http://www.dailytech.com/article.aspx?newsid=850

look at that.

I say look dont read. I am unsure there is anything there that means anything. Having half the amount of actual wattage used while not changing the 90nm to 65 nm doesnt seem like it would be dropping the wattage so drastically.

I mean, how much does say a 4200 x2 use at normal voltage on max and on idle anyways?

Of course, the bigger question is, not do these numbers mean anything, but did they ever?

I dont want to wait for Am2. I would have to buy ddr2 800 mhz ram anyways.

oh yeah, that's how they are going to get faster performance if you havent read lately, ddr2 800. In other words, it isnt faster, the memory is. That's a lame way to show speed increase, letting the memory, which you didnt make go faster on a board that you also didnt make.

hm.

Hard to tell. I'd bet if anything these are simply normal Athlon64, but running at a lower voltage.

Um....so, like given that anyone with an A64 & C'n'Q or BIOS undervolting can do that themselves, what was the point of AM2 again (for silencers)?

Better manufacturing process, producing cooler running chips? New core revision that's more optimised for power savings?

As far as I can tell, AMD didn't go to AM2 for increased performance. They've already told us that they expect AM2 systems to perform some 5-8% better only.

The only point of AM2 is the DDR2 memory. Since memory doesn't make sound......................

That's the thing, I'm looking at the TDP numbers from the link at the top of this thread, and they are virtually identical to the TDP's of the current S939 chips (some are even higher!). So no power-efficiency increase, no or little performance increase, ?????



Will it really make a significant difference to performance? Especially since AMD has some of the best memory architecture anyway (integrated controller, Hyper Transport bus, etc)?

The main reason for AMD to move to DDR2 is that Intel has moved to DDR2. The fact Intel made that move means that somewhere next year, maybe even this year, DDR memory will become more expensive than DDR2 memory. If AMD would still be using DDR, they would have a problem. This is the reason they are moving both A64 and Sempron to AM2. Would be a nice joke: budget CPU (Sempron) can only be paired with expenive memory (DDR).

The (for silencers) more interesting things AMD has on their roadmap (different SOI, slow transistors, 65nm, partial C'n'Q) will come later. They could make all those changes at once, but that would be taking a huge risk. They are taking the cautious approach: one step at a time.

Some of the confusion about the TDP numbers comes from not actually understanding what the "TDP" is. It is not how much power that particular chip will consume.

Rusty, I'm not a noob, I know what TDP is and that it does not reflect what a chip will actually dissipate in normal use. However, surely if a chip is more power-efficient than its predessor, it should have a lower TDP? For example my P4M cpu puts out about half as much heat as my A64 3000 at the same Vcore and clock speed, and indeed the TDP of the P4M is about half that of the A64, 30W vs. 67W.

There's also the matter of public perception: before the K8, AMD was Intel's shadow who brought modest improvements to chip designs based on sockets Intel had abandoned, all aimed at the "value market" (going all the way back to the i386 architecture). Even the K7 was based on Intel's P3 architecture (with some modest clock for clock performance improvement), but because of litigation AMD was forced for the first time to implement their own unique socket which forced them to briefly, and unwillingly, enter the chipset market. With the K7 they, also for the first time, surpassed Intel in performance but it was only with the K8 that they really made the leap from innovating (on existing Intel design) to inventing (with the onchip mem controller, use of hypertransport, etc).

Since ddr2 is no rambus, and AMD's already fought and won that battle (well kinda sorta, since Intel themselves abandoned proprietary rambus in favour of ddr), it only makes sense for them to adopt the higher performance ddr2 going forward, rather than willingly continue to saddle themselves with "legacy" ddr and all that association implies.

Totally off-topic, this was an integral part of AMD's "Virtual Gorilla" strategy formulated by then-CEO, Jerry Sanders:



source: wiki

Please clarify...

A lot of times when TDP is stated, this is short for TDP(max), and in this case, it does mean the maximum power the chip will consume.

In other cases, they are talking about the TDP "envelope" - the maximum dissipation for a package/architecture/etc.

These different uses create confusion, I think. Do you agree?

For those who are reading this thread and have no idea what's going on:

AMD typically uses TDP max when they give out numbers, and the actual TDP is always lower than the stated TDP.

Intel on the other hand report TDP differently. I do not know the specifics, but basically Intel's TDP numbers are for CPUs under 75% load, as opposed to 100% load for AMD.

The Pentium M CPU has a TDP of 25w, but it is actually higher than that under load. Having said that, a laptop running on battery using the Pentium M typically lasts longer than a Turion 64 laptop with similar components.

