AMD LE-1620 PIB 45W 2.4GHz 1MB cache AM2 uniprocessor

[Felger Carbon]

CPU is AMD LE-1620.

PIB part number is ADH1620DHBOX. ~65-70 bucks US on Froogle, folks are accepting advance orders due in ~2 weeks. One core. For those of us with lite-duty tasks, one core makes sense becuz core 2 does nothing but sit and dissipate unneeded power.

4 times the cache of the current 45W Athlon 64 Limas (.65u) for about the same price. *No info* whatever on the official AMD site, but then parts aren't yet on the retail shelves.

[HueyCobra]

New low-power single-core Athlons coming

The two processors are scheduled to launch this quarter, but they're not based on AMD's new architecture, nor are they built around the company's 65nm process technology. Instead, X-bit labs says the CPUs are 90nm Athlon models clocked at 2.2 and 2.4GHz with 1MB of cache and a thermal envelope of just 45 watts. The 2.2GHz model will be dubbed Athlon LE-1600, while its 2.4GHz sibling will be known as the Athlon LE-1620. Prices for both models should be lower than $70, X-bit labs suggests.

Update: X-bit labs seems to have pulled the story. However, the information appears to be valid: we can see the two chips listed as available for pre-orders at some European online retailers. Italian e-tailer ePrice, for instance, lists the LE-1600 for €44.27 ($62.66) and the LE-1620 for €49.18 ($69.61).

[ronrem]

the first LE Sempron (65nm Sparta core) is already at New Egg and 45.99.
1.9 ghz.....256 l2 cache....should run way cooler than the already very cool Manila 90nm equivalent.

[drees]

AMD Sempron LE-1100 Sparta 1.9GHz 256KB L2 Cache Socket AM2 Processor @ NewEgg

Figure we need a link.

[jaganath]

Intel competitor chip is Celeron 420 (Conroe-L), no speedstep but lower TDP, 35W.

[HueyCobra]

It would be interesting to see how the Athlon LE's performance compares to the Sempron LE and Celeron 400.

lower TDP, 35W.

Eh ... Maybe AMD's "Average CPU Power" will be more equivalent to Intel's TDP ratings.

• Average CPU Power white paper (82KB PDF)
• Introducing Average CPU Power

• AMD's Latest Intel Counterattack
• Barcelona: AMD Unveils New Power Metric for CPUs
• The nuts and bolts of the quad-core Opterons

[accord1999]

It would be interesting to see how the Athlon LE's performance compares to the Sempron LE and Celeron 400.

lower TDP, 35W.

Eh ... Maybe AMD's "Average CPU Power" will be more equivalent to Intel's TDP ratings.

No, since CPU power measurements show that if anything Intel's TDP are more over-rated than AMD's.

[smilingcrow]

Why did you create two separate threads for this CPU?
No one responding to your first thread so why not just bump the first one.

[Felger Carbon]

Why did you create two separate threads for this CPU?

The SPCR site has been acting up for me. Dunno if it's me, my ISP, or SPCR's ISP, but I get frequent "page not found" errors from time to time. When I entered that post the first time, I got the "page not found" error and believed it had not entered. It had. So in ignorance (a common state for me) I entered the post again.

AFAIK, I can't delete a post once entered, including my own.

[HueyCobra]

No, since CPU power measurements show that if anything Intel's TDP are more over-rated than AMD's.

Could you please clarify what you meant? I read that a couple of times and came up with about 10 different interpretations. If you've seen any comparisons between AMD's new rating system and Intel's, please include the links too.

[andyb]

As we know, AMD and Intel both have brackets for their CPU usage instead of saying exactly how much a CPU draws at max they say how much a whole group of CPU's draw at max i.e. 45W

This CPU apart from its price doesnt look that good if you just read the 45W number, as the dual core CPU will also use a max of 45W but has 2 cores, so this would suggest that the single core CPU is half as efficient.

My guess though is that most of the extra power usage is due to the extra cache, but we dont know how much power this CPU actually uses, this will be interesting if the real power draw is sub 30W.


