Ryzen 3950X Build Help

[moproblems99]

In most standard cases, the air is designed to be taken in at the bottom front and exhausted from the top rear.

 

[erek]

hi, just got done reversing all those fans, and most cable mgmt can you review it?

 

[erek]

Decent Airflow Design now?


Also is it possible to run tight timings including 1T with 4x16GB DIMMs? i don't want to go 4x if my timings have to get looser to run

 

[oxrufiioxo]

View attachment 138775

Decent Airflow Design now?


Also is it possible to run tight timings including 1T with 4x16GB DIMMs? i don't want to go 4x if my timings have to get looser to run

It's going to dependent on the quality of the ic's. Airflow wise you should be good.

 

[Zach_01]

You can start by 16-18-18-18-36
Do you know your current ICs (Thaiphoon) and what tRFC run now?
First run a AIDA64 momory test to know your starting point.
Ideally, prior to all, must set the CPU to static clock (low or high doesnt matter... like 4.0~4.1GHz) until you find DRAM settings because the (all over the place) boost of ZEN2 is going to mess with results.

 

[erek]

You can start by 16-18-18-18-36
Do you know your current ICs (Thaiphoon) and what tRFC run now?
First run a AIDA64 momory test to know your starting point.
Ideally, prior to all, must set the CPU to static clock (low or high doesnt matter... like 4.0~4.1GHz) until you find DRAM settings because the (all over the place) boost of ZEN2 is going to mess with results.

 

[A Computer Guy]

An interesting video with 3950x and memory tuning comparisons...

 

[oxrufiioxo]

An interesting video with 3950x and memory tuning comparisons...

Yeah I posted that in another thread i think that memories like $300 for a 16gb kit though.


You wouldn't have issues with timings on that kit @erek lol.

 

[erek]

I just wonder if i can get the F4-3600C16D-32GTZNC kit to work @ XMP after RMA that if i go with two sets of them for 64GB with all 4 DIMM slots filled if i can still run @ XMP
DDR4-3600MHz CL16-19-19-39 1.35V with 1T

or if with running 4 DIMMs will force me to loosen up to 2T and other worse timings?

 

[oxrufiioxo]

I just wonder if i can get the F4-3600C16D-32GTZNC kit to work @ XMP after RMA that if i go with two sets of them for 64GB with all 4 DIMM slots filled if i can still run @ XMP
DDR4-3600MHz CL16-19-19-39 1.35V with 1T

or if with running 4 DIMMs will force me to loosen up to 2T and other worse timings?

My guess is it should work but most likely your sub timings are going to be terrible and I dont think ryzen dram calculator has profiles for your ram ICs yet. You'd probably need to fiddle with your timing even with 2 dimms honestly to get decent performance.


Trc and trfc can be a pain on non Bdie ram

My sub timings even on my 3200 14-14-14-31 bdie are pretty terrible with XMP though.

 

[Deleted member 185158]

I just wonder if i can get the F4-3600C16D-32GTZNC kit to work @ XMP after RMA that if i go with two sets of them for 64GB with all 4 DIMM slots filled if i can still run @ XMP
DDR4-3600MHz CL16-19-19-39 1.35V with 1T

or if with running 4 DIMMs will force me to loosen up to 2T and other worse timings?

Just because sometimes memory is 2T command rate, doesn't really make it a bad thing.
Most of the latency read and write gains comes from pure clock speed even at loose timing sets.
People tweak the timing tighter to squeek out performance.
You want to run the FLCK as high as it goes, and the memory speeds to match that.
Personally, I wouldn't worry to much on the timings in your case figuring you want a stable rig.

 

[EarthDog]

64gb at 3600 may take some tweaking...that's a lot of ram and at the end of the road for 1:1 operation.

 

[Zach_01]

I just wonder if i can get the F4-3600C16D-32GTZNC kit to work @ XMP after RMA that if i go with two sets of them for 64GB with all 4 DIMM slots filled if i can still run @ XMP
DDR4-3600MHz CL16-19-19-39 1.35V with 1T

or if with running 4 DIMMs will force me to loosen up to 2T and other worse timings?

I believe your 16GB modules now are 2 ranked right? (2dimms, 2 rank)
For the record the RyzenDRAMcalc does not change timings or CR 1T->2T when given a config of 4dimms/2rank at a given speed. What changes is the termination settings (ProcODT, RTT_x)
For sure the are other timings, not visible/accessible that XMP profile auto sets or the board auto sets when XMP is disabled.

