Why are people praising AMD Ryzen 9 3950X over the i7900k?

RedRizla said:

At first I thought people were praising the AMD Ryzen 3950x due to it's price and since I'm looking to build another computer I thought I was in for a bargain. That is until I checked the price of the AMD Ryzen 9 3950x and saw it's £689. So why all the rave for AMD Cpu's? The Intel i7 9700k has better performance in games and it's cheaper at £570. 

The best answer - is always : different work loads - different strokes. For example, I am loving my new Ryzen 3700x. 8 cores 16 threads is amazing for my programming set of work and multitasking usage. While maybe not the top performer in all games do to Intel still having the upper hand in clock speeds - it still does 90-95% range in most games while still costing way less. While the 3950x might not be everyone - there are other CPUs in the product stack that can and will if you compare it right to the Intel lineup costing 100 or less in most cases that can be thrown back to the GPU where it will make up the performance differences in most cases (short of 1080p gaming). Then again - who really needs 300+ fps that is in meaningful numbers. Just seems silly to me.

One key use case for the 3950x would be more VMs for testing reasons. I run a number of VMs, still want to play games on the side, and listen to music and do research (15+ tabs) in my chrome:P Sand-boxing programs in the future might make use of this as it would allow the program to access data it needs - but keep it from burning everything else down. Chrome does this already. So does a lot of A/V programs.

pyroth309 said:

bigmike20vt said:

Yeah I forgot about thread title ;). So sure for gaming if you go top of the line  intel has an edge because they still dominate in single threaded performance ,(but less than they used to) so for pure gaming in general if you go top end intel still win. AMD shine however where multiple cores benefit more.

Just for gaming tho I would never go top end tbh. It's just wasted money you are far better buying a faster GPU or if you have that already then put it towards a holiday when lockdown ends

Yea exactly, unless there was some old niche title that is CPU heavy that you plan to play primarily. It's a shame game engines and developers haven't gotten better at using multiple cores. Even the ones that support it tend to load up a single core more than the others and do a bad job distributing the load.

That's because in programming - there isn't really a "good" or "best" way to share workloads between all the cores. For example, in Python we just more recently just got the ability to do share memory between cores. It was possible already if you wrote your own manager, but now that it is support module, I see a big jump in users using it more already. I do blame hardware for not making it easy for use in some ways, but software does take time to play catch up as well. 

From a hardware stand point - it would be super crazy to see parts of the chip redesign themselves for difference purposes and smart memory to help off set the process needed to ask where the data should go that also sits between everything.
1) Any time you move data between you cost time and power <-- bring stuff closer to it thus have more sections/big smart cores that can do it all - This is way a CPU core has more stuff vs a simple GPU core.
2) Any time you move a size of data you also cost time and power <-- compress data - GPUs are famous for this
3) Any time you have to wait for a section of the Core to open up you cost time and power <-- allow more process access to the same core (threading).
4 etc etc

Future improvements:
1) Sections of the chip that can do more than one type/reconfigure themselves a bit to fit the work better. For example, turn some of the IO part of the core into a math unit or have the math unit - wait for for IO data. Thus more core is used and more work can be done per cycle.
2) Make it easier to get to the same data. Right now its just takes way too long to access the same data from one end of the chip to the other end. This is one of the reasons why current CCXs will be reconfigure in Ryzen 4000 to share the same pool of memory inside each chip-let vs set of cores.
3) Lots of smaller dummer cores from Intel and ARM attach to a few bigger smarter cores. This is a bit of cheat to the reconfigure chips I talk about - but same idea in that the CPU it self will spread the work better depending on the work load needed to get done.
4) A chip that is seen as a single unit vs a chip that is seen as multi cores. Sort of like how MT works - it would be a lot nicer from a programming stand point to say the CPU is just a single giant core that just did the work for us in spreading up the work we give it. Even though we would still have to think/protect our selves from dead locks - it still would be a lot easier to program like we did in the past vs trying to wrap our brains around having to manually spread the work ourselves.

RedRizla said:

I had an AMD Athlon back in 2002 and it served me well. I think I switched to Intel after hearing they could withstand higher temperatures if I'm not mistaken. I'm not sure if AMD are as good as Intel now when it comes to high temps, I haven't checked.

How similar we all are, i started my first build with a celerons, then moved to Antlon and then went back to intel due to some features like temp throttling to save the CPU from burning out if the fan died, unlike the athlons back in the day that relied on the mobo bios to tell it if the fan had failed, while of course allowing the thing to continue to heat up with no cooling. lol