Intel Core i9-13900K "Raptor Lake" Tested Again, 30% Faster Than Predecessor in Cinebench R23

[ratirt]

Several times in the span of 3 years that I have worked for my previous employer, I have had a firmware update (BIOS) as a part of other ubuntu updates being run at some point.
It does warn to keep the laptop powered by cable while the operation is underway just to be safe, even though the battery was mostly full.

My point being, that BIOS updates are so easily done, that they were in my case done automatically for a Dell ubuntu mobile workstation, whichever model it was. I'm pretty sure it was like that for the whole family of those products.
Many motherboards let you do a bios upgrade without even having a processor enabled, so it's pretty easy to do.

Download and click that is it. All you need to know is to download the proper file but if the file is wrong, nowadays it will tell you it is wrong. Pretty hard to mess things up.

 

[mahirzukic2]

Download and click that is it. All you need to know is to download the proper file but if the file is wrong, nowadays it will tell you it is wrong. Pretty hard to mess things up.

Point is, I had to do nothing. Not even download (go and look for the download) and click.
I got a popup saying there's new updates. Do you want to update now? Clicked yes and off it went.
BIOS update was done automatically on the next reboot.

 

[95Viper]

Get back on the topic, please... and that is: "Intel Core i9-13900K "Raptor Lake" Tested Again, 30% Faster Than Predecessor in Cinebench R23"

 

[Gica]

Because consumption is always discussed, let's do a case study based on this topic.
Gentlemen, from what I can see, an Intel processor can run a lot of applications and the consumption of the processor is not even close to that of an AMD in idle. I ran youtube 8K decoding 60fps and the processor consumes current that an AMD owner wouldn't even dream of having it idle. It's like saying that my car consumes a lot of gas at 300 km/h, but yours consumes as much as a tractor in the city.
P.S. Today, in 30 minutes of net browsing (forums and war news), the average consumption of the processor is 2.7 W. Not having a video card dedicated to this system, his big problem was to somehow make sure that the entire consumption does not fall below 25W because it enters the power supply in protection and ... system shutdown.
Let's leave consumption and focus on performance. Are you saddened by the rumors about Raptor performance?

 

[JustBenching]

Because consumption is always discussed, let's do a case study based on this topic.
Gentlemen, from what I can see, an Intel processor can run a lot of applications and the consumption of the processor is not even close to that of an AMD in idle. I ran youtube 8K decoding 60fps and the processor consumes current that an AMD owner wouldn't even dream of having it idle. It's like saying that my car consumes a lot of gas at 300 km/h, but yours consumes as much as a tractor in the city.
P.S. Today, in 30 minutes of net browsing (forums and war news), the average consumption of the processor is 2.7 W. Not having a video card dedicated to this system, his big problem was to somehow make sure that the entire consumption does not fall below 25W because it enters the power supply in protection and ... system shutdown.
Let's leave consumption and focus on performance. Are you saddened by the rumors about Raptor performance?

Yeah, the idle and light load consumption on zen 2 and zen 3 is insane, but that's because of the iod.

 

[ratirt]

Because consumption is always discussed, let's do a case study based on this topic.
Gentlemen, from what I can see, an Intel processor can run a lot of applications and the consumption of the processor is not even close to that of an AMD in idle. I ran youtube 8K decoding 60fps and the processor consumes current that an AMD owner wouldn't even dream of having it idle. It's like saying that my car consumes a lot of gas at 300 km/h, but yours consumes as much as a tractor in the city.
P.S. Today, in 30 minutes of net browsing (forums and war news), the average consumption of the processor is 2.7 W. Not having a video card dedicated to this system, his big problem was to somehow make sure that the entire consumption does not fall below 25W because it enters the power supply in protection and ... system shutdown.
Let's leave consumption and focus on performance. Are you saddened by the rumors about Raptor performance?

I don't know what Intel CPU you have but according to this reviews, 12600k idles with 55watts whole system and that is what 5950x whole system idles at. It is the whole system though but still it is roughly the same. You can assume these two processors idle with same or very similar power?
Didn't know idle power is suddenly very important. If it consumes 50watts or 35watts at idle, is that something that should reach the headlines? I doubt it. If it hit 100w idle you might have something to talk about. Also, it is hard for me to believe you switch on your PC and do nothing. Idle power is a nice metric but it is not that important as max power needed for the stock settings for instance.

 

[JustBenching]

I don't know what Intel CPU you have but according to this reviews, 12600k idles with 55watts whole system and that is what 5950x whole system idles at. It is the whole system though but still it is roughly the same. You can assume these two processors idle with same or very similar power?
Didn't know idle power is suddenly very important. If it consumes 50watts or 35watts at idle, is that something that should reach the headlines? I doubt it. If it hit 100w idle you might have something to talk about. Also, it is hard for me to believe you switch on your PC and do nothing. Idle power is a nice metric but it is not that important as max power needed for the stock settings for instance.

You can't look at system power consumption and conclude anything. One motherboard consuming 5-10w more than other and everything is out of the window. Especially considering he is testing an asus hero that has a screen and full of leds compared to a barebones naked for the am4 system, the calculations can be way off. I know for a fact I can browse the web / watch youtube at around 10w on my 12900k while the 3700x is usually at 30 with peaks at 50 watts, which is insane.

 

[ratirt]

You can't look at system power consumption and conclude anything. One motherboard consuming 5-10w more than other and everything is out of the window. Especially considering he is testing an asus hero that has a screen and full of leds compared to a barebones naked for the am4 system, the calculations can be way off. I know for a fact I can browse the web / watch youtube at around 10w on my 12900k while the 3700x is usually at 30 with peaks at 50 watts, which is insane.

