AMD Execution Thread [2024]
- Thread starter del42sa
- Start date
Worth noting that this is AVX 512 which is now full rate on Zen 5. AMD's stated geomean IPC increase over Zen 4 is 16%x.com
x.com
Can I quote another thread here because it's saying a lot to me about how AMD executes now?
This is absolute BS. Lying straight to everyone's faces and expecting it to slide? No, this is wrong!
This is absolute BS. Lying straight to everyone's faces and expecting it to slide? No, this is wrong!
That’s a dangerous game though. Intel still has significant mindshare. If they retake the lead for any length of time people will abandon AMD.
Its an extreme cherry picking yes and I certainly don't condone it but they aren't outright lying. And they're for literally 2 gen old AM4 CPUs which don't sell much anyway really. Intel's results with their extreme power profiles are perhaps worse, even leading to silicon degradation.
Not really, Ryzen has gained a significant reputation for performance, and in gaming in particular with their X3D line. And the longevity of the AM4 and AM5 platforms have also been popular with users. To add there are a lot of Intel 13th and 14th gen K series users who have been burned by Intel. So no one's going to abandon AMD, yet.
Intel faces a far more precarious decision on whether or not they'll keep using their fabs at this point than AMD choosing to cherrypick ...
Those two things aren’t related in any way.
Intel is so far behind in PPW, they won’t be worthwhile competition anytime soon. They are also very poor value with their constant socket changing shenanigans.
Intel's "significant mindshare" advantage has to be followed up by actually releasing genuinely competitive products which is easier said than done with their own current fabs because otherwise it's easy to give AMD the pass for their nasty plays ...
Not sure about that, it wasn't that long ago Intel were nowhere near AMD in graphics/iGPU perf and in the last couple of years they've blindsided AMD with their gains and are now competitive. If AMD underachieve for a couple of generations and Intel do well and fix a few significant flaws then gaps close far faster than you'd expect. Going by Intel marketing Skymont's become seriously competitive in PPA for example. Just like Intel 10 years ago before Zen except the gap is significantly smaller - don't take things for granted, always assume your competition will make significant progress and it'll be difficult to be caught offguard. It doesn't mean Intel/ARM competition will outpace AMD and it doesn't mean AMD can't continue to make significant gains, but you can't operate under the assumption your competition is incompetent because sooner or later it will bite you
Yes the premise is that Intel can produce a competitive part. AMD’s current advantage is significant but vulnerable if Intel get their act together. .
It'll be interesting to observe what each of their timelines in the near future will look like either way ...
Intel is only currently 'so far behind' in performance per watt cuz they're a generation behind in process technology compared to AMD, and they're having to push the hell out of the CPU's to ensure they're performance competitive.
Lunar Lake will probably beat AMD's mobile offerings in terms of efficiency. And Intel's own fabs are likely to catch up to the processes that AMD will generally be on(who dont use TSMC's latest and greatest), which should change the picture in terms of desktop competition. And for servers, I do think AMD's chiplet approach is very advantageous in terms of scaling up cores+performance with good power efficiency, but Intel has some coming answers to that as well on top of the other advancements they've been making.
AMD may rightfully feel like leaders in the CPU space at the moment, but it's not unrealistic to suggest that this could change quite quickly.
And as for Intel being 'very poor value' simply cuz they dont offer 3+ generations on the same socket, I think that's a massive exaggeration. Upgradeability on the same socket is certainly nice to have for those who like to upgrade their CPU every 2-3 years, but the majority are not going to upgrade their CPU for a good while.
Subtlesnake
Regular
While I expect Lunar Lake to close the gap, Meteor Lake was not super impressive on Intel 4.
I think you would find a good amount of those users would upgrade their CPU to a noticeably faster model if they didn't have to replace their motherboard and RAM as well. I've upgraded my CPU twice since 2017, starting with a Ryzen 1700 and now running a 5800X. I'm far from a niche PC gamer in this regard.
I would have kept my older CPUs a lot longer if I didn't have that path, as you mentioned happens. But I did have a path, and took advantage of that. And that's an extremely positive feature to me.
True, Meteor Lake did not impress at all and was late to market vs originally planned. Lunar Lake looks like it will be a competitive part, though it does have a node benefit with TSMC 3nm vs Strix Point which is still on 4nm. Though Lunar Lake is more a mid range part and the real competition for Strix will be Arrow Lake Mobile. Intel is definitely catching up though and Panther Lake will be on Intel 18A. AMD needs to move to TSMC 3nm/2nm fast or Intel will have a node advantage for sure.
The upgrade path has definitely been a plus point for many users. If you see the list of top selling CPUs on Amazon (or Mindfactory), AMD CPUs are far more popular. AMD has even promised support for AM5 till 2027+ which likely means Zen 6 as well. If you were to buy an Intel platform today, there is no further upgrade possible.
For you, yes. I think hardware enthusiasts who talk about this stuff online, much like gamers online in general, have a huge tendency to think they're more representative of the overall market than they really are, though.
You also seemed to have timed it well and got a bit lucky too. Remember the plan was originally NOT to support Zen 3 stuff on earlier AM4 motherboards. Plus the entire market isn't all buying into the earliest generation of a given platform, meaning the scope for upgrading is lessened. Most people will simply not want to upgrade to just a one architectural generation ahead CPU, given the mere 20-25% gains you'll typically get under good circumstances for the extra $250+ cost.
It's a nice option to have for sure, but it doesn't inherently make an Intel CPU 'very poor value' as the other person claimed. That's just a terribly hyperbolic claim.
Last edited:
Intel 4 is basically like what Intel 14nm was in the beginning. Ok, but nothing great. Took a year with 14nm+ before it really reached a decent level of maturity to show off what it could really do. Intel 3 should be that node.
I expect the same thing with 20A->18A. There's a reason Intel are selling Intel 3 and 18A as their customer nodes.
I believe AMD is planning on using TSMC 3nm for Zen 5c in server form, but overall, Intel likely wont have the deficits in process technology against AMD going forward anymore. And might even pull ahead at some point soon.
I'm otoh still holding on to my Skylake-X i9 ( though for sure by now with the significant single thread improvements i would have been happier with some mid range parts released 2 years ago or less )
I also recognize that there would be far less people like me (buying high end but rarely upgrade), than people like you. It would be interesting to know how much more people (relative to the dyi the market size, which we know is pretty small already) would change cpus often like this
And now that I'm thinking about it, for people who like upgrading cpu whenever it makes sense like that, choosing an AMD socket is a no brainer, provided they switch motherboards at the begining of a socket lifecycle. Being quite cost effective and getting a steady performance increase is nice
Similar threads
