3 hours ago, jaxgator said:

I hope you made sure that the memory that you purchased is compatible with your A10.  Alienware boxes tend to be quite finicky when it comes to RAM.

On the Newegg reviews it seems these are Samsung B die chips and the OEM units are Kingston as well.  A double good omen for a Dell based AMD box. 

I must admit i know very little about RAM in general.  If it gets me anywhere in the neighborhood of what the chipset on the mobo is rated for and plays nice with XMP i don't see myself tweaking anything else. 

4 hours ago, Sethos said:

That should do it, at least have you running at 6000 MT/s, 2000 FCLK and 3000 UCLK. Then use the overhead from backing down from 6200, to improve your timings.

If you want to run 6200MT/s, just set your FCLK to 2066Mhz instead, 2066 x 3 = 6198, rounded up 6200. Then that's also 1:1

Just ordered a set of G.SKILL Flare X5 Series DDR5 2x32GB 6000MT/s CL30-40-40-96 1.40V.  Should be here on Sunday.

It appears as you are quite knowledgeable in memory timings so if I can't figure it out through this discussion, I may be back with some questions.  😉

4 hours ago, psolk said:

Wow, I have been out of the PC building game for a bit but that is some really in depth knowledge!  Is this all specific to AMD or DDR5?  

Impressive stuff.  

AMD.

Not that @Sethosdoesn't know his Intel stuff as well. Gave him his props for running a delidded Intel chip and OCing the hell out of it on his previous system a few months back when a previous PC building discussion came up. 

Edited September 28, 20241 yr by Maxis

I upgraded from 32GB of DDR5-4800 CL40 to 64GB of DDR5-5600 CL28 today and did a (very) crude benchmark with both a single 32GB module and then both 32GB modules to see how the memory latency and the increase in RAM affects the FPS.

Specs:
i9-12900K
RTX 3090
Asus 690-A Prime MB
Samsung 980 Pro 2TB (Windows) and 960 Pro 2TB (MSFS)
Win11 with all latest updates/drivers

MSFS settings:
3440x1440, 100 render scaling
TAA
Ultra preset
Air traffic off for these tests

Each test was performed after a clean reboot. I loaded the A300 at runway 4R of DD's MDW scenery using the scattered clouds weather preset at 2:02pm. I also have DD's Chicago landmarks installed, which is known for being a performance hog. Also installed in the area are FSDT's ORD, UK2000's GYY, FlightFX's PWK, and PuffinFlight's We Love VFR. Once loading in, I let the sim sit for a minute in order to load the textures before clicking "Ready to fly". I sat on the runway until the FPS stabilized and tried the grab the highest FPS. I then closed the sim and repeated the test to ensure the results were similar.

 

Crucial 32GB DDR5-4800 CL40 (two 16GB DIMMs installed):

G.Skill 32GB DDR5-5600 CL28 (single 32GB DIMM installed):

G.Skill 64GB DDR5-5600 CL28 (both 32GB DIMMs installed):

 

You can find the full last screenshot here: https://imgur.com/KAVmrkc

Just a note, the Uncore frequency is low because I opened CPU-Z when the system was idle and I use balanced power mode when on the desktop.

I'm no benchmark expert, but overall I gained 2fps sitting on the runway. I'm inclined to believe this is due to the faster memory, as the single module (which is hindered by running in dual channel mode vs quad channel mode) still gave me a 1fps increase and did not dip below 30fps when I took the screenshot as the slower memory did.

In the air... well, the DD scenery causes microstutters as it loads, and the extra memory did not fix that. But I'll take those few extra FPS, thank you!

8 hours ago, psolk said:

Wow, I have been out of the PC building game for a bit but that is some really in depth knowledge!  Is this all specific to AMD or DDR5?  

Impressive stuff.  

3 hours ago, Maxis said:

AMD.

Not that @Sethosdoesn't know his Intel stuff as well. Gave him his props for running a delidded Intel chip and OCing the hell out of it on his previous system a few months back when a previous PC building discussion came up. 

Yeah, this is very specific to AMD, how their infinity fabric works, which is a bit unconventional. Then you have another layer of specific wisdom for the X3Ds due to their cache and their own sweet spot at around 6000MT/s. The wisdom has even changed a few times on AM5 with X3Ds as people found previous knowledge might not have been entirely correctly or it changed due to AGESA / BIOS updates allowing for more headroom or fixing previous bugs. I appreciate the kind words Maxis, I will admit though my knowledge is usually a step beyond surface level, just so I know how to squeeze the most out of my current setup, then I like to do a bit of a deeper dive.

Also means my knowledge of other platforms in the moment is usually completely non-existent, as it seems to change every generation, every chip refresh, what the best RAM overclock method is.

 

8 hours ago, Cpt_Piett said:

I was able to get 30-38-38-30 at 6000MT/s. At CL 28 the system wouldn’t boot. I tested with TestMem5 and it’s stable. Haven’t tried the sim yet.

Interestingly, even with a physical memory load of 98%, there's negligible page file usage. 

That looks great, CL28 isn't obtainable for most so don't feel too bad about that. Though you should definitely try tweaking your secondary and tertiary timings to see if you can hit what Buildzoid has. If you can copy that almost 1:1, you got an excellent setup.