Andy

[Felger Carbon]

...we dont know how much power this CPU actually uses, this will be interesting if the real power draw is sub 30W.

The speculation is that this is half an X2, so AMD gets to use the CPUs it has set aside that "half work". Most of them are 69W if whole, but can exactly 50% of a half-OK X2 be laser-isolated?

[HueyCobra]

As we know, AMD and Intel both have brackets for their CPU usage instead of saying exactly how much a CPU draws at max they say how much a whole group of CPU's draw at max i.e. 45W

According to Athlon 64 for Quiet Power and ACTUAL Intel Processor Power Consumption!, Intel's TDP ratings are typical rather than maximum.

[accord1999]

No, since CPU power measurements show that if anything Intel's TDP are more over-rated than AMD's.

Could you please clarify what you meant? I read that a couple of times and came up with about 10 different interpretations. If you've seen any comparisons between AMD's new rating system and Intel's, please include the links too.

A number of reviews have shown that C2D based processors have measured real-world maximum power consumption noticeably lower than their rated TDP.

http://www.behardware.com/articles/674- ... -ddr3.html
http://www.hardtecs4u.com/reviews/2007/ ... index9.php
http://www.matbe.com/articles/lire/385/ ... /page5.php

[ronrem]

Not sure how that "45W chip" with the big cache will stack up. The Sempron Sparta will no doubt draw less wattage than a similar 90 nm Manila or Palermo...and those probably sucked less than 45w real world.

Obviously....a Dual core and a single core at a similar clock won't suck the same power-and a chip with a lower clock+ 1/4 the cache (single core) won't also draw the same.

The "45 Watt" stuff only means in general AMD tried a bit harder to have the chip run low power and cool. Otherwise these wattage numbers are relative only in a very very general way.

I want to know what temps it hits with a fanless Ninja in an average room outside a case...or some similar context....and specific to each model...not an abstract worst case for the hottest unit in a family of chips that might,like Windsor. range from a 3600 to a 6000. Why is this simple bit of info treated like a military secret?

[Felger Carbon]

I want to know what temps it hits with a fanless Ninja in an average room outside a case...

In the entire universe, how many Ninjas do you think operate outside a case, with absolutely no fan (not just passive, but actually fanless)? If you want to know that, you're gonna have to measure it yourself. Ain't nobody, but nobody's gonna measure it for you.

[HueyCobra]

A number of reviews have shown that C2D based processors have measured real-world maximum power consumption noticeably lower than their rated TDP.

While that is happily the case (although is it the case for literally all Core 2s, or do some exceed typical TDP?), the same is obviously true for AMD processors given their max rating system, and the margin for reduction with the new "ACP" rating may be even greater. Whether or not average/typical TDP is desirable is another debate, but having comparable metrics seems inarguable to me.

[smilingcrow]

I’ve given up considering TDP as an especially useful metric especially when trying to compare AMD and Intel. I’d rather look at the wealth of real-world data that shows the relative performance and power consumption of CPUs and motherboards along with over-clocking and cooling data.

[ronrem]

I want to know what temps it hits with a fanless Ninja in an average room outside a case...

In the entire universe, how many Ninjas do you think operate outside a case, with absolutely no fan (not just passive, but actually fanless)? If you want to know that, you're gonna have to measure it yourself. Ain't nobody, but nobody's gonna measure it for you.

Okay...perhaps there's got to be some small airflow.

Looking at user reports on New Egg..for the regular 4000 Brisbane..one user says OC to 2.6 ghz,...stocks HSF...29C idle and 35C load. Another says gets 32C with a 2.5 ghz OC. Others claim pretty good OC at stock volts.

I'd figurwe the BE and LE chips can top that..and that a Ninja or other good HS can outperform the stock HSF easily. These new 65 nm chips obviously
are intended to be cooler than a 4000 X2 Brisbane,or a 90nm Manila Sempron. Those chips have managed to run quite cool without all the SPCR type tricks. I find it a bit frustrating that with all the "reviews" "tests" we really are left with just a rough estimate.

