Slashdot Mirror
AnandTech Peeks At The Athlon 4
tupac writes: "AnandTech has posted a story about AMD's new Athlon 4 processor and also included some information about Silicon on Insulator transistors in their article. SOI technology has been used by IBM in recent history and AMD will begin using it in 2002." This is the chip which has been known for a while as Palomino. Reader Diabolus points to the same article, saying "the big news is a 20% reduction in power consumption, and that they'll be using the exact same chip for servers, workstations, desktops and even notebooks. The article details exactly what is new compared with the Tbird."
bull... the athlon is the a 7th generation chip from the x86 instruction set; it was released 2 years ago and beat the crap out of its competition. intel on the other hand just recently released it's 7th generation chip, the p4, and it got beat by it's own predecessor. in terms of developing what's new, intel has lumbered away on ia64 with nothing to show for it but some itanium simulation software. amd is preparing x86-64 and according to the roadmap will be out by next year. to say that amd is putting out a second rate product is a load of crap; amd chips regularly beat intel's, and don't have nearly as many production woes.
You cannot decouple "efficiency" and clock speed like you are. The processor you are talking about, the K5, as well as many of the cyrix designs were more "efficient". They got a lot more work done than the pentium or pentium pro, possibly even the athlon in a single clock cycle. The trade off is that the very efficiency you long for is what holds back the clock speed. By sacrificing the work done per clock cycle they can ramp up the MHZ speed. If you look back to the K5, I think it only went up to a 133 MHZ speed. What makes it difficult for non uP engineers to understand (I'm one too for sure, just a laymen who likes to read) is that design and the ability to manufacture that design are inherently linked. You can't just say, if only they could manufacture that design faster and better. The design influences how easy it is to manufacture it at higher clock speeds.
It's pretty well known that the TBird core was thermal limited, not architecturally limited in speed. It looks like the TBird will top out at 1.4GHz. The Pally core supposedly uses 20% less power, so we can assume it runs about 20% cooler. This means they should be able to achieve 1.7GHz or so out of it, which is pretty competitive with Intel's PIV.
.13 process and should be able to hit 2GHz+
Then they'll move to the
Let's just hope they don't recall it like Intel did :-)
___
___
If you think big enough, you'll never have to do it.
That was a good idea, actually.
I just hope AMD doesn't also adopt Intel's shady power consumption reporting as well.
The enemies of Democracy are
I've a PDP11/70 still running, I am sure we could work out a deal. RSX BABY, the way to go...
Vax support is going to evaporate really quickly here too.
errr....umm...*whooosh* *whoosh* Is this thing on ?
I am behind you 100% Q2 was awesome, UT is just plain FUN but Q3 is STOOPID. Is like playing a silly cartoon. Not to impune skills required for either game but I just can't concentrate on Q3.
Even Tribes2 with its plain textures and graphics, PLAYS better than Q3.
errr....umm...*whooosh* *whoosh* Is this thing on ?
3Ghz of cpu power:
$1500
2GB of DDR RAM:
$900
Watching others' faces when you compile the kernel in 0.06 seconds:
Priceless
--
"You never know when some crazed rodent with cold feet
might be running loose in your pants."
"You never know when some crazed rodent with cold feet might be running loose in your pants."
-Calvin
The article notes that due to the size of both the processor and the potential chipsets it will be impossible to use the Athlon 4 in the "thin to light notebooks" category; instead, the target would appear to be notebooks in the "desktop replacement" category. As the area of increasing demand is in the "thin to light" segment and as prices are being cut for previous laptop models to clear out inventory, I am doubtful this initial launch of notebook Athlon 4's will have much impact.
Actually for each new CPU generation the voltage gets lower since that alows the chip run faster.
...than having my new expensive CPU chip turn into a useless piece of silicon oxides when it burns.
The question is rather when and how often does it happen.
And I wonder which corporations that make laptops (IBM, Dell, etc) will make a product line with the new AMD cpu?
If you bothered reading the article qou woud find this passage.
Compaq has already announced and will be shipping notebooks based on the mobile Athlon 4 processor.
