Question Alder Lake - Official Thread

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Ajay

Lifer
Jan 8, 2001
15,468
7,873
136
I would look up what the word poll means for devs if I would care enough but I doubt that it's exclusively used for assembly level hardware poking (pulling/pushing, whatever it would be) .
The whole point was that the game checks for user input, indifferent from how this is accomplished.
It does matter how it is done. I'll try not to bother you anymore.
 
  • Like
Reactions: lobz

Hulk

Diamond Member
Oct 9, 1999
4,228
2,016
136
Very nice comparison, Alder Lake Cinebench R20 performanse per watt........................ :grimacing:


Assuming this data is legitimate there are some interesting conclusions to be drawn.
First, as we know in a highly MT application like CB lots of efficient cores running at "lowish" frequency are going to do well. Hence the great 5950X score at 142W. We can see moving from 12 to 8 cores, even though frequency increases, due to the nonlinear nature of increasing frequency vs power, performance and efficiency decreases. As expected. Still it's nice to see things behaving as they logically should.

The same behavior holds for ADL at 125W and 241W. Those Gracemont cores are definitely helping out. Also notice as the Alder Lake parts go from 125W to 241W as you move to the lower core parts the increase in performance decreases as the additional power is basically wasted trying to eek out the last few MHz.

Finally notice that the 12700K at 125W performs about as well as the 5800X at 142W. Does this mean the 12700K is more efficient than the 5800X? Not so fast. There is some nuance here. The 5800X has 8 cores while the 12700K has 12 cores so those 12 cores aren't pushed as far into that nonlinear zone of the V x F graph. I would think the 5800X wouldn't lose much performance is backed down from 142W to 125W.

There is a lot going on here. It's going to take some good reviewing and keen readers to realize that these parts are looking to excel in different scenarios and then price points much be considered as well. Just about 24 hours until the NDA lifts!
 
Last edited:

Hitman928

Diamond Member
Apr 15, 2012
5,323
8,009
136
Assuming this data is legitimate there are some interesting conclusions to be drawn.
First, as we know in a highly MT application like CB lots of efficient cores running at "lowish" frequency are going to do well. Hence the great 5950X score at 142W. We can see moving from 12 to 8 cores, even though frequency increases, due to the nonlinear nature of increasing frequency vs power, performance and efficiency decreases. As expected. Still it's nice to see things behaving as they logically should.

The same behavior holds for ADL at 125W and 241W. Those Gracemont cores are definitely helping out. Also notice as the Alder Lake parts go from 125W to 241W as you move to the lower core parts the increase in performance decreases as the additional power is basically wasted trying to eek out the last few MHz.

Finally notice that the 12700K at 125W performs about as well as the 5800X at 142W. Does this mean the 12700K is more efficient than the 5800X? Not so fast. There is some nuance here. The 5800X has 8 cores while the 12700K has 10 cores so those 10 cores aren't pushed as far into that nonlinear zone of the V x F graph. I would think the 5800X wouldn't lose much performance is backed down from 142W to 125W.

There is a lot going on here. It's going to take some good reviewing and keen readers to realize that these parts are looking to excel in different scenarios and then price points much be considered as well. Just about 24 hours until the NDA lifts!

12700k has 12 cores, no?
 

Asterox

Golden Member
May 15, 2012
1,026
1,775
136
Intresting no doubt, if we look at this comparison and what he said in year 2016.Raptor Lake will just follow same road or PL1 around 250W.


"William Holt, head of Intel’s Technology and Manufacturing Group, made the announcement at the International Solid State Circuits Conference (ISSCC) this past week, when discussing some of the options Intel is evaluating. These technologies aren’t coming next year or the year after — all of the tech in question would be introduced after 2021."

Zen3

yellow is stock, R5 cant eat more than 90W

green, Eco mode


Zen 4 will have the same or maybe even a little lower CPU power consumption.

Intel, "hm sacrifice speed to reduce power consumption". Unexpected competition arrived, blah forget it we don't care. :grinning:

Maybe i'm crazy, but 125W TDP should be the absolute maximum TDP for Desktop processors.In real world year 2021 and future, "150W CPU package power" should be maximum and not a hair more.

2021-11-03_140255.jpg
 
Last edited:

Ajay

Lifer
Jan 8, 2001
15,468
7,873
136
Maybe i'm crazy, but 125W TDP should be the absolute maximum TDP for Desktop processors.In real world year 2021 and future, "150W CPU package power" should be maximum and not a hair more.

Yes, you are crazy :p. Honestly, I'd crank up my 5900X to 250 Watts if it gave me a 30% performance boost - but that’s not possible (on anything other than LN2). But, I’m part of that small DIY enthusiast market.
 

TheELF

Diamond Member
Dec 22, 2012
3,973
731
126
Zen3

yellow is stock, R5 cant eat more than 90W

green, Eco mode


Zen 4 will have the same or maybe even a little lower CPU power consumption.

Intel, "hm sacrifice speed to reduce power consumption". Unexpected competition arrived, blah forget it we don't care. :grinning:

Maybe i'm crazy, but 125W TDP should be the absolute maximum TDP for Desktop processors.In real world year 2021 and future, "150W CPU package power" should be maximum and not a hair more.

