Is AMD's triple-core Phenom no more than marketing hyperbole?
Summary: Today I'm coming across a fair bit of coverage of AMD's triple-core Phenom processor - but is it anything more than marketing hyperbole?
Today I'm coming across a fair bit of coverage of AMD's triple-core Phenom processor - but is it anything more than marketing hyperbole?
What we know about the triple-core Phenom (codenamed "Toliman") is sketchy - it has three cores and will be out in Q1 2008. That's it. No pricing information. No clock speeds. What there is spade loads of is marketing hype. However, when reading through AMD's info on the triple-core Phenom I did come across this interesting statement:
In addition, triple-core processors from AMD can provide significant performance advantages over similar dual-core AMD processors in key industry standard benchmarks, including SYSmark® 2007 and 3DMark™ 2006, as well as similar quad-core AMD processors in certain gaming and digital content creation scenarios.
This is easy to rationalize - AMD's current desktop processor lineup suck. If Toliman couldn't deliver a performance advantage over AMD's existing lineup then what would be the point of releasing it. What really matters is how it compares to Intel's lineup, and not Intel's current range, but what they have out at the time Toliman is launched.
Call me cynical, but I can't help but feel that all AMD is trying to do here is fragment the processor market and cause consumer confusion. Most people buying PCs aren't going to know (or care) about benchmarks, they're going to assume that three cores is better than two and buy accordingly.
Maybe AMD is onto something here though. After all, has H.L. Mencken once said "No one in this world, so far as I know — and I have researched the records for years, and employed agents to help me — has ever lost money by underestimating the intelligence of the great masses of the plain people." Me, I think I'll wait until I see some real benchmarks.
Thoughts?
Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.
Talkback
First, AMD may not perform with...
Second, what software out there today, or tomorrow, will be able to take advantage of 3 cores? Or 4 cores for that matter.
I agree
Why multi-core works even without special software
So unless ALL you do is gaming, for many of us, who are doing day to day work who may have lots of processes running (and doing something) it makes a lot of sense. I'm probably a bit atypical but not so much that my argument doesn't apply to the masses.
Multi threading is also becoming
Let's See ....
Probably just quad cores with defects in 1 core
Sure, a good way to get rid of defects, but, they might also manufacture 3
silly amd fanboy
Let me get this right, you are arguing that three cores are not enough, and
All I can say is that the sweet spot for mult-core is probably around 2-3 right now and going up all the time. But, there will still be plenty of room at the low end for 2 and 3 cores 2 years from now.
Funny to see the weird arguments pop up here!!
Look at it from the other side.
Turned-off defective core
Agreed. I recall that back in the day of the 386 and 486 processors, Intel managed to salvage many "defective" dies by snipping the connector pin for the FP portion of the processor (that had failed a heat/temperature test) and labeling them as "386sx" or "486sx". The "good" processors received a "dx" label. And a higher price.
And then, Intel spun and marketed the hell out of them, eventually creating an entire line of [b]intentionally[/b] brain-damaged processors for a market they (Intel) created. And now, history repeats itself.
I will say, though, that Intel never claimed that the "sx" processors could "provide significant performance advantages", since they knew damned well that it wasn't true. I'm willing to suspend judgment on AMD's claim until it's actually tested.
These will perform better than 2 core but not as good as 4 core. Pretty
Crippled processors
"I wonder if AMD will intentionally cripple the chips as Intel did though."
Did you actually bother to read what I wrote? Initially, the Intel 386 and 486 "sx" processors were not "intentionally" crippled. The floating point (FP) section of the processors failed a simple heat test. In order to salvage the expensive "failed" processors (which were otherwise completely functional), Intel removed the pin that fed the FP portion of the processor, and charged a lower price for it.
It wasn't until later, after the cheaper "sx" processors became wildly popular (due to the price difference between themselves and the more expensive "dx" processors) that Intel intentionally snipped the FP connection, whether the chip was "defective" or not. There was no need to build a separate production line for the "lesser" chips.
In fact, if you decided that you needed floating point processing power, you could purchase a separate FP chip--387 or 487, which lived in a separate socket on most system boards produced at that time.
How difficult is this for you to understand?
"With the flexibility of their architecture, it should not be to hard to lay out a 3 core part and manufacture it. It would be a smaller die. Of course they could also sell the 4 core chips with one bad core. Heck, they could conceivably come out with a 5 core."
How long have you been designing microprocessors?
If the architecture of the AMD Phenom is so "flexible", then AMD would have designed it with three cores on a smaller die in the first place. Right? At least, that's according to [b]your[/b] logic.
The three-core Phenom [b]is[/b] (or will be, once it's released) a four-core die that has one "defective" (for whatever reason) core.
Sheesh. Kids. Ya can't teach 'em anything these days...
Ummm...
That was early on. The whole SX story is very complicated and
re: Ummm...
I find that I must make corrections to some of my previous statements. Mea culpa due to brain bit decay. It's only been 15 or so years. :)
The i386sx processor was a low-cost version of the original i386. Internally, it was a 32-bit processor but had a 16-bit external data bus and could only address 16Mb of RAM. This was done intentionally by Intel. The i386 was later renamed i386dx to "avoid confusion".
Source:
http://en.wikipedia.org/wiki/Intel_80386#i386SX
The i486sx processor was originally a complete i486 processor with a defective math co-processor element that had been disabled. In later production, Intel intentionally removed the math co-processor element from the die to save space and cost.
The optional i487 math co-processor, when installed, would disable the i486sx main processor.
Source:
http://en.wikipedia.org/wiki/Intel_80486#Models
Thanks ^_^
Yes, we understand very well that Intel did not intentionally cripple the
And, there is no reason that AMD could not manufacture BOTH 3 and 4 core chips. With the 3 core, they would get more chips from each wafer, and thus it would cost less. I kind of doubt it would be cheaper to cripple a perfectly good 4 core chip.
As another poster noted though, if one of the cores goes south at a low clock speed, it might be best to call it a 3 core and raise the frequency.
Understanding
Is that the editorial or imperial "We"? ;)
It's possible that AMD might remove a core from the Phenom, but I think that's unlikely and do not see what purpose it would serve.
At the 45nm level, it's unlikely that AMD would be able to squeeze out very many more 3-core dies to make up for the costs of the redesign to produce them. It might, however, help to improve stability and reduce the amount of heat generated in the die itself.
In my opinion, AMD's decision to issue a 3-core Phenom model is mainly to reduce the cost of wastage in regards to Phenom dies which have one "defective" (for whatever reason) core, so long as the rest of the die can be used in production processors. After all, reducing wastage will tend to shore up the bottom line somewhat.
"As another poster noted though, if one of the cores goes south at a low clock speed, it might be best to call it a 3 core and raise the frequency."
I would be rather leery of using a "factory overclocked" processor which already has one bad core.
about the 486SX
I understand that the "math co-processor" for the 486SX was, in fact, a complete working 486DX CPU. When installed, the 486SX was disabled.
clever marketing huh?