10w just the CPU or the whole system? I need to check what my CPU uses while idle. Just the CPU cause I assume 10w is just the CPU?
Different motherboard more features uses more power. Not shock here.

 

[Solid State Brain]

From personal testing, with a minimal configuration (2 memory sticks, 1 NVMe SSD, CPU, mouse+keyboard), all C-states active (for both core and package), no discrete GPU and display off, the idle power consumption of an Alder Lake desktop system can get in the 15–20W range and possibly even less with a good low-power PSU.

Depending on the motherboard, several of these power savings might be disabled by default, and just having a discrete GPU installed and active (even if just idling) will cause idle power consumption to increase by a significant margin (percentage-wise) unless it can be completely switched off.

 

[JustBenching]

10w just the CPU or the whole system? I need to check what my CPU uses while idle. Just the CPU cause I assume 10w is just the CPU?
Different motherboard more features uses more power. Not shock here.

Yes just the cpu. Alderlake idles at less than 2watts.

 

[Valantar]

Because consumption is always discussed, let's do a case study based on this topic.
Gentlemen, from what I can see, an Intel processor can run a lot of applications and the consumption of the processor is not even close to that of an AMD in idle. I ran youtube 8K decoding 60fps and the processor consumes current that an AMD owner wouldn't even dream of having it idle. It's like saying that my car consumes a lot of gas at 300 km/h, but yours consumes as much as a tractor in the city.
P.S. Today, in 30 minutes of net browsing (forums and war news), the average consumption of the processor is 2.7 W. Not having a video card dedicated to this system, his big problem was to somehow make sure that the entire consumption does not fall below 25W because it enters the power supply in protection and ... system shutdown.
Let's leave consumption and focus on performance. Are you saddened by the rumors about Raptor performance?

I mean, you're running hardware accelerated video decode which leaves the CPU essentially idle and mostly power gated, so ... this is hardly surprising? Also: software power reporting is notably untrustworthy. It's not necessarily wrong, but it's quite opaque in what is actually being measured and how this reflects the overall picture. For reliable CPU power measurements, you need to measure power draw through the EPS12V cable.

Yeah, the idle and light load consumption on zen 2 and zen 3 is insane, but that's because of the iod.

No, it's because of through-package IF. Ultra high bandwidth interconnects like that going through a PCB require quite a lot of power compared to similar interconnects in silicon. There's a reason why TR Pro and EPYC these days sees 100+W power draws for uncore - with 8 CCDs and a bunch of IF links over relatively long distances, that requires a lot of power.

You can't look at system power consumption and conclude anything. One motherboard consuming 5-10w more than other and everything is out of the window. Especially considering he is testing an asus hero that has a screen and full of leds compared to a barebones naked for the am4 system, the calculations can be way off. I know for a fact I can browse the web / watch youtube at around 10w on my 12900k while the 3700x is usually at 30 with peaks at 50 watts, which is insane.

I agree that full system power consumption measurements are very problematic, but idle power is actually one metric in which they are useful - after all, idle is the state in which all the various draws of the system add up most visibly, which highlights platform configuration differences. This also means you idelly need to test across a range of motherboards - it's quite well documented that low-end motherboards with less stuff on them consume less power, after all - but as long as the boards being compared across vendors are comparable in their featuresets (on-board hardware features such as controllers, VRMs, etc.), then the numbers are roughly comparable for low-load system power draw. And they provide some interesting perspective: if one CPU is reporting 3W package power and the other 30W, but both are measuring and averaging the same power for the full system at the wall, where are the differences stemming from? Is it down to other hardware, software power reporting inaccuracies, or something else?

 

[JustBenching]

No, it's because of through-package IF. Ultra high bandwidth interconnects like that going through a PCB require quite a lot of power compared to similar interconnects in silicon. There's a reason why TR Pro and EPYC these days sees 100+W power draws for uncore - with 8 CCDs and a bunch of IF links over relatively long distances, that requires a lot of power.

I agree that full system power consumption measurements are very problematic, but idle power is actually one metric in which they are useful - after all, idle is the state in which all the various draws of the system add up most visibly, which highlights platform configuration differences. This also means you idelly need to test across a range of motherboards - it's quite well documented that low-end motherboards with less stuff on them consume less power, after all - but as long as the boards being compared across vendors are comparable in their featuresets (on-board hardware features such as controllers, VRMs, etc.), then the numbers are roughly comparable for low-load system power draw. And they provide some interesting perspective: if one CPU is reporting 3W package power and the other 30W, but both are measuring and averaging the same power for the full system at the wall, where are the differences stemming from? Is it down to other hardware, software power reporting inaccuracies, or something else?

Yeah, I've seen motherboard tests where for the same platform and same CPU there is up to 50w difference. That's under load ofcourse, and you could argue some of it is down to voltage differences, but 50w is a LOT to be attributed down to just voltage difference.

On another note, it's kinda sad that some apps merely by being open - doing absolutely nothing, double or even triple idle wattage. Namely steam / icue / bnet and other similar stuff, if I have all of them open at the same time idle wattage skyrockets from less than 2w to 10 or even 15 on alderlake.

 

[tps3443]

This means that it's still going to be around 30% slower than 7950X

^ Wow! How wrong was this guy. 13900K matches the 7950X in multithreaded and outpaces the 7950X in single threaded.


The 12900KS isn’t even 30% slower than a 7950X!