Felger....as you know I'm now pretty gung ho on the idea of a big fan at very low rpm. The 3600 and 4000 Brisbanes came pretty close to being coolable with only a big fan at under 500 rpm----or at least that appears to be so. If the BE dual cores can run a bit cooler....then a 1 fan computer is pretty much do-able. The Sparta Sempron is especially interesting in this context. With a big,high grade heatsink...and a 220 mm bigfan....as the ONLY fan ----how low can the fan's rpm be and still keep the CPU under 50 C worst case? That's basically what I'd want to have some info on. So...yeah...I guess the idea of a Ninja but no powered airflow at all would be a yardstick hard to apply. I'd be content with a report on that Sparta with SOME kind of "silent" cooling that I can accept as a guide...say a case with one 800 rpm 120 mm fan and a passive heatsink....or a big fan at whatever... just SOME quiet oriented setup to hint at how quiet these CAN get.

[HueyCobra]

Ars Technica - AMD launches new line of energy efficient desktop processors:

AMD has launched a new series of 45W parts designed to appeal to DIY users looking for ultra-quiet PCs, OEM system builders looking to emphasize power-per-watt, and companies trying to exceed the EPA's ENERGY STAR (Version 4) specification. All of the CPUs are listed as 45W parts, per AMD's decision to list maximum potential thermal dissipation, rather than actual expected dissipation under real-world loads and conditions.

According to EETimes, AMD's new 45W product line is designed as follows:

• AMD Athlon 64 X2 BE-2400 (2.3GHz): $104
• AMD Athlon 64 LE-1620 (2.4GHz): $53
• AMD Athlon 64 LE-1600 (2.2GHz): $47
• AMD Sempron LE-1250 (2.2GHz): $53
• AMD Sempron LE-1200 (2.1GHz): $48
• AMD Sempron LE-1100 (1.9GHz): $37

Given the disparity between AMD's TDP (Thermal Design Power) and the actual amount of energy consumed when measured in real-world conditions, I'd expect some of these chips, particularly the sub-$50 Semprons, to offer an excellent performance-per-watt ratio, particularly in single-threaded workloads or when, say, powering a home-built Windows Home Server or HTPC.

[austinbike]

OK, here's the TDP low down. I am going to put this in terms of xeon vs. opteron, I don't know if it holds for desktops as well.

Intel has circuitry that will lower the speed of the processor if it gets too hot. So they can claim a TDP of, say 85-90% of the max power. So an 80W clovertown is really ~90W+ of max power (what the system builder has to design to).

AMD doesn't have this circuitry, so max power is max power.

One could argue that the intel design is more elegant, but the reality is you want the system to throttle down when you are doing nothing, not when you really need the performance the most.

Intel TDPs cannot be compared to AMD. You best bet is with a power meter at the wall. Don't look at TDPs other than as a guide to give you a rough feel. It will never be a clear indicator of the actual consumption.

I have a 65W processor in my system and my total system draw is ~45-47W these days. Obviously the processor is drawing significantly less than the TDP.

[jaganath]

the reality is you want the system to throttle down when you are doing nothing

which is exactly what Speedstep (EIST) does.

[smilingcrow]

Having measured real-world power consumption comparing Prime95 to H.264 encoding on the same system, it seems as if the TDPs that are used by both companies will be hard to match unless you are benchmarking or stress testing your system:

Q6600 @ 3GHz 1.275V
Prime95 – 156.5W
H.264 encoding – 134W
Delta – 22.5W AC, ~16W DC

Q6600 @ 2.4GHz 1.05V
Prime95 – 107W
H.264 encoding – 92W
Delta – 15W AC, ~10.5W DC

Both applications stressed all cores 100%.

Has anyone written an article looking at which applications (not benchmarks) load the CPU the most in terms of power consumption? It would have to ignore games as the load on the GPU makes it hard to quantify how much the CPU is consuming.

