SunDevil56

Memory questions...

Recommended Posts

Thinking of upgrading my memory, should I just concentrate on the memory's speed, or should I also go for the lowest latency...??

Share this post


Link to post
Help AVSIM continue to serve you!
Please donate today!

1 hour ago, SunDevil56 said:

Thinking of upgrading my memory, should I just concentrate on the memory's speed, or should I also go for the lowest latency...??

U definitely want to pay attention to the latency.  What speed are u looking at?

Share this post


Link to post
27 minutes ago, mpw8679 said:

U definitely want to pay attention to the latency.  What speed are u looking at?

3466. My mobo supports up to 4000 I believe.

Edited by SunDevil56

Share this post


Link to post
1 hour ago, SunDevil56 said:

3466. My mobo supports up to 4000 I believe.

I would go no higher then a cl16.  Or look at 3200 at cl14.

  • Upvote 2

Share this post


Link to post

See Bob's post in this thread:

Greg

  • Upvote 2

Share this post


Link to post
1 hour ago, lownslo said:

See Bob's post in this thread:

 

Thanks Greg.

Yes, I have read that thread before, and Bob posted some great info there...

Using Bob's formula i've come up with a figure of 9.2ns "true latency" for the memory i've selected, (G.Skill Trident Z DDR4 3466 CAS Latency 16) and this would qualify as at, or very near, "performance level" memory by Bob's calculations....

Bob's recommendations on "true latency" for memory.....

9ns is performance-level, 10ns is average, and above 10ns is slow consumer-grade stuff.

 

Edited by SunDevil56
  • Like 1

Share this post


Link to post

Another important variable in this is your budget.

Greg

Edit: Just noticed you have a very cool co-pilot! 🥂

Edited by lownslo
The cat!
  • Upvote 1

Share this post


Link to post
11 hours ago, SunDevil56 said:

this would qualify as at, or very near, "performance level" memory by Bob's calculations....

Bob's recommendations on "true latency" for memory.....

9ns is performance-level, 10ns is average, and above 10ns is slow consumer-grade stuff.

Did you realize that Bob's recommendation are numbers that are extremely close together. I would definitively not to for those 3466MHz CL16 modules. As mentioned, 3200MHz CL14 should be the starting point with 8.75ns. Either choose those (there are a lot available to a reasonable price) or try to go lower on latency. In the end, latency is slightly more important than bandwith (simply because bandwith is already quiet high with DDR4 modules above 3000MHz), so I would in any case prefer the 3200MHz CL14 over the 3466MHz CL16 modules.

Share this post


Link to post

for kicks a giggles but maybe somewhat relevant if OP decides to overclock the new sticks-

GSkill's - 3200 c14 slightly overclocked


6pcpVTD.png

 

...

 

  • Upvote 2

Share this post


Link to post

Hey Ken, thanks for the suggestion, but overclocking CPU's GPU's, and RAM has just never really been in my wheelhouse. As Steve put it (more or less).... Some people prefer to just slap in the best hardware they can afford, open the barn door, and let the horses run...😁

FWIW, with just a very few cfg tweaks, including an AM graciously supplied by Steve, i'm getting great overall performance, it's smooth, virtually stutter free, and very good fps, even in some of the most demanding scenery and weather environments, and I have a lot of higher end scenery and AC installed, and my P3D settings are mostly very high or above.... So I guess you could say i'm a pretty happy camper overall with the way my system and setup are working. Even so, i've learned a lot from you guys, and i'm always looking at what people are doing to improve the P3D experience...

Cheers,

Scott

  • Upvote 1

Share this post


Link to post

I understand and thanks for the education.  Now does it really make much of a difference when running P3D?

Share this post


Link to post
11 minutes ago, diajohn said:

I understand and thanks for the education.  Now does it really make much of a difference when running P3D?

Not sure who this was directed at, but that's kind of a proverbial "$64,000 question" to put it mildly...

If you can be a little more specific about what it is you want to know it would be very helpful to anyone who might want to respond here...😉

Share this post


Link to post
17 hours ago, AnkH said:

Did you realize that Bob's recommendation are numbers that are extremely close together.

Not sure I follow here...the difference between cutting-edge 8.25ns RAM and average 10ns RAM is 17.5%...not close together at all when related to the CPU speed.  A 5GHz CPU will tick through almost 9 additional idle clock cycles waiting for data with the average RAM vs the high-end stuff.

Does it make a difference?  It sure seems to, especially in areas of dense autogen and/or photoscenery (think ORBX SoCal).  When you see performance hits like dropping frame rate and/or stutters, yet neither the CPU nor GPU are close to maxxed out when it's happening, I believe that limited memory bandwidth is the culprit in that situation.

Regards

 

  • Upvote 2

Share this post


Link to post
7 minutes ago, w6kd said:

Not sure I follow here...the difference between cutting-edge 8.25ns RAM and average 10ns RAM is 17.5%...not close together at all when related to the CPU speed.  A 5GHz CPU will tick through almost 9 additional idle clock cycles waiting for data with the average RAM vs the high-end stuff.

