Power consumption and efficiency in gaming: AMD vs Intel

[FredFlint]

I just spent some time trying to find the idle power consumption between Intel and AMD, no reputable reviewer that i could find bothered with any of that, so far as i can ascertain its because the difference is so small no one thinks its worth bothering with, its 5 to 10 watts vs 10 to 15 watts.
Other than one lol Youtube video who tried to claim that due to "Intel's efficiency cores AMD's idle power consumption is 10X higher" and said that while he had HWINfo open with a big red arrow pointing at a 50 watt load on a Ryzen 7950X on the right hand screen with with an image stabilisation workload running on the left side of the screen
Its such an obvious shill attempt i'm not even going to post it here to give it views.

So still don't get it, the difference in idle power consumption between AMD and Intel is 5 watts, or 10 watts if you want to give Intel the benefit of the lowest value vs AMD's highest value, the difference in gaming is 60 to 200 watts, so if its 10 watt idle then you need to be AFK for 16 hours for every 1 hour of game time to even it out, if you cared that much why is your system running when you're not using it???? because where ever the system is actually doing something the Intel CPU is pulling more watts.

If for some strange reason you care about your power bill but also never turn the system off and game only for 1 hour per day then you would save £8 per year at 30p per KWh.
If you game for 3 hours per day well then you're ______ if you're on Intel.

The 7800X3D is £380
The 14700K is £400

So after 3 years you save.......... £4, double digits, £12 in fact if you go 4 years. £12 saved over 4 years and you still have the slower gaming CPU, this also ignoring the 7800X3D is so efficient it will run cool on a £25 CM Hyper 212 cooler, you need a high end £100 cooler for the 14700K.

Ridiculous arguments.

AM5 chips idle at 17-40W depending on RAM speed, using EXPO/XMP adds ~10W for me (5600 RAM). The lowest idle i have seen is ~16.5W, that is running ECO-65 and stock RAM speeds (4800) and iGPU disabled. Its normally ~22W, with default BIOS + EXPO its normally ~35-40W which is over 4x Intels 8-10W idle.

 

[mattyfez]

I can confirm my i5 idles at 10-20w.
When I day Idle, I mean i'm running steam, and firefox, and surfing the net, so not totally idle.
If I shut down everything and just stare at my desktop, it will idle at under 10w*.

But i've also undervolted it too, so that will be playing a part.

Just as an example, I've tabbed out of Forza 5 to type this, so its running in the bacground,and my 13600k is pulling about 30w.

 

[jigger]

I think a techtuber should do a video of all the various Intel chips sat idle. They would have to let the room heat normalise. A 4 hour video of each should do it.

 

[humbug]

AM5 chips idle at 17-40W depending on RAM speed, using EXPO/XMP adds ~10W for me (5600 RAM). The lowest idle i have seen is ~16.5W, that is running ECO-65 and stock RAM speeds (4800) and iGPU disabled. Its normally ~22W, with default BIOS + EXPO its normally ~35-40W which is over 4x Intels 8-10W idle.

Interesting that, you have 25 watts for the CPU and the SoC die is using 2X as much power as the CPU.

No matter what speed i ram at the idle power consumption is exactly the same, it makes 0 difference, 5800X.

I use a bit more power, about 5 watts if i run the ram at 3800Mhz vs 3200Mhz, but only under load.

Asus strikes again?

Edit you know what, this is genuinely interesting, fire up Ryzen Master for me please and take a screenshot while at idle.

I don't think i would be happy with 40 watts on the SoC die, that was cause the cooler to ramp up or overheat the chip at idle, i don't think the SoC die is even designed to pull that much power, again.... Asus, you at it again?????

 

[Rroff]

My 14700K is all over the place - the CPU averages 6.2w when "fully" idle but can drop to 3-4w for long periods of that - if I shutdown a bunch of Windows 11 **** it drops to about 4w average, but anywhere from 8 to 18-19w under light desktop use and can spike to 80-90w at times just opening YouTube, etc.

Personally I really don't care that much though, lower is better but it doesn't affect my purchasing decision.

As I recently mentioned quite impressed by the Ryzen Z1 Extreme though - limited to 5w TDP it can actually match or beat my old Xeon 1650 V2, gives a desktop Ryzen 5800X a good run for the money (after allowing for thermal soak of the cooling).