Should also try messing with the PowerDown feature, listed as 'enabled' on yours. I had some problems getting it to work on older BIOS versions but seems to function now. When disabled, it means the RAM never powers down to true idle which improves latency and performance, plus RAM sip so little power, it's not something you'd notice. Think Buildzoid also mentions it in his video.

4 hours ago, jaxgator said:

Just ordered a set of G.SKILL Flare X5 Series DDR5 2x32GB 6000MT/s CL30-40-40-96 1.40V.  Should be here on Sunday.

It appears as you are quite knowledgeable in memory timings so if I can't figure it out through this discussion, I may be back with some questions.  😉

That's a solid kit and I'll gladly help. Though if you can install them, set the EXPO profile and then just copy Buildzoid's timings, you are golden

3 hours ago, Sethos said:

Though you should definitely try tweaking your secondary and tertiary timings to see if you can hit what Buildzoid has. If you can copy that almost 1:1, you got an excellent setup.

I did that a while ago and got the exact same secondary timings 👍🏻

Il’ll disable PowerDown, thanks for the suggestion!

Edited September 28, 20241 yr by Cpt_Piett

I'm trying to use as much RAM as possible. Currently:

• FlyTampa's EHAM
• Fenix with APU on
• GSX
• Active Sky
• GAIST
• We Love VFR
• FSHud with 65 total aircraft
• MobiFlight connector
• aviaServer
• REX Atmos
• YourControls (shared cockpit)
• Navigraph Charts
• Chrome with 18 tabs and playing a 8K YouTube video
• Installed a few apps
• Some simple image editing in Paint
• Optimizing the C drive using O&O
• Recording a 4K video with OBS Studio
• Editing then rendering a 4K video with Wondershare Filmora

So far up to 29GB used and 6% page file use (2.3GB). Which means that in this situation 64GB would be useful.

EDIT: Not sure how to interpret the data. According to HWiNFO, virtual memory load (physical + page file) was up to 99.8%. Apparently I've got 68GB of virtual memory. After shutting down everything except Chrome, virtual memory load is 44% (30GB) which seems high for just using Win11. 

Some nerdy stats during this crazy session:

• max CPU load 89% with a max thread load of 100%
• max CPU temp 79°
• max GPU load 99%
• max GPU temp 51°
• max VRAM usage 22.3GB

Edited September 28, 20241 yr by Cpt_Piett

1 hour ago, Cpt_Piett said:

So far up to 29GB used and 6% page file use (2.3GB). Which means that in this situation 64GB would be useful 🙂

I appreciate you posting this as I’m trying to get a sense of just what it would take to push me over 32gb.  I flew the new 2024 preorder waterbomber around Ottawa the other night and was only using half of my RAM.  That’s a pretty lightweight test though so tonight I’ll fly the Fenix out of Toronto, which is a typically heavy performance area, plus all of my usual other programs:

- beyond ATC

- Navigraph Charts

- Skalarki hardware / software

- Voltana

- FS ATC Chatter

- Air and Sea AI traffic

- FSFO

Thats about the extent of a Torture test that I can pull together.  Obviously 2024 will be a different sim so it’s not fully comparable but I hope to get a sense.

6 hours ago, Sethos said:

That's a solid kit and I'll gladly help. Though if you can install them, set the EXPO profile and then just copy Buildzoid's timings, you are golden

Thank you sir!  Looking into Buildzoid's timings now.  My mobo has 2 EXPO profiles and I am not sure what the difference is.  Is there any chance you might know and if so, should I use EXPO 1 or 2?

8 minutes ago, jaxgator said:

Thank you sir!  Looking into Buildzoid's timings now.  My mobo has 2 EXPO profiles and I am not sure what the difference is.  Is there any chance you might know and if so, should I use EXPO 1 or 2?

It doesn't really matter. EXPO 2 just contains additional tighter sub-timings over EXPO 1 but it's the same timings you'll be tweaking using Buildzoid's numbers. So you just activate an EXPO profile, just to get the speed and primary timings, restart so it can train up the RAM and then start tweaking.

21 minutes ago, Sethos said:

It doesn't really matter. EXPO 2 just contains additional tighter sub-timings over EXPO 1 but it's the same timings you'll be tweaking using Buildzoid's numbers. So you just activate an EXPO profile, just to get the speed and primary timings, restart so it can train up the RAM and then start tweaking.

Ok, great.  Thanks again for your help Sethos.

5 hours ago, Cpt_Piett said:

• max GPU temp 51°

That's a decently nice low max temp on the gpu.  Is it waterblocked?   I need to monitor my Gigabyte 4090 a little more closely.  I think I remember it being around 55-65C in a typical sim session, moving to the lower end of that spectrum when I have my fan blowing ambient air directly onto the card.

I should have gone 64 when I built this thing 1.5 years ago.  I debated it then.   I think ram prices have fallen since that time, too.

Just now, Mace said:

That's a decently nice low max temp on the gpu.  Is it waterblocked?

Just stock fans which works impressively well. I have the PC mounted on a test bench though, so there’s no airflow restriction.

5 minutes ago, Cpt_Piett said:

Just stock fans which works impressively well. I have the PC mounted on a test bench though, so there’s no airflow restriction.

Which 4090 card do you have?

14 minutes ago, Mace said:

Which 4090 card do you have?

MSI Gaming Trio X RTX 4090