It would be nice if AMD and Intel can standardize on a TDP measurement methodology so that people can make an apple to apple comparison.

This is something you've made up yourself. TDP "max" or TDP "envelope", they still usually aren't applicable to specific products. On these boards, as well as most others, it doesn't matter how many times the term TDP is explained, it still doesn't sink in.

Thermal Design Power is exactly what it sounds like. It's a figure used in the design of products such as motherboards, CPU coolers and the likes. Example from the URL in the original post:

Athlon64 X2 5000+ (2.6GHz) = 89W TDP
Athlon64 X2 3800+ (2.0GHz) = 89W TDP

Does it make any sense that these two CPUs manufactured on the same process and running on the same voltage (atleast it won't be higher on the 3800+) are outputting the same amount of heat? Nope. The 89W figure is simply given to other manufacturers so that they can verify that the line (3800+ to 5000+) will function correctly on their hardware.

In this case, the 5000+ might not even use 89W. It might use 50W, for all we know. The ONLY thing we know is that neither of these CPUs uses more than 89W. Common sense also tells us that the 3800+ won't be near the 89W ceiling.

For this new EE range of AMD processors, it seems as if AMD has chosen to have two different TDPs (35W and 65W). This also makes it easy for companies to support some of the line up. Just as in the previous example, it is evident that the X2 4800+ 65W and the X2 4000+ 65W don't output equal amounts of heat in the real world.

Another example: My brother's Sempron 2800+ rated at 62W TDP is considerably cooler running than my old AthlonXP-M 2600+ 47W. Again, the Sempron is very far from the 62W roof.

When are people going to stop using TDP to draw flawed conclusions? At best, it gives a rough indication of what to expect.


Again: That number is for the whole Pentium-M line (actually 27W) and as far as I know, no P-M has really been that power hungry in reality. For example, my P-M Dothan 1.7GHz seems to output around 17W stock. Other measurements done on Pentium-M systems show the same thing.

No, I didn't make that up. In the Intel datasheets for processors, they have ratings for TDP(typ) and TDP(max). Here is one just pulled at random: http://www.intel.com/design/intarch/prodbref/27334803.pdf. Clearly these are power consumptions for a processor running at different operating conditions.

What you are describing, where different processor grades have the same TDP, is the TDP "envelope," and I agree with you that it is relating to the platform/architecture.

You can keep it clear in your mind by ignoring the other use of TDP - but you are misunderstanding why there is confusion on this point, which is fundamentally because the same term is used for different purposes.

TDP typical and TDP max are simply measuring two different things. I don't argue against that. The max figure is probably derived from simulations where more or less all transistors are switching and the typical figure is derived from "real world" conditions. Also, it is quite clear that the TDP in this case refers to a specific product.

What I'm opposing is the distinction made between TDP envelope and TDP and that these are used differently by the manufacturers. They generally just use "TDP" and that may be for a specific product, although in that case it is often stated as "power consumption". More often than not, TDP is simply a guiding figure for a whole product line. That is exactly what these AMD figures that the thread discusses are, which makes it hard to draw any decisive conclusions based on them.

Regardless of the seeming across the board 65w and 35w TDPs, if these AM2 Athlons do run cooler than the current generation S939 Athlons, then I may reconsider waiting for AM2 to come out before upgrading from my current Athlon XP. After all, cooler components typcially make for a quieter computer.



My mistake, I was thinking of the Turion 64 MT series when I typed in 27w. Based on what I've read in the past Intel typically reports TDP number at 75% load (and some other fine details that I don't recall). I suppose in the case of the Pentium M, actual TDP really is less than the estimated TDP.

That's for AM2.
http://forums.silentpcreview.com/viewtopic.php?t=29255

It's not a matter of usage. AMD issues TDP numbers as a guideline for hardware developers, and the numbers reflect the maximum wattage draw for a "series" of CPU's.

A hypothetical example:
"Series 1" are unvealed with a TDP of 50 watts. Later, AMD releases a "Series 2" line of chips that are more efficient, but AMD has roadmapped them to scale to much higher clockspeeds than Series 1. Clock-for-clock the "2" chips are more efficient, but their TDP may be listed as higher than the "1" chips, to allow for the future upscaling.



Only in Intel-land. Mikael explains it pretty well. AMD's "TDP" is not directly comparable to any of Intel's TDP(max), (case), (junction), (kitchen sink), etc. nomenclature.

Personally, I'll wait until Xbit reviews the new DDR2 chips before I worry about their power consumption too much. They're doing the best "real world" CPU testing I've seen lately.