AMD's desktop CPU share passes Intel's for first time in 15 years

This is testing passmark share. Not a market share at all.
I have to agree with you: while the data from Passmark is certainly a good indicator, it would be nice to see information from other sources too, or, if there is an independent industry-wide survey covering it all, that would be best (I know, I should keep dreaming :) )
I am genuinely interested to see how AMD is progressing (I'm rooting for them, yes), and how the rivals react.
 
These reports are incomplete. I personally procured and manage 11 Epyc servers (gen 1 and gen 2) at work and Passmark never touched them. Windows never touched them either. You would have to use HP sales data or Red Hat activations or something to know they're in use. Epyc is best at heavily multithreaded things, virtualization and the like, so I imagine there's a lot of deployed AMD servers that never get picked up by these reports. And yes, if you're curious, the servers have been very well received.
Yeah, Passmark is mostly about home computers. The problem is that it's literally impossible to know what everyone is using. Things like Passmark and Steam Surveys, while flawed, are the best tools that we have available. They do at least show an objective change in the trends and those changes are in the right direction. That much cannot be denied.
 
People are funny. They say they want better graphics and go for higher resolutions than ever and praised ray tracing like godsend. And then they also praised to the skies a new high tech graphics downgrading technology DLSS.
Well, DLSS v1.x was terrible but DLSS v2.x is apparently amazing. Steve Burke himself said so (even if nVidia was holding him up with a heat gun...lol).
 
But this statement on AMD's takeover of the desktop market represents only Passmark's stats. Shouldn't other indicators, such as actual sales, be considered?
 
But this statement on AMD's takeover of the desktop market represents only Passmark's stats. Shouldn't other indicators, such as actual sales, be considered?
Sure, but those records aren't exactly public domain so all we have to go on are things like Passmark records and Steam Surveys. Are they perfect? Of course not, but they're what we've got and they do show a definite trend. If you want to look at something to do with sales, in the top 20 of Amazon's CPU bestseller list, AMD occupies spots 1, 2 ,3, 4, 6, 7, 8, 9, 11, 12, 18 and 20.
Amazon Best Sellers in Computer CPU Processors
Again, not perfect, but definitely impactful. Of Amazon's 20 best-sellng CPUs, 12 are AMD and they completely dominate Intel with 8 of their CPUs in the top ten and occupying the top-4. It doesn't prove that AMD is outselling Intel overall (this is only desktop CPUs sold by themselves) but it definitely proves a trend. In this instance where Intel used to absolutely dominate AMD, the opposite is now true. This much cannot be argued.
 
Last edited:
I feel that Passmark and Steam user stats are going to be heavily weighted in favor of gamers and enthusiasts - people who build their own rigs.
 
Corporate cares about AMD or Intel if someone on corporation cares and has high enough rank.

Timed delivery and Intel. Yeah. Basically put, Intel almost always have more serious problems delivering stuff to OEM's than AMD has. Past 3 years Intel has been very very badly capacity constrained because 10nm tech is late about 3 years. Basically put, overall Intel has had much more delivery problems to OEM's than AMD has when looking this century overall. Intel delivers on time is simply myth. Talking about servers, same thing. AMD delivers, Intel don't https://www.tomshardware.com/news/u...percomputer-delayed-due-to-intels-7nm-setback

I highly doubt TCO is higher on AMD products than on Intel products.

Your confusing headlines about intel having capacity issues and truely grasping the *scale* of Intel's capacity. AMD single sources from TSMC and TSMC can't produce nearly enough to meet server market demand. Notice how AMD is having hige problems meeting *retail* demands for their CPUs right now? And lets not forget that Apple bought the entire first output of TSMCs 5nm (apple and amd are likely why nvidia is sourcing their retail products from Samsung).

TSMC (and thus amd) do not have the capacity to absorb intels server market. Not yet. Even the ARM competitors can't threaten huge swathes of that market yet for the same reasons. But time *is* running out. Intel better pick up the pace and quick. TSMC is building a fab in the US (another issue with TSMC is nearly the entirety of its fab capacity is next door to China who kept threatening to take it over). And Samsung is quickly moving to smaller nodes and should be at parity with TSMC in 2021 or 2022 and possibly pass them in 2023 or 2024.

Its an exciting time in the business.
 
Your confusing headlines about intel having capacity issues and truely grasping the *scale* of Intel's capacity. AMD single sources from TSMC and TSMC can't produce nearly enough to meet server market demand. Notice how AMD is having hige problems meeting *retail* demands for their CPUs right now? And lets not forget that Apple bought the entire first output of TSMCs 5nm (apple and amd are likely why nvidia is sourcing their retail products from Samsung).

Intel had serious supply issues because 10nm didn't work and 14nm chips were larger. 14nm demand was so high that not-so-high volume products (like OEM laptop chips) had several months wait time.

AMD does not seem to have problems on retail market. Ryzen 5xxx series is sold out because of scalpers but 3000-series is available very well. What this has to do with "server chip shortage" is confusing. AMD has somewhere around half of retail market but still just few percentage of server market. Zen 3 server chips aren't even announced yet. So far there is not single evidence of AMD's "server chip shortage", they just don't sell at all. Stupid buyers, that's only explanation. AMD does not need 5nm for long time, even Zen 2 server chips are better than Intel will have in two years.

TSMC (and thus amd) do not have the capacity to absorb intels server market. Not yet. Even the ARM competitors can't threaten huge swathes of that market yet for the same reasons. But time *is* running out. Intel better pick up the pace and quick. TSMC is building a fab in the US (another issue with TSMC is nearly the entirety of its fab capacity is next door to China who kept threatening to take it over). And Samsung is quickly moving to smaller nodes and should be at parity with TSMC in 2021 or 2022 and possibly pass them in 2023 or 2024.

Its an exciting time in the business.

"To absorb" and get "more than few percent" are different things. Again, AMD has just few percent of server market (Intel has over 90) despite AMD is much better on servers than on desktop (compared to Intel). AMD's market share should be much higher than just few percent. It isn't, buyers are to blame.

I doubt Samsung is trying to compete with TMSC as Samsung uses most of it's capacity for own chips. Unlike TMSC that only does what others want. I expect that Samsung will stick with lower performance processes leaving Intel and TSMC competing on high performance processes.
 
Well Done AMD this has been an uphill battle. And I'm not surprised at the server market share that is a much harder nut to crack. Those servers usually are tied to multi year support contracts. And companies don't just switch vendors even if one product is better than the other, its a lengthy process.

I would have liked to see a historical graph of the server space share. At one time AMD was making great strides in this area from what I recall. I think that was short lived too, and never eclipsed intel entirely. It'd be nice to see it though. But the reason is not the contract thing, but rather the system management, monitoring and security Intel puts in their chips that works with supporting hardware for server room purposes (like a true "lights out" environment.} AMD only made good movement when they had better math abilities than Intel years ago.
 
Back