Question How does 7nm actually improve performance for the ryzen 3000 series cpu's compared to ryzen 2000 cpu's?

Jan 10, 2019
15
0
10
0
By performance i also mean if multitasking gets better. I actually was disappointed that the ryzen 3 3200 wasnt 6 cores and 12 threads. Guess the leak was fake. Anyways im planning to build a pc once ryzen 3000 comes out. However i have like a bunch of questions.

1. Whats the difference between the ryzen 3 3200g and ryzen 5 3600g other than 4 cores and 4 threads and 4 cores and 8 threads for the ryzen 5. I do want some multitasking ability and im not sure if the extra 4 threads does benefit in multitasking, does it? Is the cpu faster with 4 extra threads?

2. The title. Im on a really tight budget but im not sure if i should go for the 7nm ryzen 3600 with 6c and 12t.

3. Would the b450m gaming plus be able to hold the ryzen 5 3600. Definitely not getting x570 but i want to know if the motherboard can withstand or cool or support the ryzen 5 3600.

4. What ram speed should i aim for with the ryzen 3000 series? Would 2666mhz suffice or should i get 3000.

Probably no one will answer this thread but if u do i appreciate your help so much and im sorry for having pain in the a$$ questions. Ty
 

13thmonkey

Titan
Moderator
  1. I believe a better iGPU on the 3400g vs the 3200g.(there is no 3600g) https://www.anandtech.com/show/14523/amd-ryzen-3000-apus-up-to-vega-11-more-mhz-under-150
  2. 7nm itself will have an effect, but also the enhanced design has an effect, clock for clock the 7nm processors will be quicker than 12nm, and the 7nm will clock faster and use less power and therefore be cooler.
  3. that'll be fine.
  4. 3200 minimum, ryzen loves fast ram.
Are you getting a GPU, or do you NEED the iGPU, if you are getting a GPU then go 3600.
 
Reactions: DMAN999

InvalidError

Titan
Moderator
The whole point of extra hardware threads is improved performance (25-40%) at almost no extra silicon cost in more heavily threaded workloads from leveraging more of the CPU cores' often under-used resources.
The IGPs on the 3200G and 3400G are the same Vega8 and Vega11 from the 2200G and 2400G, only clocked a bit higher with the additional benefit of the APU officially supporting faster RAM.
Since DDR4-3000 (and even 3200 sometimes) costs about the same as DDR4-2133, there is no point in settling for anything slower unless you already own the slower RAM or have a good deal on a second-hand set.

For Ryzen 3600 and up, having twice as much L2 and L3 cache along with a few other memory-related architecture tweaks should reduce sensitivity to memory performance by a fair amount. Otherwise, it would be a horrible amount of extra die area overhead for nothing.

I wouldn't worry about it until the chips have actually launched since anything until then is mostly speculation. Let numbers from launch-day reviews speak for themselves and decide from there. Anything we write before then may ultimately get proven wrong due to any aspect turning out worse or better than expected. The biggest unknown IMO is how well the re-structured caches and IF will mitigate the likely increased latency of having to go through the IO die for cores on the other chiplet on 12/16C models, IO and RAM.
 
Reactions: DMAN999 and Mandark

13thmonkey

Titan
Moderator
The whole point of extra hardware threads is improved performance (25-40%) at almost no extra silicon cost in more heavily threaded workloads from leveraging more of the CPU cores' often under-used resources.
The IGPs on the 3200G and 3400G are the same Vega8 and Vega11 from the 2200G and 2400G, only clocked a bit higher with the additional benefit of the APU officially supporting faster RAM.
Since DDR4-3000 (and even 3200 sometimes) costs about the same as DDR4-2133, there is no point in settling for anything slower unless you already own the slower RAM or have a good deal on a second-hand set.

For Ryzen 3600 and up, having twice as much L2 and L3 cache along with a few other memory-related architecture tweaks should reduce sensitivity to memory performance by a fair amount. Otherwise, it would be a horrible amount of extra die area overhead for nothing.

I wouldn't worry about it until the chips have actually launched since anything until then is mostly speculation. Let numbers from launch-day reviews speak for themselves and decide from there. Anything we write before then may ultimately get proven wrong due to any aspect turning out worse or better than expected. The biggest unknown IMO is how well the re-structured caches and IF will mitigate the likely increased latency of having to go through the IO die for cores on the other chiplet on 12/16C models, IO and RAM.
I'm also interested to see if there is oddness associated with >8 cores as two die as used.
 

ASK THE COMMUNITY

TRENDING THREADS