Plus the calc is not a infrangible rule, cause at the end the CPU's UMC is the major factor in the equation... and that is lottery

 

[Nordic]

This does seem to indicate a lot, ~5% gain for 3600 @ CL17 compared to 2400

AMD Zen 2 Memory Performance Scaling with Ryzen 9 3900X

We take a close look at memory scaling on AMD's new Zen 2 Ryzen 3900X, testing both application and gaming performance at seven different memory speed and timing combinations ranging from 2400 MHz all the way up to 4000 MHz.

www.techpowerup.com

5% is not a lot. Upgrading from my 1060 3gb to a 2080 would net me about a 270% increase in gpu performance according to TPU reviews. Going from my old 4790k to a 3900x has net me an increase of about 315% in multicore performance by my own testing.

I am personally very interested in single threaded performance. Going from my overclocked 4790k to a 3900x has net me a 7% increase in benchmarks. Going from 2133mhz memory to 3600mhz memory only increases my performance in benchmarks about 3% in single threaded performance. That 7% single threaded from the cpu upgrade and 3% from fast ram doesn't add up to a whole lot in comparison.

 

[Deleted member 185158]

5% is not a lot. Upgrading from my 1060 3gb to a 2080 would net me about a 270% increase in gpu performance according to TPU reviews. Going from my old 4790k to a 3900x has net me an increase of about 315% in multicore performance by my own testing.

I am personally very interested in single threaded performance. Going from my overclocked 4790k to a 3900x has net me a 7% increase in benchmarks. Going from 2133mhz memory to 3600mhz memory only increases my performance in benchmarks about 3% in single threaded performance. That 7% single threaded from the cpu upgrade and 3% from fast ram doesn't add up to a whole lot in comparison.

3% is taken from what? What is the control benchmark that you would produce this type of figure>?

 

[A Computer Guy]

I just wonder if i can get the F4-3600C16D-32GTZNC kit to work @ XMP after RMA that if i go with two sets of them for 64GB with all 4 DIMM slots filled if i can still run @ XMP
DDR4-3600MHz CL16-19-19-39 1.35V with 1T

or if with running 4 DIMMs will force me to loosen up to 2T and other worse timings?

You can set 1T and if you have issues set Geardown=enabled before trying 2T. In any case no guarantee XMP settings are still going to work when you have 4 dimms installed unless your QVL (grain of salt) suggested it should work. You might discover some adjustment is necessary for the reasons other posters have mentioned. But then again it just might work.

One of the more difficult aspects for OC'ing your system is that DRAM Calculator for Ryzen didn't have a profile for your ram IC's yet (no guidance available) - so you will need to do research finding others with similar setups to find out what works (or what doesn't work), or go though a lot of trial and error your-self. Your playing with silicon lottery not just with your ram but motherboard, cpu (IMC), and ram all trying to work together. The situation can change if you update BIOS/UEFI later. Just things to keep in mind depending on how you want to spend your time.

How important are these things to you and how do you value them for your build and usage of your PC: Performance, Capacity, Aesthetics, Budget, and Stability?

If you'd rather go for capacity (64GB) and aesthetics (4 dimms) then potentially you may need to sacrifice on some performance (freq or timings).

There are people who are ok with these trade-offs depending on their preferences, and quite frankly with your level of hardware, if it came down to it sacrificing some ram speed for capacity would not be a big deal as your gaming performance potential should already be really great anyway. In the worst case scenario if you had to drop down to DDR4-3200 and/or maybe CL18 it wouldn't be the end of the world. If you didn't pay attention to it I would hazard to guess it wouldn't change the outcome of your overall gaming experience much anyway unless you have some very specific needs.

If you're running stable with decent timings I don't think you need to be overly concerned with having the most optimal timings depending on how difficult it will be to get those optimal timings. For example. I'm currently testing a different ram kit at 64GB 20-20-20-40-66 from 32GB 18-17-17-36-58 and it really didn't make any perceivable difference (at least to me) for the games I'm playing currently.

 

[Nordic]

3% is taken from what? What is the control benchmark that you would produce this type of figure>?

