• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD's Infinity Fabric

nlel1975

nlel1975

nlel1975

Soldato
Joined
22 May 2007
Posts
3,504
Location
GJ
Firstly I am not a troll. I was just wondering, many people on here say Ryzen CPU performance is much much much better with faster memory (I'm talking loads better) due to the infinity fabric.

So I was just wondering how come AMD CPU's don't trounce Intel CPU's which don't have Infinity Fabric at the same price point.

Most on here make it sound like the difference is a major major concern and it makes loads and loads and loads of difference.

Once again not a troll. Just wondering.
 
The architecture is completely different. There is no apples to apples comparison.

The reason why ram speed affects Ryzen is the speed of the ram is the speed of the infinity fabric which is the interconnect between the cores.

So faster ram = faster cpu.
 
Infinity fabric isn't a feature that improves performance like HTT. It's just the means to transfer data across all linked components.

https://en.wikichip.org/wiki/amd/infinity_fabric

By using faster RAM it can operate more quickly. Intel chips are of a different design and don't use the RAM in the same way as AMD chips. To use the car analogy, there are lots of different engines but they all get the job done sightly differently.
 
Ok, but like how much extra performance are we talking, 10% to 15% isn't a whole lot and I doubt it makes anywhere near 50% difference. If you're building a top spec PC to last years and have the money then I don't disagree the 10% to 15% is worth having.
 
As one of the people that has been trying to explain how faster ram impacts the infinity fabric, i want to point out that i never said that it "is a major major concern and it makes loads and loads and loads of difference."

I did say right at the start:

"I just fact checked my previous post.

The speed of the infinity fabric within the cpu is tied to the speed of the ram.

I make no claims about how this affects performance. But faster ram does = a faster infinity fabric.

This is the case with ryzen cpus."


Every post of mine after this said that the speed of the infinity fabric, which is the bus speed within the cpu is directly tied to the speed of the ram. So having faster ram absolutely does have an impact on zen based cpus like ryzen.

This was in answer to your comment that ram is ram and has no effect on the speed of the cpu outside of what a normal ram speed increase has done historically.

I was correcting an incorrect assumption and provided both the link to the info provided by nvidia, and a link to testing done that shows this to be the case.

In fact if you do your own research into the infinity fabric and ram speed you will find both the articles i linked and more.

I personally have not accused you of trolling.

Edit: link for reference. My posts start at 739
https://forums.overclockers.co.uk/threads/ryzen-ddr4-memory-what-you-need-to-know.18770248/page-37
 
Last edited:
I've never tried to deny there is a performance improvement, my point is that is it really a significant improvement? In my opinion 10% to 15% isn't a fat lot, its an improvement for sure but not loads and loads and loads and loads better. I agree with your point that you've never made claims about performance increases just that there is an improvement, which I've never tried to deny.
 
nlel1975 said:
I've never tried to deny there is a performance improvement, my point is that is it really a significant improvement? In my opinion 10% to 15% isn't a fat lot, its an improvement for sure but not loads and loads and loads and loads better. I agree with your point that you've never made claims about performance increases just that there is an improvement, which I've never tried to deny.
Click to expand...

But you did.

nlel1975 said:
Faster RAM means faster access times, it doesn't make the CPU itself perform any better. Faster RAM will show up in RAM benchmarks but it won't make any difference to CPU benchmarks.
Click to expand...

It's this erroneous belief that i was trying to correct, since other people were asking about how ram affects ryzen and I wanted them to have the facts. My point was that for zen based cpus faster ram does make the cpu perform better because faster ram = faster infinity fabric. How much better i leave to the various sites on the internet that have done the testing.

I'm going to stop now because i feel like I'm going around in circles.
 
Distracted said:
I'm going to stop now because i feel like I'm going around in circles.
Click to expand...

If I did I didn't mean to, I was only ever trying to point out the performance increases aren't really worth making a big deal out of, unless you're one of those people who thinks 10% to 15% is loads and loads and loads and loads better.
 
Martini1991 said:
Why would I want my 2700 to perform 10-15% worse?
Click to expand...

In a situation where the faster memory costs a significant amount more money. Which to be fair with ocUK current pricing on 3000mhz memory isn't the case. But really 10% to 15% is nothing and a human wouldn't be able to perceive it. Of course in gaming with a 1080 and above it affects your frame rates so you don't want your memory bottlenecking your expensive graphics card.
 
nlel1975 said:
In a situation where the faster memory costs a significant amount more money. Which to be fair with ocUK current pricing on 3000mhz memory isn't the case. But really 10% to 15% is nothing and a human wouldn't be able to perceive it. Of course in gaming with a 1080 and above it affects your frame rates so you don't want your memory bottlenecking your expensive graphics card.
Click to expand...

Why mention about faster memory costing significantly more when it doesn't?

Back when I bought a 1700, 2400MHZ RAM and 3000MHZ RAM were about 10 pound difference. That 10 pound difference results in a big performance gain per pound spent.

If you're going to say that 10-15% doesn't matter then why bother buying anything better? I run a Vega 64 on a 144HZ screen, I want the performance.
 
