Intel Alder and Raptor Lake support

It depends on whether you are mainly looking for privacy or security. Which of the two do you think is more important or are they equally important?

There are some interesting browsers that unfortunately won't be available on OpenBSD.

The above options mainly focus on privacy, ad blocking and tracker blocking.
What should work perfectly is that you use Chromium with uBlock Origin and Privacy Badger.

Also good for privacy is using Tor optimally in combination with a VPN. But for me Tor is really slow and annoying to use and many things don't work optimally if you set a higher security profile in Tor because many content is blocked.

Pale Moon does not contain suspicious privacy-invading software and extensions, it is run by a privacy-oriented non-profit organization and it has robust security.

If you want security, Firefox is fine.

Something that might be safe is the Nyxt browser. Since very few people use it, very few exploits will be developed for it.
The browser is written almost entirely in Common Lisp, and 97% of programmers who develop exploits know 0.000% of Common Lisp.
In the past, people have successfully installed Nyxt on FreeBSD.
You then get something similar to very high security through obscurity.


Should iridium
under OpenBSD thanks pledge and unveil not be safe enough ?
 
Should iridium
under OpenBSD thanks pledge and unveil not be safe enough ?
You have to decide for yourself. It seems to me that they are less secure than other browsers.
I have already listed two reasons why this browser may be unsafe.

Iridium sends "Do-Not-Track" header and uses its own User-agent which makes you stand out like a sore thumb. Also latest commit was 3 months ago so browser hasn't been updated since it. But there isn't anything more to say actually. Iridium was a good idea but seems the project is pretty dead.
Simply put... The lack of updates for Iridium makes it a big security issue. To keep things clear, Iridium's lack of updates does not come from its removal of "automatic updates." The lack of updates stems from the developers simply not updating and implementing the newer versions of Chromium's source code. This is critical seeing as many browser exploits are geared towards Chromium - it's the most popular browser engine used today and therefore is the primary focus of many malicious actors.
It's currently 6 Chromium releases behind Chrome's stable channel and missing the 53 security fixes included in those releases.
Given that Chrome (and Google in general) has possibly the best defensive security team in the world, it's hard for me to take these security-oriented forks too seriously. Indeed, the last "secure Chromium fork" I heard about, WhiteHat Aviator, turned out to introduce a bunch of new vulnerabilities:
Even if the fork doesn't add bugs, you are now relying on the fork's maintainer to push security updates. Will they be as good at this as Chrome's team? This is unfair, of course: no startup or small project is ever going to have Chrome's resources. But when it comes to security, speed of updates really does matter.
Iridium's website makes a selling point of the fact that Chrome contacts Google servers, but fails to mention that it may phone home to servers of its own. Having to search the source to discover that large a privacy difference isn't proper disclosure to my mind.
I don't claim it's necessarily malicious. The explanation given in the git log makes sense, though there are better ways to accomplish the stated goal, such as replacing Google's tracking URLs with invalid ones. But leaving those changes in public releases of software supposed to increase privacy seems negligent.

I don't trust them. They have not a single contact information and no imprint on their website. Which is in fact illegal as their website ist hosted in Germany and uses a german top-level-domain.
They advertise their product as ”a secure browser“ without making any significant changes under the hood. As ”unicornporn“ said: ”privacy“ != ”security“. Especially when you replace one villain by another.

Wow, these guys seem to be using an ancient and vulnerable version of cgit.

Iridium is just another software package from entitled developers which are too bored for giving educated responses to reasonable user's questions. For example:


One thing I find strange about Iridium is that it advertises as:
"A Browser Securing Your Privacy"
But yet when you fire it up all kinds of settings changes need to be made to make it secure.
Why would then not make it secure out of the box? That is their shtick.
Mozilla does the same thing but we know they are driven by corporate sponsors.

Latest commits: 2 weeks ago

Latest commits: 1 hour ago
 
I will repost my question unless an admin can move these posts to another topic where they belong :)

Anybody using the iGPU of their Alder / Raptor Lake CPU? I want to use it for transcoding so was wondering if anyone is using it (with drm-515-kmod, only supported on CURRENT right now).
 
Given that even 12th gen (alder lake) GPU doesn't work right now with 14-BETA3 (loading i915kms from drm-515-kmod just locks up the system), I doubt there will be working raptor lake support anytime soon.
 
