[SOLVED] Intel or amd in single core?

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

InvalidError

Titan
Moderator
The Intel memory overclock record is 5.7GT/s vs 5.1GT/s for Ryzen 3xxx.
https://www.tomshardware.com/news/ddr4-memory-overclocking-world-record-ram-micron,39358.html

So Intel's memory controller lead extends all the way into extreme territory and goes a considerable bit beyond 4.4GT/s.
 
The Intel memory overclock record is 5.7GT/s vs 5.1GT/s for Ryzen 3xxx.
https://www.tomshardware.com/news/ddr4-memory-overclocking-world-record-ram-micron,39358.html

So Intel's memory controller lead extends all the way into extreme territory and goes a considerable bit beyond 4.4GT/s.
Intel overclock was on ln2 while Ryzen was on air...
That's not exactly great comparison.
 
But for a normal person that doesnt matter as virtually no-one runs ram above 4000mhz.
Most gamers will plug in a 3200mhz kit or even 3600mhz and turn XMP on with no issues. Intel or AMD.
Exactly. I was just stating the improvements AMD managed to make to their controller by putting it on a separate die, giving it much more space without impacting the yields since it's made on an older cheaper process.
And since Intel doesn't have their main data pipeline speed tied to the ram speed like Zen does it's only logical that their controller isn't as sensitive.
I don't know why InvalidError decided to bring that up though since op didn't ask for it.
 

InvalidError

Titan
Moderator
I don't know why InvalidError decided to bring that up though since op didn't ask for it.
Simple: Finstar stated that Ryzen 3 had a "very good memory controller because it is on a separate chiplet" I pointed out that Intel's memory controller is still better and still on a monolithic die. Also, Zen 2 incurs a ~15ns memory latency penalty compared to Zen+ (~65ns vs ~80ns) due to the chiplet-to-chiplet interface, which is a large part of the reason why AMD had to increase L3 cache to a ridiculously large 16MB per CCX so most frequently used data would remain in-CCX to compensate.

Splitting things up isn't intrinsically better.
 
Simple: Finstar stated that Ryzen 3 had a "very good memory controller because it is on a separate chiplet" I pointed out that Intel's memory controller is still better and still on a monolithic die. Also, Zen 2 incurs a ~15ns memory latency penalty compared to Zen+ (~65ns vs ~80ns) due to the chiplet-to-chiplet interface, which is a large part of the reason why AMD had to increase L3 cache to a ridiculously large 16MB per CCX so most frequently used data would remain in-CCX to compensate.

Splitting things up isn't intrinsically better.
Better because it can do higher clocks on ln2 than Ryzen's on air?

You're comparing apples to oranges.
AMD managed to make massive improvements to their memory controller by moving it on a separate chiplet from a monolithic die.
Intel has yet to do that so there's no reference point for Intel cpus.
 
Last edited:
I agree with InvalidError that Intels memory controller is better than ryzen 3k from what I read, but both have no issues getting to the speeds that most people buy.

As I said earlier, nobody runs incredibly fast ram. No performance increase and way higher cost.

The cheapest ddr4 4800mhz kit on pc part picker is nearly $450 for 16gb.
 

ASK THE COMMUNITY

TRENDING THREADS