These should already be in stores by publication. While no other manufacturers have announced yet we can expect notebooks from HP and probably Sony as well.
So true lower voltage in and of itself limits max speed. But, as I said lower voltage is a requirement for higher speed due to heat issues. The speed increase is created through many design elements.
Moderators seem to be erratic at best and maybe you were modded down because you'r AC...
Increase in speed will use up more power unless you do changes to your chip (and other cicuits). This is important since there is a finite amount of heat that can be absorbed and dissipated from a chip. The reason for the power consumption is that there is a certain current needed to carge and discharge the capacitance of a CMOS. The heat comes from the fact that Power = Voltage(drop) * Current, so to increase the frequency you need to do one or more of the following.
Usually all of this is done to each new CPU. Lowering the voltage does have a negative impact on performance but it is still required to enable higher frequencies due to heat factors. Actually some of the lower voltage performance hit is offset by other improvements (i.e. less power loss in interconnects)
"What Intel definitely understands is that the public thinks Mhz == speed."
Intel understands it because Intel caused it. Intel's adverstising over the past 2+ years has focused on two things. First, they make MHz seem equivalent to horsepower. Second, they make vague claims that Intel processors make "the internet" faster. AMD played Intel's MHz game for a while and did a damn good job beating the big boys at their own game. Now they're not playing anymore. They seem to be focusing now on creating good processors at a reasonable price.
-B
I don't think it's true that people are buying Intel's MHz = performance ploy, or the whoel GHZ PC hype for that matter.
The evidence is:
a) P4 is hardly selling at all
b) The whole PC market is stagnent. Most people have 2/3/4/500MHz or whatever PCs that do all they need, and see no need buy a new GHz PC
Really, outside of DVD ripping, there are few reasons for a *typical* person to need this much speed - certainly not enough reason for them to want to shell out $1000 or more for a new PC.
Please. Intel at least add instructions between...
If you add instructions you should do this for one reason only: to use 'm. This is the biggest problem right now. Compiler builders have a hard time optimizing your code and keeping up to date with the latest instructions.
These days compilers are the word, not uP's. You can make the nicest microproc but without a good compiler you are nowhere... Everyone can think of a few more instructions that are 'handy'.
nosig today
Just because they use the same socket format doesn't necessarily mean all motherboards will work with it. There's chipset compatibility to deal with, as well as voltage and BIOS support. So we may need a new motherboard anyways, even if they use the same socket.
SYSOP ('sih-sop) n.: the guy laughing at your typing.
That's a function of the Chipset, isn't? That's not something which requires explicit support in the CPU itself, as far as I know. Wait a bit, and I'm sure Via will come out with something.
And not to me, since they both agreed with my incredulous response to the parent author's statement that there is no DDR support.
Quoted directly from the article:
"This process obviously increases FSB and memory bandwidth usage and it does tend to show more of a performance improvement on higher clocked/higher bandwidth FSB/memory platforms. This does translate into DDR SDRAM being much more useful for the Athlon 4 than it was on the Athlon (Thunderbird)."
Johan V.
Chip Maker Trustworthy announced today they'll be cutting to the chase and releasing a 5terrahertz chip which is the fastest on any market, touting a catchy slogan "Powered by God."
"We didn't want to get involved in the whole marketing game at all. We've had these chips for years but had to sell all other lower speed chips in order to make money. Well all that is in the past, from now on we will release things to the public immediately. No more lies." stated Swedish born CEO Karl Karlssson who is now a converted Born Again Christian Chipmaker.
The company however faces a lengthy delay after residents of South Beach Miami claimed to own the patents to create the sand used to make the silicon used to make the chips. "What about dew prossis" claimed a big breasted South Beach Miami blonde wearing an Intel t-shirt looking in the mirror.
"We decided to give the chips to the people, without any side stepping. As is, and we're confident the powers of the Pope will annoint this chip and smoat the "Blue Man Group" of Intel who resemble satanists." stated Karl.
Stay tuned for this lengthy battle.
Q & A with John Young of Cryptome.org
Want Root?
