AMD CPU's, An Open Letter to the Framework Team

Not that much is open source. AMD also is not a laptop manufacturer.

1 Like

I want an AMD Framework but I had a 32 core threadripper and my intel hexacore i7 single threaded performance was much better than my threadripper.

I write multithreaded software in Java, Rust and C, so I like having multiple cores and using multithreading as opposed to multiprocessing.

If you don’t care about performance per watt and care about overall performance, then you might choose Intel. I bought Intel twice lately. The CPUs are good and as a software engineer, the performance is acceptable.

The Ryzen is a good chip, but so is Intel. Intel was better than AMD for a long time and now it might be the other way around but it doesn’t mean that Intel is worthless. Try having perspective, it means life is easier.

I don’t understand the need to stick it to a company for not doing what you want. AMD is behind on laptop chips compared to Intel so that’s where the business is, it is what it is.

2 Likes

Well, generally speaking this is a laptop, and one with notoriously poor battery life compared to its peers, so…

Honestly, I have no idea who has the better chip, I just have a personal vow not to give Intel a single penny after the way they handled the VISA situation (and honestly the fact VISA even shipped at all, you should not be making this kind of accident if you are a $140 billion company where if you fail, identity theft of your customers is extremely easy), their decision to place a fab in Arizona of all places, the like five different unpatchable hardware exploits found in their CPUs in 2021 alone, making their backdoor diagnostic for sysadmins to get into headless boxes into a necessary component for their low power YouTube watcher chip to even boot… and yeah, AMD is barely better, but they’re at least marginally better as a stopgap while the ARM or I guess RISC-V board is being worked on.

I assume you mean not buying a FW (at least, new) until AMD arrives. The answer for me, though others will likely be different, is that I actually really want to buy one and the Intel CPU and wireless card are the only two blockers. The company it is being stuck to is Intel, and Framework is unfortunate collateral. Maybe I’ll get an 11th gen off someone for now and just get parts from Framework.

Could you tell me what is wrong with the wireless card? Is it the vPro?

Yes I was aware of the poor battery life when I decided to buy a Framework, I want to support the idea of long term reuseable repairable hardware and I tend to be sat on the sofa or in a coffee shop with my laptop, so I don’t worry about battery life. As long as I can find a socket!

I am thinking of long term performance of the laptop, that’s important to me. I understand that AMD has superior multicore performance, but single threaded performance is important too. I’m not sure who is superior, I think AMD is better for multiple core, multithreaded performance and Intel is superior to AMD in single core performance. But I suspect it depends on model and workload.

The Ryzen supposedly outclasses Intel’s lineup for single core performance but I’ve not seen the benchmark confirmed anywhere else. So I’m not sure if it is actually true. (It was a user submitted benchmark)

it’s intel in this case, the user in question refuses to give any money to intel, period.

I’m no fan of intel’s cpu division, I can’t stand them, but in my experience, intel nics, both wired and wireless, are the best on the market.

Zen Vulnerability:

Looks like it’s a blessing that we don’t have an AMD option yet.

https://www.amd.com/en/corporate/product-security/bulletin/amd-sb-1039

1 Like

imo that security threat is basically meaningless… It’s unlikely to become a big problem, and will likely become patched before long!

Although it’s just my opinion I don’t believe that these security bugs are that severe considering they only trace proccesses on a instruction level! A hacker would basically need physical access to the machine in order to do anything.

1 Like

According to this entry on Red Hat, they need local privileged access (not physical). But if they have local privileged access, you’re screwed anyway in some shape or form. Though this seems to be specific to co-location. (e.g. Say, in the cloud, Admins of one organisation, can get encryption key of another co-hosted organisation’s VM)

CVE-2021-46778- Red Hat Customer Portal

2 Likes

try looking at how many CVEs intel have had over the last 5 years.

1 Like

It’s not about the numbers of CVEs.

(More about the CVE’s impact, and mitigation impact…In this case, mitigation impacts performance…by disabling SMT?)

My point still stands. Intel CVE microcode patches have caused MASSIVE performance loss repeatedly over the last 5 years, and they have repeatedly tried to foist their performance killing kernel patches onto AMD machines as well despite AMD not being vulnerable to the flaws the patches were intended to fix. AMD having a flaw is small beans compared to intel at this point

1 Like

I can vouch for this. The Spectre-meltdown double whammy beat 30% of the performance out of some processors. One could say that CPUs from the mid-2000s shouldn’t be in use, but that’s a bad take because functional hardware is good hardware, and having a quad-core processor suddenly take a beating for no reason beyond a manufacturer design issue is plainly unfair to the customer.

Processor design flaws like this are unacceptable.

2 Likes

No. Your point was only about “how many CVEs intel have had over the last 5 years.” No mention of severity. Do not outsource the responsibility of communication clarity to others to infer.

Now this is the point you should have started with. And that, I would agree, had experienced with as well, with their patches…time and time again. Painful performance drops…multiple times.

There’s Hertzbleed as well back in June…both Intel and AMD.

my point followed directly on from your suggestion that its a good thing that we don’t have an AMD board because of a CVE that needs a mitigation that causes performance loss. Intel have had at least a dozen CVEs needing mitigations that cause performance loss over the last 5 years. You framed the discussion, I responded within that same frame.

over 60% on some workloads. Plus they’ve removed AVX512 instructions from some newer cpus via microcode update as well. That leaves a very bad taste for me.

2 Likes

Yeah, that AVX drop was also pretty awful. Nothing like getting a chip, getting software developed that’s optimized for a feature set, and then having that feature set withdrawn without any explaination beyond market differentiation.

It’s intentional, too, and has its roots as far back as the late 80s, when they released the SX series of processors. What Intel learned back then is that if a feature isn’t consistantly provided, it gets overlooked. In this case, they did the reverse: encourage development, and then took it away to force buyers who then relied on the feature into upgrading to devices that features the unplugged pieces. It’s cynical, and it’s a waste of developers time.

The point to a platform is that it’s consistent. Sure, premium features can exist in other areas, but Intel wouldn’t dare yank Virtualization features now that it’s a standard Windows feature (and necessary for security). This essentially creates fragmentation where it didn’t exist before. Intel is, in this case, their own enemy.

1 Like

Is the idea of AMD CPUs still being considered by the Framework Team? I’d like my next chipset to be from AMD for power consumption reasons, seeing as the 11th and 12th gen are power hungry. Are there no options for us more power-conscious? (I need longevity more than performance for my office-oriented tasks).

I see competitors have solutions, but I’d like my next purchase to be from the Framework Team.

Did we get any official word on whether it was ever considered?

4 Likes

lmao I was about to say that…

Intel chips can also be very power efficient, it just depends on what settings you tweak. It seems right now most of the power problems are from the software/firmware not being optimized on Framework’s side.
(Although AMD 6000 chips do have far better power to performance last time I checked)

2 Likes

I don’t think I’ve ever seen the Framework team officially mention that they were “considering” AMD CPUs, but I challenge this forum to prove me wrong. The closest thing I know of is that Linus Sebastian has tried multiple times to get a conversation started between Framework and AMD, but I don’t believe that has panned out either time.

3 Likes

Yeah, that’s the extent of my recollection as well.

2 Likes