[derekva]

http://www.dailytech.com/article.aspx?newsid=9215

More low-power AMD goodness...now with lower prices.

[djkest]

I believe that Mike could make a very good article on budget and "green" CPUs. I'd like to see perhaps using an 80+ or pico PSU, micro atx mobo with onboard video, and a laptop hard drive with some of these new amd cpus, and some of the low power low speed Intel CPUs.

I'd like to see power draw, thermal output (which should be very similar to power draw), and lastly BENCHMARKING to see how fast these low power chips are compared to full power counterparts.

[ryboto]

Having measured real-world power consumption comparing Prime95 to H.264 encoding on the same system, it seems as if the TDPs that are used by both companies will be hard to match unless you are benchmarking or stress testing your system:

Q6600 @ 3GHz 1.275V
Prime95 – 156.5W
H.264 encoding – 134W
Delta – 22.5W AC, ~16W DC

Q6600 @ 2.4GHz 1.05V
Prime95 – 107W
H.264 encoding – 92W
Delta – 15W AC, ~10.5W DC

Both applications stressed all cores 100%.

Has anyone written an article looking at which applications (not benchmarks) load the CPU the most in terms of power consumption? It would have to ignore games as the load on the GPU makes it hard to quantify how much the CPU is consuming.

what are the specifications of the system that's working in? Those numbers sound too good to be true. With my system, even undervolted, when I stress both cores the system pulls 109W from the wall. I imagine the Q6600 system has only 1 hard drive, and no video card?

[smilingcrow]

what are the specifications of the system that's working in? Those numbers sound too good to be true. With my system, even undervolted, when I stress both cores the system pulls 109W from the wall. I imagine the Q6600 system has only 1 hard drive, and no video card?

The G0 revision chips undervolt very nicely; details here.

Added:
The point of my post was to highlight how meaningless TDP can be. Here’s a CPU rated at 95W DC but when undervolted the whole system consumes 92W AC with all 4 cores at 100% load encoding H.264 video. 92W AC equates to roughly 74W DC with this setup.

From an SPCR perspective it means that 3GHz quad core is easy to cool silently which surprised me.

[ryboto]

Added:
The point of my post was to highlight how meaningless TDP can be. Here’s a CPU rated at 95W DC but when undervolted the whole system consumes 92W AC with all 4 cores at 100% load encoding H.264 video. 92W AC equates to roughly 74W DC with this setup.

From an SPCR perspective it means that 3GHz quad core is easy to cool silently which surprised me.

I'm just curious though, it must be a pretty basic setup for it to achieve such a low system power draw.

[smilingcrow]

I'm just curious though, it must be a pretty basic setup for it to achieve such a low system power draw.

Hardly. It lacks a VGA card which adds about 9 or 10W, but if you used a more modern PSU that would help reduce the power draw. If you follow the link that I gave above you can fully quench your curiosity.

With a mobile hard drive, less RAM and a more efficient PSU you could reduce the power draw even more. But regardless of this, it just shows that TDP can be very removed from real world power usage. This chip is so undemanding that I was able to cool it passively with a Ninja+ with just a single case fan; Nexus 120mm @ 540 RPM. I hit one stumbling block which was the puny Northbridge heatsink on the mATX board that I use. With a 120mm fan on the Ninja+ at 410 RPM everything was fine even at 3GHz.

[ryboto]

With a mobile hard drive, less RAM and a more efficient PSU you could reduce the power draw even more. But regardless of this, it just shows that TDP can be very removed from real world power usage. This chip is so undemanding that I was able to cool it passively with a Ninja+ with just a single case fan; Nexus 120mm @ 540 RPM. I hit one stumbling block which was the puny Northbridge heatsink on the mATX board that I use. With a 120mm fan on the Ninja+ at 410 RPM everything was fine even at 3GHz.

The TDP might still be accurate, you've only shown TDP data for overclocked and undervolted conditions. Very impressed with the power draw.