Where on earth did you learn how to write? Anyway, I "think" you mean that marketing will triumph. If so, I tend to agree. I was reading a ZDnet article the other day about standards. They mentioned a number of times how Windows is a standard, simply because it's everywhere, even though there's no official "desktop standard". We gotta long fight ahead of us!
Dive Gear
--- Think of it as evolution in action ---
20% reduction on the chip powerconsumption, does that give me a lot to think about when using a 350 Watt power supply? I know powerreduction on the cpu is interesting to know when looking at the heat production/cooling and stuff but Kilowatt's are getting expensive in Europe. Just curious.
Courtesy of ViaHardware, comes this link: http://www.viatech.com/products/KN133.htm.
The chipset is called KN133 (basically a mobile KT133), and offers PowerNow support as well as integrated Savage4 graphics, but no DDR-DRAM. DDR-DRAM is offered only by the mobile ALi MaGiK1 chipset, which so far seems to offer underwhelming performance, and has not been included in any of the current release of Athlon 4-powered notebooks.
Get off my virtual lawn, you damned virtual kids!
Cache improvements, thermal diode on the CPU, 3 - 3.5 hours claimed battery life, significant performance gains over P III when both systems run on battery power, overall a performance gain over the current desktop Thunderbird.
In a nutshell.
No DDR, which if it were implemented would actually help laptops save more power.
Get off my virtual lawn, you damned virtual kids!
If it's not that hard, how come you follow up somebody who got it right, with something wrong?
The original TBird and the "Athlon C" are basically the same core, with a minor revision to eek out a few hundred more megahertz of speed.
1) K7: Original Slot A Athlon at 0.25 microns
2) K75: Slot A Athlon at 0.18 microns
3) Thunderbird. AKA "Athlon with Performance Enhancing Cache". Socket A, although some Slot A's were made. This may be a faster cache, but it is also smaller than the 512K external cache that the first two Athlons had. The original version of Thunderbird barely made it to 1.2GHz. The "C" revision looks like it'll go to about 1.5GHz.
4) Palomino, or "Athlon 4".
I hope they use a different name for the desktop version. It sounds good as a mobile version, because Intel doesn't have a Pentium 4 mobile, but I think AMD will want to make sure that people make fundamental comparisons on the desktop rather than superficial.
Bryan
Should have called it the athlon 5
The current Slashdot moderation system is made by gay communists!
Contrast this to Intel, who is so bent on shoving stuff down our throats that they willingly sell products that have a short or no real life span expectancy. (p462 anyone?)
Sorry to nitpick, but Socket 462 is the same as Socket A, for AMD Athlon/Duron.
The current P4 socket is Socket 423, whereas P4 Northwood will require Socket 478.
/Dervak
Ace's Hardware has a nice summary and set of links for the Athlon 4.
Unfortunately Sharkyextreme and HardOCP do not have reviews of the chip up for comparison yet.
Tom's does have a review up.
The Bi Athlon (in addition to being a familiar, if stupid, Olympic event, this name has sexual connotations and should slightly benefit from men who want a faster proc for pr0n)
The Athlone Ranger (OK, this one's a bit of a stretch)
Athlon as You're Happy, That's All that Matters (may appeal to moms)
--
It's rare that you're presented with a knob whose only two positions are Make History and Flee Your Glorious Destiny.
Sexium being after Pentium doesn't even make logical sense
I'm aware of that... but you're trying to tell me it makes significantly less sense than "Pentium?" And that anything that comes out of the Intel marketing department makes logical sense? AND that Sexium isn't just a significantly funnier/better name than Hexium?
:-)
sig fault
...was created so that Intel could have an enforcable trademark on their chip name (unlike 386 or 486).
:-) Time to roll out the Sexium.
So now, because marketing got so attached to the name "Pentium," AMD can again match version numbers in their product names.
That's hilarious. Heads are gonna roll.
sig fault
Looks good.
Well, to answer the gentleman who said "Athlon 4" was a "stupid marketing ploy", consider - The readership of Slashdot, and persons who assemble their own PC's, period.. Are a far, far smaller percentage of the computer-buying public than the cluebies who buy PC's at Best Buy as though they were $1,000 toaster ovens. Branding is important; It's how THOSE people will remember "Athlon." YOU remember specs, performance - Technical information. THOSE people need a counter to Blue Man Group's P4 commercials. Like.. Stomp hopping around on trash can lids, singing the praise of Athlon Instead.