So Voltaire did you receive your 13th gen CPU? Is it working well for you? :)
Also, anybody successfully using the iGPU of his 13th gen CPU?
I may be going to buy a new CPU today or tomorrow but I'm actually thinking about the i5-12600KF.
This CPU is currently the best option for me, so I might not end up choosing Raptor Lake after all, but it is largely the exact same architecture as Alder Lake.
 
I may be going to buy a new CPU today or tomorrow but I'm actually thinking about the i5-12600KF.
This CPU is currently the best option for me, so I might not end up choosing Raptor Lake after all, but it is largely the exact same architecture as Alder Lake.
Thanks for letting me know. So perhaps I might be the first to report something on Raptor Lake here on the forum in the future :D, probably CPU will work just fine. For the iGPU I have found out drm-5.17 is needed, which has not landed yet but there is a WIP.
 
Thanks for letting me know. So perhaps I might be the first to report something on Raptor Lake here on the forum in the future :D, probably CPU will work just fine. For the iGPU I have found out drm-5.17 is needed, which has not landed yet but there is a WIP.
I think the underlying architecture is 99.9% the same for Alder Lake and Raptor Lake.

We've known for a while that Intel would reuse its Alder Lake CPU architecture with some of its latest 13th Gen non-K Raptor Lake processors to improve yields and reduce waste. In the case of the i5-13400F, Intel accomplished this strategy by re-purposing i9-12900/12900K dies into i5-13400 CPUs by disabling some of the P-cores and the CPU cache.
HWCooling says that virtually all retail i5-13400Fs are the Alder Lake version, while the Raptor Lake models are found in the OEM market.
Both flavors of the i5-13400F were tested in several CPU benchmarks, including Cinebench R23, FLAC, and F1 2020. The chips were neck and neck with each other in nearly all benchmark tests, with a maximum performance delta of 5%. In Cinebench R23, for example, the Raptor Lake i5-13400F scored 16,131 points in the multi-core test, while the Alder Lake version scored 16,038 points — a 0.6% difference in score.


I ordered the Intel i5-12600KF yesterday and I should get it on Thursday or Friday.
Will it work fine on FreeBSD 13.2 or is it worth upgrading to version 14?
 
I think the underlying architecture is 99.9% the same for Alder Lake and Raptor Lake.

We've known for a while that Intel would reuse its Alder Lake CPU architecture with some of its latest 13th Gen non-K Raptor Lake processors to improve yields and reduce waste. In the case of the i5-13400F, Intel accomplished this strategy by re-purposing i9-12900/12900K dies into i5-13400 CPUs by disabling some of the P-cores and the CPU cache.
HWCooling says that virtually all retail i5-13400Fs are the Alder Lake version, while the Raptor Lake models are found in the OEM market.
Both flavors of the i5-13400F were tested in several CPU benchmarks, including Cinebench R23, FLAC, and F1 2020. The chips were neck and neck with each other in nearly all benchmark tests, with a maximum performance delta of 5%. In Cinebench R23, for example, the Raptor Lake i5-13400F scored 16,131 points in the multi-core test, while the Alder Lake version scored 16,038 points — a 0.6% difference in score.


I ordered the Intel i5-12600KF yesterday and I should get it on Thursday or Friday.
Will it work fine on FreeBSD 13.2 or is it worth upgrading to version 14?
It should work just fine. Except for the graphics. I do not think it is supported yet. My 12700K has been working fine on all cores since 13.2 came out.
 
It should work just fine. Except for the graphics. I do not think it is supported yet. My 12700K has been working fine on all cores since 13.2 came out.
The Intel i5-12600KF does not have integrated graphics. When you see the word K at the end of an Intel CPU's name, it means it has overclocking support.
The F means the CPU does not have integrated graphics.

I'm going to keep using a Nvidia GPU, and in a few months I might buy a new AMD GPU so that this system is (completely) up-to-date.

I would like to ask, what result do you get with the 12700K in WebXPRT 4 if you use FreeBSD 13.2 and Firefox 118?
 
The Intel i5-12600KF does not have integrated graphics. When you see the word K at the end of an Intel CPU's name, it means it has overclocking support.
The F means the CPU does not have integrated graphics.

I'm going to keep using a Nvidia GPU, and in a few months I might buy a new AMD GPU so that this system is (completely) up-to-date.