Does it make a difference?  It sure seems to, especially in areas of dense autogen and/or photoscenery (think ORBX SoCal).  When you see performance hits like dropping frame rate and/or stutters, yet neither the CPU nor GPU are close to maxxed out when it's happening, I believe that limited memory bandwidth is the culprit in that situation.

Regards

 

Thanks Bob, I appreciate the clarification on this.

BTW, fellow Ham here, callsign N3TTN, 73'..😎

Edited by SunDevil56

Share this post


Link to post

It would be interesting seeing a real benchmarking with different RAM (and same everything else) in the P3D v4 world.

 

Share this post


Link to post
2 hours ago, w6kd said:

Not sure I follow here...the difference between cutting-edge 8.25ns RAM and average 10ns RAM is 17.5%...not close together at all when related to the CPU speed.  A 5GHz CPU will tick through almost 9 additional idle clock cycles waiting for data with the average RAM vs the high-end stuff.

Does it make a difference?  It sure seems to, especially in areas of dense autogen and/or photoscenery (think ORBX SoCal).  When you see performance hits like dropping frame rate and/or stutters, yet neither the CPU nor GPU are close to maxxed out when it's happening, I believe that limited memory bandwidth is the culprit in that situation.

Regards

 

QFT!

Share this post


Link to post
31 minutes ago, Wanthuyr Filho said:

It would be interesting seeing a real benchmarking with different RAM (and same everything else) in the P3D v4 world.

 

Paging FunknNasty, paging FunnNasty about memory tests... :biggrin:

Greg

Share this post


Link to post

Sorry about lack of clarity.  I wanted to know if anyone could say that faster ram actually makes a user detectable difference with other components being equal, in P3D.  Usually a RAM change is done with other components this blurring the equation.

 

John

  • Like 1
  • Upvote 1

Share this post


Link to post
2 hours ago, lownslo said:

Paging FunknNasty, paging FunnNasty about memory tests... :biggrin:

Greg

Answering page ....

I will try to put up some entertaining graphics on the subject tomorrow .....And I wonder why no one wants to talk memory. 🙂

-I installed an 8086 and 4400 mems today ...the good bad and ugly

  • Like 1

Share this post


Link to post
14 hours ago, diajohn said:

Sorry about lack of clarity.  I wanted to know if anyone could say that faster ram actually makes a user detectable difference with other components being equal, in P3D.  Usually a RAM change is done with other components this blurring the equation.

 

John

Hi John,

I believe this article helps to put things into proper perspective:

https://www.velocitymicro.com/blog/important-ram-speed/

Edit: although Prepar3D may be a special case in the way CPU cache, VRAM and System memory all  interact for best performance.

Regards,

Mike

 

Edited by Cruachan

Share this post


Link to post

 Cheers!

Edited by FunknNasty

Share this post


Link to post
3 minutes ago, FunknNasty said:

Proper perspective?

That's showing some real appreciation for the guys behind P3D, don't you think? I mean that two paragraph artical says nothing about how Prepared compares to Grand Theft Auto ....or did I miss something.  Your post gets a 'C'mon man!' 🙂

He, he, I was typing my edit as you posted your response! However, that article does still have some validity...only my opinion, of course!

Mike

Share this post


Link to post
1 hour ago, Cruachan said:

Hi John,

I believe this article helps to put things into proper perspective:

https://www.velocitymicro.com/blog/important-ram-speed/

Yikes...this article does not reflect an in-depth understanding of system operation.

First, the article says that the RAM frequency is "the number of commands it (the RAM) can process per second."  Absolutely NOT.  The frequency is the number of clock ticks per second (from the IMC).  It is the pacing mechanism used by the memory module...but the memory does not execute a "command" each tick, but instead steps forward in the process of reading or writing data to RAM.  On one tick the memory might receive the read command, then the next tick it might set the data ready line to a false state, then it might wait 10 or more ticks for the data to stabilize on the bus, then it sets the data ready line to a true state, etc.  The total number of ticks needed to go from a memory request to completing that request is the CAS latency.  This is, of course a simplified example, but the characterization in this article of RAM frequency as a command rate is...wrong.

Then, in an even more glaring misstatement, the author says "system RAM will largely not be used when gaming."  This is patently wrong...system RAM is used for virtually every process running on a computer.

Since the background discussion in this article is so far gone, I can't with good conscience even consider the conclusions they draw as anything but flawed.

Regards

 

  • Like 2
  • Upvote 2

Share this post


Link to post

No 'c'mon man' award for you then...

Sorry to have to do this to you Mike, but after reading Bob's post you’re going to have re-accept the 'c'mon man' award ..... sorry man, All in good fun, of course.

 

Edited by FunknNasty

Share this post


Link to post

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now