THIS SPACE INTENTIONALLY LEFT BLANK.
From the article:
That notebook you saw on QVC is real. It is the desktop model that will not ship until August.
Now that AMD's new cpu core (Palomino) will run over 20% cooler than the Athlon, I wonder how long it will take for this chip to actually APPEAR in a laptop?
And I wonder which corporations that make laptops (IBM, Dell, etc) will make a product line with the new AMD cpu?
I would love to see some "uptime" tests between the P4 and the Athlon, with temperatures and all.
[Connection closed by foreign host]
And they added "3DNow! Professional" - 3D Now! plus Intel's SSE.
Wu-Tang Name: Half-Cut Skeleton Get your own Wu-Na
Dancing is illegal.
They sweep Loch Ness for the monster.
Douglas Adams dies.
Hmm..wait a tick. All those things already happened. ; Just might be the reason that my boxes are all running on AMD.
~~ What's stopping you?
Always a bonus... I've still got my PPro here, which is running essentially the same core as todays P3s. In fact, if I were a sick individual, I could get either an adaptor that would let my place this chip in a modern slot1 board, or an adaptor to put a s370 celeron in its place (P3s might work, but PPro boards only go up to 66MHz FSB).
my sig's at the bottom of the page.
IBM's pages have a lot of info on SOI. Here's a few links:
Press release,
Introduction - the pdf linked to from this page is a little more indepth,
Appears to be the start of an html version of the pdf.
All these pages of blah-blah and they could not cut to the chase and do a single review/benchmark of Q3A??
Gimme a break!
2)Socket A Thunderbird with larger/faster cache.
3)"Athlon C". Thunderbird core with 266 MHz bus.
4) Upcoming Palomino core discussed in the article.
It's not that hard, people
-atrowe: Card-carrying Mensa member. I have no toleranse for stupidity.
Wouldnt it be better if Intel focused on improving performance without adding more instructions all the time. If they hadnt added more instructions to the P4 (for example) it would not have performed so badly in benchmarks (that are based on old code).
;)
In the end, most applications and OSes will still be built on 386 or Pentium compatible code (with the possible exception of Multimedia applications - which by the way are the only ones that requires performance
That was a really well written article. *pleasently suprised* I'm not actually too hot on Hardware but I understood most of that article. The last few pages detailing Athlon's roadmap was quite interesting.
/. community.
I've also been thinking, You know you've hit the big time when any and every performance test for hardware uses your software.
Some interesting numbers to quote from AMD include a 6% advantage under Quake III Arena...
Anyway just thought I'd pass this useless thought on to the
Pinky: "What are we going to do tomorrow night Brain?"
Pinky: "What are we going to do tomorrow night Brain?"
Brain: "I would tell you Pinky but this 120 char limi
The other night flipping through channels I noticed a lady on QVC selling this laptop. Nothing unusal, but when she mentioned Athlon 4 and I saw the name on the screen, I got interested. I think they priced it at ~$2k, but since Athlon 4s are not going to be available until Q3 2001, I wonder what kind of ploy QVC invented this time... "Sure, we'll take your money, but you will get the laptop in September." or "Athlon 4, Pentium 4, come on, what's the difference? They all are fast, so who cares."
Will the processor use ACPI support to use the power-stepping features? ACPI doesn't work in Linux yet.
Got friends?
To be fair, though, (at least in the local computer store here) I've actually overheard store clerks recommending Durons instead of Celerons (and even P3s in some case) and Athlons instead of P4s. The store clerk is there to help out the consumer, and I think less and less Intel chips are being sold because of simple word-of-mouth.
.sig: Open Source, Open Mind
Granted Athlon has had two previous cores but whatever happened to the third one? Unless they go in powers of two (1,2, 4.. Athlon 8 = Sledgehammer? After all it would be K8..) .18 micron core(2), the T'bird(3), and now the Palomino(4).
