Why don't AMD fix their single core performance?

Why don't AMD fix their single core performance?

Other urls found in this thread:

agner.org/optimize/blog/read.php?i=49#829
dtu.dk/english/service/phonebook/person?id=78696&tab=2&qt=dtupublicationquery
agner.org/optimize/blog/read.php?i=49#49
twitter.com/SFWRedditVideos

Why doesn't your mom fix her single porn performance?

...

People dont get one thing.
We got single core performance so high that hypertreading was introduced to make a fucking use of all that power and rather spread that performance between more threads.

single core is bretty gud when you compare clock-to-clock performance from jewtels offerings.

agner.org/optimize/blog/read.php?i=49#829

Why hasn't Intel improved its single thread performance for 4 years?
Do you seriously think Intel will be able to hide behind higher clockspeeds on low core parts for more than a year?

Just wait for graphene, Intel will once again take the lead with 10GHz

I'm waiting for their unobtanium based CPUs. Gonna hit 100 THz on air.

Wow, this is brutal.

If Agner fucking Fog tells you something like this, Intel is truly damn fucked.

FTL Quantum cores will be able to compute into the future and obtain answers even before they were queried

>agner.org/optimize/blog/read.php?i=49#829
>The single thread IPC is higher than that of anything intel
is this guy retarded or just ignorant?

Neither: dtu.dk/english/service/phonebook/person?id=78696&tab=2&qt=dtupublicationquery

Insanely smart and highly trusted in the industry for his words due to long history of reliable contribution to the computing industry.

because from bulldozer until very recently, AMD was banking on every application becoming extremely multi-threaded *and* calling on the CPU to handle that parallel workload. Something which hasn't....really happened, sure there's a use for more than 4c/8t in workloads like video production, but we're also seeing a lot of the workloads that would require such parallelization being offloaded to the GPU more and more, since those are already vastly more parallel. A trend that'll probably continue, leading to 8c/16t being the sweet spot for a while to come.

I'm still expecting 6c/12t to be the new standard.

>He doesn't know about Agner "Jew Wrangler" Fog

agner.org/optimize/blog/read.php?i=49#49

Of course it will, intel has nowhere else to go until next architecture and amd is still very focused on multithreaded performances. And when old games like GTA V are nearly bottlenecked by i5s, it's pretty obvious that the consumer industry will have to shift to more cores. It's also more power efficient than intel 4.7GHz housefires so you can expect laptops to up the core count too.

Except he is completely wrong

Says you.

Except he's not

why doesn't intel fix their multi core performance?

>pic related: $330 CPU vs $1600 CPU

...

Ryzen does not have more IPC than anything intel has to offer

From what I understand, hyperthreading only kicks into effect once more threads are being used than you have physical cores. If you're playing a game designed for a quad core CPU on a quad core CPU without hyper threading, the CPU will use all four cores without hyper threading instead of using two cores with hyper threading.

Shit, I meant quad core CPU WITH hyperthreading.

>AMD just killed entire INTEL division
NVIDIA is next

>AMD just killed entire INTEL division
Shit I can really feel their market share still being above 60% what is intel gonna do

Because only poorly made programs rely on single threaded performance nowadays. The future is packing as many cores in as possible while maintaining an acceptable level of single threaded performance for legacy applications.

you realize that's 30% less than last year?
for giants like intel it's a disaster, the bigger you are the bigger every little blow becomes

...

honestly, I love the fact that cpu performance has stagnated because now I don't feel the need to upgrade for a very long time

Single thread will never disappear. Not every program can be parallelized, even in theory, and writing parallel code is much more expensive. Even parallel engines rely on a master loop to keep things sync'd. The future is only in parallel CPUs because we've hit a wall in single core performance. Taking advantage of 2-4 cores isn't hard but 8? Yeah... very tough thus DX12 to get something out of it.