I didn't do extensive testing but cinebench, 7zip benchmark, and cpuz benchmark. Take it with a grain of salt. I used TPU reviews for more serious numbers. I think I made my point that 5% isn't a whole lot.

 

[Deleted member 185158]

I didn't do extensive testing but cinebench, 7zip benchmark, and cpuz benchmark. Take it with a grain of salt. I used TPU reviews for more serious numbers. I think I made my point that 5% isn't a whole lot.

Depends on how you look at 5% and how large the number is that you base that 5% from.

Like I cannot justify buying a 3700x even though its x% faster.

So 5% of 500 is lower than 5% of 5000 for example. Thats really the reason why I had asked because it would make a difference to some one to inquire some numbers.

Curious. 5% isnt a big number by itself.... But its end result might be

 

[Nordic]

Depends on how you look at 5% and how large the number is that you base that 5% from.

Like I cannot justify buying a 3700x even though its x% faster.

So 5% of 500 is lower than 5% of 5000 for example. Thats really the reason why I had asked because it would make a difference to some one to inquire some numbers.

Curious. 5% isnt a big number by itself.... But its end result might be

In terms of a performance boost it really isn't a good post. I chose an example of a gpu upgrade being a 270% increase in performance. That is a big difference.

 

[Zach_01]

Depends on how you look at 5% and how large the number is that you base that 5% from.

Like I cannot justify buying a 3700x even though its x% faster.

So 5% of 500 is lower than 5% of 5000 for example. Thats really the reason why I had asked because it would make a difference to some one to inquire some numbers.

Curious. 5% isnt a big number by itself.... But its end result might be

If you mean it like...: in 1 case 5% more FPS mean nothing in real life benefit and gaming experience and on another case 5% shorter rendering/editing time meaning for some people more money then yes it does making difference some times.

 

[Deleted member 185158]

Yes and Yes.

5% can make or break a benchmark score PR.

270% is a big number. Additional 5% memory performance increase with an additional 3% cpu performance increase = X time shortened for rendering, and better benchmark scores.

Gains are gains in my opinion. But that's a result of competitive benchmarking...... I'll take 1/4% if I can eek out a better score than someone else lol.

 

[oxrufiioxo]

https://www.reddit.com/r/Amd/comments/e8cbve

binning for these doesn't seem that much better than the 3900X at lest from silicon lotteries sample size. I'm most surprised by only 12% hitting 1900mhz fclk.

I'm also pretty sure AMD sent golden samples to reviewers their chips all clocked higher at lower voltages and hit 1900mhz in every review I read.

 

[moproblems99]

I just wonder if i can get the F4-3600C16D-32GTZNC kit to work @ XMP after RMA that if i go with two sets of them for 64GB with all 4 DIMM slots filled if i can still run @ XMP
DDR4-3600MHz CL16-19-19-39 1.35V with 1T

or if with running 4 DIMMs will force me to loosen up to 2T and other worse timings?

I have this kit F4-3600C16D-32GTZNC and I run C-Die timings at 3800 1:1 at 1T.

 

[Nordic]

Yes and Yes.

5% can make or break a benchmark score PR.

270% is a big number. Additional 5% memory performance increase with an additional 3% cpu performance increase = X time shortened for rendering, and better benchmark scores.

Gains are gains in my opinion. But that's a result of competitive benchmarking...... I'll take 1/4% if I can eek out a better score than someone else lol.

My whole point originally was about performance gain per dollar. Someone was saying if you are spending $2000 on a pc, what is another few hundred for ram. My point was that the gains are much smaller, as in ~5% to as high as maybe 300% or more. The dollar per percent gained is very high and some may not see that as worth it.

 

[Deleted member 185158]

My whole point originally was about performance gain per dollar. Someone was saying if you are spending $2000 on a pc, what is another few hundred for ram. My point was that the gains are much smaller, as in ~5% to as high as maybe 300% or more. The dollar per percent gained is very high and some may not see that as worth it.

Well certainly if you have 2k for a pc, you're going to buy high end high quality parts....
Doesnt make sense to spend big on a cpu and board then purchase cheap memory and then expect better performance.
Or 700$ cpu and buy a 30$ cooler and wonder why the cpu runs hot.

Then its not really all about expensive, its about choices.
Lead horse to water...... But let him buy value ram for 3950x.... Eek