The article states the following:
"Technically speaking, the Palomino core does mark the fourth AMD Athlon core since the release of the original K7 core in 1999. If we begin counting at the K7 core there was the 0.18-micron Athlon which was based on the K75 core, then the 0.18-micron Thunderbird with on-die L2 cache and the fourth Athlon core would be the 0.18-micron Palomino core."
So there's the origional core(1),
i wish i was but oh well
Indeed. But it would almost certainly cut the cost of developing the new motherboard versions to support it. Chipset compatibility could well mean upgrading to a newer chipset - but then this could be pin-compatible with the older one anyway! Voltage level and BIOS support are both trivial changes (from a hardware point of view).
So - yes you're right. It doesn't *necessarily* mean motherboard compatibility (eg look at early and late versions of the celeron... I for one have written off both chip and motherboard by plugging the wrong one into the socket!). It means cheaper motherboards available faster because of lower design costs.
Athlon 4 actually is the fourth version of the Athlon core (says so in the article). First was .25 micron Alu. Then .18 micron Alu. Then Alu Tbird, and then Cu Tbird although the copper version doesn't really count as separate.
It supports both SDR and DDR chipsets. For some reason the Compaq laptop which is available right now uses SDR, however that is not because DDR chipsets are not available (Ali makes DDR laptop chipset). DDR should become very popular for portables because it uses a lot less power than SDR.
The amount of misinformation I've read in less than a minute is making me sick. What is the ratio of noise to real information on /. now? About 1 out of every 1,000 of you know what you are talking about, most of you seem to get off at seeing your nickname in the fucking thread. salimma1 you have no idea how many cores of the Athlon there have been. The previous to this AMD was on the model 4. A quick read of the tech docs tells you that. VAXman, so it is a paper launch eh? Then what do you call the GHz mobile PIII? There is not enough data yet to determine how many of these are available yet. We probably won't have that data until the conference call AMD has in about an hour, or maybe until the Q2 earnings conference call, calling it a paper launch is premature. If AMD wanted to paper launch it then why didn't they announce when the PIII was announced. That does not make sense. nathanm, do you seriously think you will be seeing a mobile P4 anytime soon? If you do you need to do some research. You can't even get a P4 into a 1U server, much less a notebook. The Athlon will scale well past what the PIII will be able to do. Everyone knows that, or they should. "All AMD ever does is rip Intel's work and force them to release inferior products. " WTF? wyrm, no shit they don't have reviews, what are they going to do? buy a few laptops and compare them? This is a mobile part that probably doesn't work in the desktop because it has a different amount of pins. Even the boy wonder what he is saying some of the time couldn't get a laptop. In fact the screensavers guys I talked to Saturday are the only ones I know that have one.
ignorance is bliss. googlefiberatx.com
The most important part of this article was AMD's statement that the socket-A layout will continue throughout this series and the next two.
Contrast this to Intel, who is so bent on shoving stuff down our throats that they willingly sell products that have a short or no real life span expectancy. (p462 anyone?)
I think this statement from AMD may actually help them among the fence sitting OEMs who still seem glued to AMD. By keeping the same socket it allowd manufacturers to refine the product, instead of trying to figure out a new one.
* Winners compare their achievements to their goals, losers compare theirs to that of others.
--
True enough. I hadn't considered that. I have some of the same questions about SOI, but I'm no EE, and therefore have no freaking idea ;-)
Where's my lobbyist? Right here.
All of these changes sound good on paper, but it seems AMD has overlooked one detail. The pentium 4 core was built to allow for an amazing ramp-up of clock speeds. That's why it has a 20 stage pipeline. What Intel definitely understands is that the public thinks Mhz == speed. Yes, the enthusiast crowd knows this is false, but it sells anyway. I wonder how far they can push the improved Athlon core before they hit architectural limits on speed....
Where's my lobbyist? Right here.
You're right. We should just go back to using Vaxen. We could roll things back even farther, but the PDP-11 is far too old to be of use, so we'll only go back to the VAX.
If our friends in the patent office and in the courts had a clue, the patent for integrated circuits would have been awarded to Texas Instruments instead of the guy who went on to found Intel. But the patent wasn't awarded on who got there first, at least not when the courts got done with the issue.