I would like to ask, what result do you get with the 12700K in WebXPRT 4 if you use FreeBSD 13.2 and Firefox 118?
I can't answer for that. I am on FreeBSD14-STABLE and have undervolted my cpu and overclocked it to 5.4mhz on the P-cores (E-cores (4.2). So my system is not a stock one. But running cool (under 60 degrees celcius at all times) and stable. But I have run the test on it. Hope it is of some use for you :)
WebXPRT-4-logo.svg


Test complete!

Your Score:

223




WebXPRT 4 (v3.73)

Test ID: 178000

2023-10-04 16:09:14

Browser: Firefox 118


Confidence interval: +/- 4
What's this?



WorkloadmsVariation(%)


Photo Enhancement
380+/- 0.92


Organize Album using AI
1788+/- 4.37


Stock Option Pricing
98+/- 1.18


Encrypt Notes and OCR Scan
1174+/- 0.7


Sales Graphs
308+/- 4.29


Online Homework
1317+/- 3.19
 
I can't answer for that. I am on FreeBSD14-STABLE and have undervolted my cpu and overclocked it to 5.4mhz on the P-cores (E-cores (4.2). So my system is not a stock one. But running cool (under 60 degrees celcius at all times) and stable. But I have run the test on it. Hope it is of some use for you :)
Test complete!

Your Score:

223
It is very useful :) I currently have an Intel i3-3240 and it scores around 128 with Firefox on FreeBSD, which is about exactly as fast as Clear Linux and faster than I can score with windows or other Linux systems.

My impression is that Firefox scales nicely even with faster CPUs as the power of the CPU increases, e.g. with the R5 PRO 3400G I get a score of exactly 161 with Firefox and Clear Linux.

I had thought you would score around 330 based on these (old) results:
2023-10-04-174338_1920x1080_scrot.png

What you sometimes see is that overclocking can have a negative impact. It could also be that you're using very slow RAM.
Maybe you are using Firefox extensions that have a negative impact. Or maybe during the test you had your PC doing other tasks.
Either FreeBSD is still very poorly optimized for Alder Lake.

If you have time you can see if you score higher or lower with Chromium.

In a few days I will also have an Alder Lake CPU and then I will be able to see exactly whether FreeBSD already performs decently with these CPUs or not.
 
It is very useful :) I currently have an Intel i3-3240 and it scores around 128 with Firefox on FreeBSD, which is about exactly as fast as Clear Linux and faster than I can score with windows or other Linux systems.

My impression is that Firefox scales nicely even with faster CPUs as the power of the CPU increases, e.g. with the R5 PRO 3400G I get a score of exactly 161 with Firefox and Clear Linux.

I had thought you would score around 330 based on these (old) results:
View attachment 17059
What you sometimes see is that overclocking can have a negative impact. It could also be that you're using very slow RAM.
Maybe you are using Firefox extensions that have a negative impact. Or maybe during the test you had your PC doing other tasks.
Either FreeBSD is still very poorly optimized for Alder Lake.

If you have time you can see if you score higher or lower with Chromium.

In a few days I will also have an Alder Lake CPU and then I will be able to see exactly whether FreeBSD already performs decently with these CPUs or not.
I had a some more tabs open in the browser. Around 15. So the browser did not have the machine alone. When compiling big packages. I have run the test again. A similar result 227. My system is compiled and optimized for alderlake (both sytem and packages. But FreeBSD is not optimized for balancing P and E cores. Windows and Linux are. But I am sure it will improve over time ;)
 
I had a some more tabs open in the browser. Around 15. So the browser did not have the machine alone. When compiling big packages. I have run the test again. A similar result 227. My system is compiled and optimized for alderlake (both sytem and packages. But FreeBSD is not optimized for balancing P and E cores. Windows and Linux are. But I am sure it will improve over time ;)

Windows and Linux are presumably very different in their Alder Lake support. Windows is supposed to use the "thread director", about which very little information is available. The Linux scheduler is contributed by Intel, but I don't see anything super smart in there. They have some recognition of specific applications which IMHO is cheating.

I suspect that the Windows support is a lot more primitive than we have been led to believe, too. Could be as simple as recognizing games and fast-core those, mixed with a lot of hand coding for specifically known applications in the scheduler.