At 5w - with max 10w, average 4.85w (on battery power):

AMD Ryzen Z1 Extreme @ 3941.98 MHz - CPU-Z VALIDATOR

[bac4ni] Validated Dump by Anonymous (2023-12-25 02:17:38) - MB: LENOVO LNVNB161216 - RAM: 16384 MB

valid.x86.fr

At 30w TDP (on mains):

AMD Ryzen Z1 Extreme @ 4214.64 MHz - CPU-Z VALIDATOR

[xf86bg] Validated Dump by Anonymous (2023-12-21 16:14:55) - MB: LENOVO LNVNB161216 - RAM: 16384 MB

valid.x86.fr

EDIT: Though saying that shame it can't run even lower TDP by disabling parts as the performance is way overkill for these handhelds which are using it and it could make them really useful for non-gaming stuff if the CPU could drop back to like 1-2w which would still give great performance for light web-browsing, etc. while massively increasing the battery life (dunno if Ryzen Master can be used to that end here).

 

[humbug]

My 14700K is all over the place - the CPU averages 6.2w when "fully" idle but can drop to 3-4w for long periods of that - if I shutdown a bunch of Windows 11 **** it drops to about 4w average, but anywhere from 8 to 18-19w under light desktop use and can spike to 80-90w at times just opening YouTube, etc.

Personally I really don't care that much though, lower is better but it doesn't affect my purchasing decision.

As I recently mentioned quite impressed by the Ryzen Z1 Extreme though - limited to 5w TDP it can actually match or beat my old Xeon 1650 V2, gives a desktop Ryzen 5800X a good run for the money (after allowing for thermal soak of the cooling).

At 5w - with max 10w, average 4.85w (on battery power):

AMD Ryzen Z1 Extreme @ 3941.98 MHz - CPU-Z VALIDATOR

[bac4ni] Validated Dump by Anonymous (2023-12-25 02:17:38) - MB: LENOVO LNVNB161216 - RAM: 16384 MB

valid.x86.fr

At 30w TDP (on mains):

AMD Ryzen Z1 Extreme @ 4214.64 MHz - CPU-Z VALIDATOR

[xf86bg] Validated Dump by Anonymous (2023-12-21 16:14:55) - MB: LENOVO LNVNB161216 - RAM: 16384 MB

valid.x86.fr

You're not kidding....

just shy of 6300 MT at 30 watts. What the funky? I get about 6850 with my curve optimised 5800X at 80 watts. And i think, or thought the performance per watt there was pretty damned good even now....

Nope!

Also: LENOVO Legion Go?

 

[chaparral]

My 14700K is all over the place - the CPU averages 6.2w when "fully" idle but can drop to 3-4w for long periods of that - if I shutdown a bunch of Windows 11 **** it drops to about 4w average, but anywhere from 8 to 18-19w under light desktop use and can spike to 80-90w at times just opening YouTube, etc.

Personally I really don't care that much though, lower is better but it doesn't affect my purchasing decision.

As I recently mentioned quite impressed by the Ryzen Z1 Extreme though - limited to 5w TDP it can actually match or beat my old Xeon 1650 V2, gives a desktop Ryzen 5800X a good run for the money (after allowing for thermal soak of the cooling).

At 5w - with max 10w, average 4.85w (on battery power):

AMD Ryzen Z1 Extreme @ 3941.98 MHz - CPU-Z VALIDATOR

[bac4ni] Validated Dump by Anonymous (2023-12-25 02:17:38) - MB: LENOVO LNVNB161216 - RAM: 16384 MB

valid.x86.fr

At 30w TDP (on mains):

AMD Ryzen Z1 Extreme @ 4214.64 MHz - CPU-Z VALIDATOR

[xf86bg] Validated Dump by Anonymous (2023-12-21 16:14:55) - MB: LENOVO LNVNB161216 - RAM: 16384 MB

valid.x86.fr

EDIT: Though saying that shame it can't run even lower TDP by disabling parts as the performance is way overkill for these handhelds which are using it and it could make them really useful for non-gaming stuff if the CPU could drop back to like 1-2w which would still give great performance for light web-browsing, etc. while massively increasing the battery life (dunno if Ryzen Master can be used to that end here).

Rroff have you tried undervolting you're 14700k ?

This review says the temps can drop by up to 20c and reduce the power draw by huge amounts

 

[FredFlint]

Interesting that, you have 25 watts for the CPU and the SoC die is using 2X as much power as the CPU.

No matter what speed i ram at the idle power consumption is exactly the same, it makes 0 difference, 5800X.

