The main issue with these Toliman processors is that their clock speeds aren't even up to Agena level. K10 very badly needs higher clock speeds since its instructions-per-clock aren't high enough to make it competitive with Core-microarchitecture processors.

I would've expected Toliman to be able to clock higher, since with a core disabled, the probability that transistors won't be able to switch reliably at a certain frequency goes down, but I suppose AMD might be binning Phenom X4s with a single faulty core as X3s, which would negate that possibility. Oh well.

Who would've imagined that AMD's latest would've been occasionally outperformed by their previous CPU architecture? Yikes.

Please don't ever use the phrase "core-botomy" again. It just sounds ....wrong.

Entertaining review as usual, but unpleasant reading for AMD. The only applications where the X3 distances itself from the duals, it's soundly beaten by the not-much-more-expensive quads. As you concluded, these would only have made sense if there was more breathing space between the top and bottom of AMD's price range.

Another failed AMD product.

This was an interesting decision on AMD's part. I was disappointed to see that triple-core didn't result in a lower advertised TDP (though it seemed to make a difference under load, at least) though.

I'm thinking the Phenom line will really start to shine with either a new spin that allows higher clock speeds or the transition to 45nm. Was hoping to pair one of these with a 780G board in my HTPC upgrade but looks like the Athlon X2 BE series is still best for that.

I've got a 6 CPU Pentium Pro server in my closet which has 2 daughterboards, each containing 3 CPU's.

I know what you're thinking - How can it have 6 CPU's when the PPro only supported up to 4 gluelessly? Unisys (made the box I think) made each group of 3 CPU's its own logical CPU for addressing purposes. Kinda slick when you think about it.

If memory serves me correctly, there was a tri socket Power Mac released a few years ago.

very interesting review, thanks scott!

one particularly interesting effect is the difference between intel and ms compilers in the lame test: the intel compiler significantly closes the gap between the b2 phenoms' performance with and without the tlb patch!

Firstly, we should get over this power of two thing. Power of two works great in theory but not so much in practice. If we do not get this into our heads we will force the industry to sell us nonsense. Harddisks for example are far from the power of two but still we get to see these nice numbers showing up in capacities of 160GB, 320GB and 640GB, when it is actually only billions of bytes and not 2^30 bytes. We can see the same with ISPs who sell us 1, 2, 4 and 8MBit connections and people actually might prefer to get a 16MBit connection over one with 20MBit. See lets embrace tri-cores.

The first thing thought that TechReport wants to do when benchmarking tri-cores is to adjust the CPU usage reading. In the report its 4 cores and I really would like to see the real usage for these CPUs as a lot of software still is not able to fully utilize neither 2 nor 4 cores.

Sven

AMD is perfectly fine trying to sell a processor with three cores...

...except in this universe, computers revolve around powers of two. Sure, there's nothing wrong with it in theory, but look what happened with the encoder. I hate to see AMD take yet another hit, but I think what they're doing (twelve-core Opterons coming up?) is going to encourage software designers to totally rethink how they deal with multithreading, and that's good news for us. (I'm sittin' here on a Q6600, how lame am I. :-p)

Bottom of page seven.
Shame.

I understand the software was using 2 cores, but it really boggles the mind how 2 cores of the X3 8750 are unable to even approach 2 cores of ye olde X2 5600+.

Didn't ya'll do a test awhile back with an opteron 275 and a 260 on a board together for 3 effective cores (or whatever the right model numbers were)?

Good review though. Shame the X3 doesn't offer anything over the X2 for all intents and purposes.