I find it to be a hard mental exercise to make scheduling decisions in a heterogenous CPU when the workload is server-like and not dominated by a recognizable desktop applications. How do you know whether a certain thing should save power or not?

The first useful step would be to run some of Phoronix' benchmarks on Linux and FreeBSD.
 
What I get from this is that if/when FreeBSD gets its scheduling optimized for hybrid architectures, it should be depending on instructions used. If all cores are utilized use whatever it wants to. But if P-cores are vacant, move processes that uses VNNI or AVX2 (maybe AVX as well ???) to those from the E-cores. Powersaving might benefit from moving some processes (without VNNI/AVX2) not fully using P-cores to the E-cores.

Just my 2 cents without knowing to much about it.
 
What I get from this is that if/when FreeBSD gets its scheduling optimized for hybrid architectures, it should be depending on instructions used. If all cores are utilized use whatever it wants to. But if P-cores are vacant, move processes that uses VNNI or AVX2 (maybe AVX as well ???) to those from the E-cores. Powersaving might benefit from moving some processes (without VNNI/AVX2) not fully using P-cores to the E-cores.

Just my 2 cents without knowing to much about it.

Well, this is what Windows does.

I am not sure this is a good thing. The efficient cores have vector instructions, too. Let's say I am watching a video on a laptop on battery. If an e-core has enough CPU power to sustain the stream I want an e-core to decode the video.

Likewise, let's say I want to re-encode an entire video library. I don't care how long it takes but I want to use as little total power consumption as possible. Wouldn't that be better on an E-core?

This is what I mean by the existing mechanisms being overly simplistic.
 
Well, this is what Windows does.

I am not sure this is a good thing. The efficient cores have vector instructions, too. Let's say I am watching a video on a laptop on battery. If an e-core has enough CPU power to sustain the stream I want an e-core to decode the video.

Likewise, let's say I want to re-encode an entire video library. I don't care how long it takes but I want to use as little total power consumption as possible. Wouldn't that be better on an E-core?

This is what I mean by the existing mechanisms being overly simplistic.
For me I would choose speed over efficiency. Otherwise I could downclock my cpu, or disable turbo mode. But efficeincy/energy consumption I think the scheduling should be going both ways, so processes not utilizing a lot of the P-cores capacity should be using E-cores. But it has noting to do with foreground/background processes as stated on windows. It is just about optimizing the use of CPU resources. And the current state on FreeBSD is just fine. ;)
 
For me I would choose speed over efficiency. Otherwise I could downclock my cpu, or disable turbo mode. But efficeincy/energy consumption I think the scheduling should be going both ways, so processes not utilizing a lot of the P-cores capacity should be using E-cores. But it has noting to do with foreground/background processes as stated on windows. It is just about optimizing the use of CPU resources. And the current state on FreeBSD is just fine. ;)

I think it is quite hard to determine from the scheduler whether a process will "utilize a lot of P-cores". What about a single-thread, 100% CPU taking process like a compiler compiling a C file? And those have no vector instructions either.

This is all highly non-trivial. As far as I can see these signals exist:
  • Uses vector instructions
  • Uses 3D (probably the case on Windows)
  • Hand coded list of known applications

That's not much to go on with to recognize a compiler from an Apache process.
 
As said above. The current situation is just fine. And it is all about efficiency and power usage. I am just a happy FreeBSD user. I did just politely (I hope so :) ) said my opinion what I could read out of the pdf you attached about the situation on Linux.
 
I had a some more tabs open in the browser. Around 15. So the browser did not have the machine alone. When compiling big packages. I have run the test again. A similar result 227.
330/227=1.454
Maybe the efficient cores are just not being used. But even then the i5-12400F currently achieves +- 255 in this test on windows.
This is a 6-core CPU while your CPU has 8 performance cores.

If someone tests an i5-12400F then we can know if it is just the efficient cores that bring performance down, or if there are other performance issues with Alder Lake on FreeBSD.

It does seem to me that there is a need for FreeBSD to investigate how this new "hybrid" architecture can achieve optimal performance.

AMD is soon planning to go down this type of road as well:
 
AMD already has "uneven" CPUs. For example, in the 7950x3d half of the cores have access to the large L3 cache, while the other half can clock higher. Good luck being smart about scheduling here.
 
Back
Top