View attachment 52262
Keep in mind that ryzen uses another ~30% of power that comes from the socket, so unless some site shows both the CPU and the mobo power draw, it's uncertain if these power numbers for zen can be believed.
According to AMD themselves getting less power from the socket can reduce performance.
We’ll quote directly from AMD’s review documentation so that there is no room for confusion:

Package Power Tracking (“PPT”): The PPT threshold is the allowed socket power consumption permitted across the voltage rails supplying the socket. Applications with high thread counts, and/or “heavy” threads, can encounter PPT limits that can be alleviated with a raised PPT limit.

  1. Default for Socket AM4 is at least 142W on motherboards rated for 105W TDP processors.
  2. Default for Socket AM4 is at least 88W on motherboards rated for 65W TDP processors.
Intresting no doubt, if we look at this comparison and what he said in year 2016.Raptor Lake will just follow same road or PL1 around 250W.


"William Holt, head of Intel’s Technology and Manufacturing Group, made the announcement at the International Solid State Circuits Conference (ISSCC) this past week, when discussing some of the options Intel is evaluating. These technologies aren’t coming next year or the year after — all of the tech in question would be introduced after 2021."
Look at how much more performance AL gets compared to rocket at the same power, so efficiency is way up from previous gen and that's done by using more cores and running them at lower speeds.
At the 125W which is your maximum it reaches close to the 5950x (about ~5% difference) which is $200 more expensive.
 

Hitman928

Diamond Member
Apr 15, 2012
5,323
8,009
136
Keep in mind that ryzen uses another ~30% of power that comes from the socket, so unless some site shows both the CPU and the mobo power draw, it's uncertain if these power numbers for zen can be believed.
According to AMD themselves getting less power from the socket can reduce performance.


Look at how much more performance AL gets at the same power, so efficiency is way up from previous gen and that's done by using more cores and running them at lower speeds.

At stock, with precision boost on, a 5950x/5900x will use up to 142W, just as the chart shows. A 5600x or 5950x/5900x/5800x in eco mode will use up to 88W just as the chart shows. Your quote is just re-affirming what is in the chart for AMD CPUs.
 

Hulk

Diamond Member
Oct 9, 1999
4,228
2,016
136
As I wrote above, power discussions are going to have to be very specific to be meaningful moving forward. For example, are you crunching numbers in perfectly scaling MT applications 24 hours a day (aka distributive computing)? Then as this point in time a 5950X cruising along at <4GHz is going to crush the competition in terms of throughput and power consumption.

On the other hand is your work flow generally light in compute with bursty moments that require a lot of ST and/or MT compute? In that case Alder Lake starts to make sense. I look at Ian's video editing example of the Thread Director being in-line with what I would expect actually in this scenario. For example, I just finished editing a video and I'm frameserving it from Vegas Pro to Handbrake for final render. But now I'm moving on to some photo editing in Photoshop. I personally want the Golden Coves with me in the foreground while I edit those photos because those are the actions I can FEEL the responsiveness (or lack of) when I'm working.

I will grant that the 5950X is still a beast of a CPU and can quite effectively cover all bases. But it is expensive and has no integrated GPU and with GPU's being rare as hens teeth these days that is a factor.

Different strokes for different folks.
 

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Keep in mind that ryzen uses another ~30% of power that comes from the socket, so unless some site shows both the CPU and the mobo power draw, it's uncertain if these power numbers for zen can be believed.
According to AMD themselves getting less power from the socket can reduce performance.
What I'd like to see is a a review that shows both package power and wall power for power consumption. Throughout the advent of the Zen architecture and subsequent iterations those chips always consume more power in gaming, even when churning out inferior fps. Something isn't right about a "power efficient" platform that is only power efficient when you measure power in software, but as soon as you measure at the wall it's no longer that power efficient.
Additionally, that platform is plagued with heat issues, and most enthusiasts resorting to undervolting and secondary cooling (even 5600x owners, a chip that supposedly consumes no more than 88watts tops at stock). Yeah, I've heard about heat spots and all that but guess what, Intel goes to 250w on the package alone and suffers from heat spotting too, and whatever other heating effects pumping 250w into your chip can cause. So, I think this needs to be looked into, the delta between wall and reported package power using a similar brand and tiered motherboard, a barebones bench table, and peripherals to reduce/eliminate any differences between setups.

1800x
1635955098565.png

2700x
1635955146567.png

3900x
1635955173803.png

5950x
1635955205672.png
 
Last edited:
  • Like
Reactions: igor_kavinski

Asterox

Golden Member
May 15, 2012
1,026
1,775
136
Look at how much more performance AL gets compared to rocket at the same power, so efficiency is way up from previous gen and that's done by using more cores and running them at lower speeds.
At the 125W which is your maximum it reaches close to the 5950x (about ~5% difference) which is $200 more expensive.

Yes, but Intel has competition or just look at Ryzen 7 or 9 Eco mode vs Alder Lake defoult 125W.

You can expect just that comparison soon, or CPU performance in various applications and games.
That's the point, vs just a CPU power consumption=score in Cinebench bench.This will be a very interesting comparison no doubt.
 
  • Like
Reactions: Makaveli

Abwx

Lifer
Apr 2, 2011
10,965
3,485
136

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Also, based on everything that we've seen so far, Golden Cove seems to be on par or better in power efficiency than Zen 3 Core for Core, at the same clock. The inefficiency only stems from the massive all-core frequencies Intel is pushing the the flagship 12900k, especially. Looking forward to some detailed power analysis tomorrow.