I use a bit more power, about 5 watts if i run the ram at 3800Mhz vs 3200Mhz, but only under load.

Asus strikes again?

Edit you know what, this is genuinely interesting, fire up Ryzen Master for me please and take a screenshot while at idle.

I don't think i would be happy with 40 watts on the SoC die, that was cause the cooler to ramp up or overheat the chip at idle, i don't think the SoC die is even designed to pull that much power, again.... Asus, you at it again?????

Bumping the RAM up increases the SOC voltage, this increases the power/heat. At 4800, the SOC is at ~1v, at 5600, its at ~1.25v. Think this is just how AM5 works, so not an ASUS thing. I played around with the setting and tested power at load and idle (using HWiNFO64) at first but now I have my PC stable and don’t want to change any setting or install anything else.

 

[Rroff]

Rroff have you tried undervolting you're 14700k ?

I only had a quick play with it - seemed my chip had more frequency headroom at stock voltage than potential for under volting - at a guess I don't think one of the cores has much margin voltage wise while 2 of the cores seem to have loads of frequency headroom and 2 not bad - I can get 6GHz boost under most gaming workloads (on air cooling) with little or minor voltage increase without anything dramatic changing thermal or power wise.

EDIT: To be honest I'm not seeing anything too dramatic power/thermals wise with my 14700K outside of synthetic loading - for most gaming situations it is sitting at 60-70C and power draw isn't terribly different to what I'm used to having mostly run stuff like socket 2011 CPUs, etc. for the last decade.

Despite the attitude of meh from the reviewers and most people when it comes to the 14th gen, and to an extent I'd agree with it when it comes to the 14600 and 14900, and some reservations as to the lack of upgrade path and/or when next generation CPUs drop, I've no regrets buying the 14700K, actually very happy with the purchase - definitely the sweet spot IMO as things stand (assuming the pricing doesn't change much).

 

[humbug]

Bumping the RAM up increases the SOC voltage, this increases the power/heat. At 4800, the SOC is at ~1v, at 5600, its at ~1.25v. Think this is just how AM5 works, so not an ASUS thing. I played around with the setting and tested power at load and idle (using HWiNFO64) at first but now I have my PC stable and don’t want to change any setting or install anything else.

You do know about Asus have had issues with ramming too many volts in to the SoC and literally frying Ryzen 7000 CPU's?
Its been fixed since with BIOS updates.

With those BIOS AMD have imposed a hard limit of 1.3v, the maximum long term safe limit is 1.25v, so its fine but i wouldn't go any higher than that.

I'm still worried about your SoC pulling as much as 40 watts, no, it should not be pulling anything like that, at all ever. Also that's more than 60% of the total TDP of some of them, its quite obviously not normal.

 

[FredFlint]

You do know about Asus have had issues with ramming too many volts in to the SoC and literally frying Ryzen 7000 CPU's?
Its been fixed since with BIOS updates.

With those BIOS AMD have imposed a hard limit of 1.3v, the maximum long term safe limit is 1.25v, so its fine but i wouldn't go any higher than that.

I'm still worried about your SoC pulling as much as 40 watts, no, it should not be pulling anything like that, at all ever. Also that's more than 60% of the total TDP of some of them, its quite obviously not normal.

I am aware of the 1.3v SOC limit. It was not just ASUS as AMD sets the max limits, other boards had high SOC values too. I did not say the SOC was using 40W, I said the idle power, meaning the total package power at idle. I have never seen > ~35W idle but I don’t run the fastest RAM and have never gone > 1.25v on the SOC, others used > 1.35v before the limit was put in place. That said, ASUS did seem to be the worst.

 

[humbug]

I am aware of the 1.3v SOC limit. It was not just ASUS as AMD sets the max limits, other boards had high SOC values too. I did not say the SOC was using 40W, I said the idle power, meaning the total package power at idle. I have never seen > ~35W idle but I don’t run the fastest RAM and have never gone > 1.25v on the SOC, others used > 1.35v before the limit was put in place. That said, ASUS did seem to be the worst.

AMD's white paper said the SoC should not go above 1.25v for day to day client use, they didn't set a maximum voltage cap because they assume, wrongly obviously that motherboard vendors are capable of reading instructions.

They left it uncapped for extreme overclocking, which is a thing.

I said back then when this first came to light and a lot of people were trying to blame AMD for this, including Asus that we should not do that, because AMD's reaction would simply be to lock voltages down, because if you can't play nice we will take away your toys.