Well in my case the memory I bought in my sig was £ 19 a stick cheaper than ocUK current cheapest 3000mhz 32GB memory kit but it was more when I bought my memory as they've recently dropped the prices. I wouldn't argue that 10% to 15% performance is probably worth an extra £ 19 per stick.
 
nlel1975 said:
In a situation where the faster memory costs a significant amount more money. Which to be fair with ocUK current pricing on 3000mhz memory isn't the case. But really 10% to 15% is nothing and a human wouldn't be able to perceive it. Of course in gaming with a 1080 and above it affects your frame rates so you don't want your memory bottlenecking your expensive graphics card.
Click to expand...

To be fair Intel has been giving us 10% improvements for years and people have been happy to pay handsomely for it. Others will buy Intel over AMD for a few FPS improvement in gaming.

The other factor to consider is if that 10-15% pushes you over a threshold you'll definitely feel the difference. The only real issue is RAM is expensive atm, then again this is an enthusiast forum. Stock speeds can be had for less on a DELL or HP system of that's all that's required.
 
nlel1975 said:
Well in my case the memory I bought in my sig was £ 19 a stick cheaper than ocUK current cheapest 3000mhz 32GB memory kit but it was more when I bought my memory as they've recently dropped the prices. I wouldn't argue that 10% to 15% performance is probably worth an extra £ 19 per stick.
Click to expand...

Fact is, Ryzen scales extremely well with faster ram, however the cost of ram at the time of ryzen was on the increase, and has only recently started to descend, so for many the cost of ram outweighed the performance bonus.

I would not call 10-15% mediocre, not when Intel struggle to give you 5% per gen increase in performance... if you can instantly add 15% performance to your chip for another £20 why would you not?
 
It's just the case that the MEMCLK and DFICLK are hard linked (at a ratio of 1:2). Higher the MEMCLK, the higher the DFICLK. The DFICLK is the clock for the Data Fabric/Infinity Fabric.

Below 3200MHz MEMCLK there is a noticeable bottleneck on the DFICLK, hence why people are aiming for 3200MHz+. Beyond the 3200MHz-3466MHz mark the performance increase diminishes slightly as this bottleneck is overcome.
 
Last edited:
nlel1975 said:
I've never tried to deny there is a performance improvement, my point is that is it really a significant improvement? In my opinion 10% to 15% isn't a fat lot, its an improvement for sure but not loads and loads and loads and loads better. I agree with your point that you've never made claims about performance increases just that there is an improvement, which I've never tried to deny.
Click to expand...

There are videos on the YT showing the effects of different memory speeds on Ryzen, I can't link them as I'm at work but I'm sure you can find them.

Your CPU can only output a result once all the parts of the CPU what you don't want is you processing core sitting idle waiting for data to be given to it because it's still travailing over the infinity fabric, ideally you want the data already there waiting to be called by processing core. Think of it like a bus service, the bus service runs optimally if all the passenger are there waiting at the bus stop but that same service would be slower if the bus had to wait for passengers who are still making their way to the bus stop. By speeding up the infinity fabric that data is getting pushed around quicker and quicker reducing delays, in day to day terms slow memory in Ryzen can manifest itself as lag in games or sub-optimal results in benchmarks. By the the same token memory once you start using memory faster then 3000Mhz/3200 Mhz it doesn't really yield tangible improvements because the processing cores aren't waiting around for data.

It's hard to given a reliable % because it really comes down to what CPU is doing, if your worried/curious then as I said above there are plenty of videos and web publications that have benchmarked this since Ryzen came out last year in various scenarios.
 
nlel1975 said:
Well in my case the only benchmark I use is 3D Mark Timespy. Here are the results of my PC (in sig) with 2133 RAM compared to a Ryzen 7 2700X machine with 3000mhz RAM... both with exactly the same graphics card.

My PC... https://www.3dmark.com/spy/4439154
Ryzen 7 2700X with 3000mhz RAM https://www.3dmark.com/spy/4342539

The differences can be put down to the CPU not the memory.
Click to expand...

There isn't enough info there with a single benchmark to make those assumptions, in fact you simply cant make assumptions at all on such a small data set. I said it in your other thread and ill say it here. Memory frequency does make a difference on ryzen as explained above the reasons are crystal clear, much like the old AMD barton cores that scaled on memory frequency and core FSB frequency as they were 1:1, ryzen and its infinity fabric interconnect scales on those same metrics. You could attribute the fact they are so close as an extra 30mhz (core) on your card which could bring you closer either way on the same like for like system you will notice some differences across varying benchmarks as you up memory frequency.
 
nlel1975 said:
Well in my case the only benchmark I use is 3D Mark Timespy. Here are the results of my PC (in sig) with 2133 RAM compared to a Ryzen 7 2700X machine with 3000mhz RAM... both with exactly the same graphics card.

My PC... https://www.3dmark.com/spy/4439154
Ryzen 7 2700X with 3000mhz RAM https://www.3dmark.com/spy/4342539

The differences can be put down to the CPU not the memory.
Click to expand...
Impossible to see the difference unless you're using the same CPU, that comparison is just telling you that a better CPU is better...
 
You must log in or register to reply here.
Back
Top Bottom