Asus and yes the one or two others who did this to a lesser extent are also to blame for not following AMD's instructions.

What i find really annoying is when Intel, or AMD or who ever leave these things unlocked to allow us to play and then when some 19 year old coding the BIOS at Asus can't be bothered to read AMD or Intel instructions get the blame for it..... NO!

 

[drakioned]

I saw this video and on the face of it, buying AMD seems like a bit of a no-brainer? Our summers are getting more miserable and I really want to start minimising the amount of heat that my PC puts out.

 

[ooStevo]

I might try OC my 5500. What should I aim for,or is it pointless? I have 3600 ram. It's not that I need to, just thought it might be fun tuning things. I miss overclocking. My best ever is an E6300 (1.86ghz) and clocking it to 3.8 GHz. I dialled it back to 3.5ghz and it ran perfectly fine for about 3 years,at which point the traces on the motherboard burned up. It was some Gigabyte mobo I think. Paid £60 for that CPU . Bargain really.

 

[humbug]

I saw this video and on the face of it, buying AMD seems like a bit of a no-brainer? Our summers are getting more miserable and I really want to start minimising the amount of heat that my PC puts out.

No its not a no-btainer, IMO there is more value in something like a 7800XT vs an RTX 4070, the rasterization, performance is a little higher on the 7800XT, its has a proper amount of VRam, 16GB vs 12GB, IMO its has a much better Driver UI and features and it overclocks much higher, its about 15% cheaper.
The 4070 has a little more RT performance overall and it uses about 20% less power.

Personally i think the 7800XT is a better GPU overall, and cheaper, but it depends on what you want, for example DLSS produces a slightly better image quality and the 4070 will put out a little less heat,

Then again FSR 3 Frame Gen works on any DX11 or DX12 game if it officially supports it or not, unlike DLSS frame gen.

Edit: you're talking about CPU's yes its a no-brainer....

 

[humbug]

The 7800X3D is the fastest gaming CPU on earth, its slightly cheaper than the 14700K and is incredibly efficient using half the power of the 14700K to produce more FPS.

Even the standard 7700X is just as good for gaming as the 14700K and about £80 cheaper with again much less power consumption,

 

[drakioned]

Edit: you're talking about CPU's yes its a no-brainer....

 

[Drumroll]

For gaming the 7800x3d is great, even more when you see how little effort is required to cool a CPU with such performance. But don’t get fooled thinking it is slow for other tasks. Isn’t. May not be the fastest, but is far from slow. And unless your main use of your PC benefits from more/faster cores, for the average user, is plenty for every task but smashes in games.
Intel may have some good performance, but compare it against the 7900 non X and again, just shows how much power is needed to deliver such performance.
Cooling 100ish W is one thing. Cooling 250-300W using warm air from a 300+W GPU is another. For the first few hours fine, for after a while, the higher temperature inside a room is noticeably and unpleasant.
Some people may have genuine reasons to choose Intel. But some won’t be so critical to the excessive consumption as they are when comparing how much more power AMD GPUs uses compared to Nvidia. I only hope AMD manages the same efficiency for next GPUs.

 

[G J]

I find these CPU consumption and efficiency debates weird in a way that some people that argue to death over it will happy have a GPU in their system that pulls 200+ watts. Though I understand that a CPU that uses less power allows people to more be flexable on the GPU front.

Intel 13900K & 13600K Temperature Myths BUSTED

There's been a lot of talk about how Intel 13th gen CPUs including the 13900K and 13600K run hot and need a lot of power. Well its time to bust some myths a...

www.youtube.com

As an example in the above video the 13900K will take an extra 50w (300w) from its stock 253w and hit the thermal limits using a 360 AIO all for an extra 150mhz or 3% extra performance.

I'm all for motherboards having set stock limits in place that have to be adhered too as some motherboards out of the box will just ignore guildlines and take all the power they can get and for what is a negligible increase in performance. Unfortunately winning a benchmark at any cost means that Intel and AMD get **** for motherboards ignoring limits and the user experience suffers.

 

[mattyfez]

undervolting is the new overclocking...
I can only speak from experince of Raptor Lake i5, so the i7/i9 13th/14th gen will be proportionatley hotter/more power hungry...

...but running my i5 at stock speeds with a negative v-core offset has yielded some very good results interms of peak temp, peak wattage and sustained boost speeds, without detriment to performance.