Overclocking RAM, Gear 1 vs Gear 2

[freeagent]

Maybe your CPU isn't 100% happy at 6400 Gear1

Nah, its good for 6600 1:1, not really sure what to say.

Edit:

FH5 feels smoother too lol. I game at 4K.

Edit:

Not FH5, I meant FM lol.. always confuse the two..

 

[Cheeseball]

I'd like to see a study on 1% and .01% lows related to ram speed on the 9800X3D. It's hard to believe there is no benefit.

I would assume that faster data-rate should mean faster loading times, right? How can the 3D cache of 96mb can make much difference when you're loading, say, 8 gigs into ram all at once?

I get a ~5 FPS difference for the low 1% FPS between these two sets:

128 GB (64GB x 2) Crucial Pro DDR5-5600 JEDEC (CP2K64G56C46U5)
96 GB (48GB x 2) G.SKILL Flare X5 DDR5-6000 EXPO (F5-6000J3036F48GX2-FX5)

One game is Apex Legends and only drops when dropping from the dropship at 1080p480 on a RX 9070 XT. 300 FPS engine maximum. According to the AMD in-game overlay, the drop is 289 FPS on the 5600 RAM, 292 with the 6000 set.

Elden Ring at 2160p240 with unlocked FPS is the other. 92 FPS average (near Moorth Ruins in the DLC) with drops to 86 FPS for the 5600 RAM, and 89 FPS for the 6000 set. Same GPU.

 

[phanbuey]

In-game smoothness?

less stutter when it did stutter it was super mild.

 

[freeagent]

less stutter when it did stutter it was super mild.

Yup, not so much about the latency for me, though the latency is slightly better at 8000 because of some tuning, but the bandwidth gain is impressive. Also, running fast ram like that lets you use less VSOC and that can potentially get you some higher sustained CPU clocks.

 

[ir_cow]

Nah, its good for 6600 1:1, not really sure what to say.

Edit:

FH5 feels smoother too lol. I game at 4K.

More bandwidth might be the the difference here. Maybe In combo with texture streaming is seems smoother.

Just guessing

 

[ShrimpBrime]

Yup, not so much about the latency for me, though the latency is slightly better at 8000 because of some tuning, but the bandwidth gain is impressive. Also, running fast ram like that lets you use less VSOC and that can potentially get you some higher sustained CPU clocks.

I love the forum calculator! Dig this dude. I looked your rig details up and set them in with mine for comparison. Though mine is not an AMD, with this sort of comparison, it's the memory only details. These latencies are swayed obviously via different overclock settings between each user, for say in instance comparing benchmarks.

Free Agent 8000mt/s settings vs Shrimp 6800mt/s settings. Obviously by the frequency, that is direct bandwidth description, his should just be higher despite memory timing differences. If it where to be lower, the memory wouldn't be set up properly!


I can definitely lower tRAS to 68. However, through some testing and benchmarking, the instability wasn't worth the minor performance increase. And I mean fine tooth minor increase. In fact I would say the efficiency actually dropped off during longer benchmark memory intensive tasks like 7zip and Cruncher 5b and higher. Might had been the memory temperature also, FreeAgent you definitely have thermal advantages for most times of the year. From what I can tell during 3D benching, I don't need to actually run 8000mt/s But if using 7Zip and Cruncher, that is important place to be. But this doesn't really always show up in gaming.

Besides, OP is utilizing a 9800X3D Cpu. He could run pretty much any memory configuration he wants, I don't think it's going to show up in most games. Maybe some select few AAA titles, but whats a single or two FPS? Margins of errors that day?

 

[freeagent]

You are awesome my man

 

[ShrimpBrime]

You are awesome my man

Everyone is awesome!

 

[ir_cow]

Besides, OP is utilizing a 9800X3D Cpu. He could run pretty much any memory configuration he wants, I don't think it's going to show up in most games. Maybe some select few AAA titles, but whats a single or two FPS? Margins of errors that day?

I go by 3% to be causes. So 3 runs, averaged together would be a bit overkill. Either way 1-10 FPS matters little if it still only adds up to 3%. Thats the way I see it anyways.

30 = 30.9
60 = 61.8
120 = 123.6
240 = 247.2
360 = 370.8

 

[AleXXX666]

I guess I should go back to my 12700k and gtx 1070. You know, to save electricity.

to replace 12700K with 9800X3D.. umm..

 

[ShrimpBrime]

I go by 3% to be causes. So 3 runs, averaged together would be a bit overkill. Either way 1-10 FPS matters little if it still only adds up to 3%. Thats the way I see it anyways.

30 = 30.9
60 = 61.8
120 = 123.6
240 = 247.2
360 = 370.8

That's excellent information. A possible 10fps, IF the fps is already really high.

 

[Pets4Ever]

to replace 12700K with 9800X3D.. umm..

It was mainly because I was locked into a dead-end socket, with a DDR4 motherboard. If I'd had a DDR5 board I would have just picked up a 14700k because it has a larger level 3 cache than the 12700k and more performance in general. Would have been an energy hog but beastly for productivity as well as competitive in gaming.

And for what it's worth, the new setup does seem to have smoothed out Helldivers 2 somewhat.

I love the forum calculator! Dig this dude. I looked your rig details up and set them in with mine for comparison. Though mine is not an AMD, with this sort of comparison, it's the memory only details. These latencies are swayed obviously via different overclock settings between each user, for say in instance comparing benchmarks.

Free Agent 8000mt/s settings vs Shrimp 6800mt/s settings. Obviously by the frequency, that is direct bandwidth description, his should just be higher despite memory timing differences. If it where to be lower, the memory wouldn't be set up properly!


I can definitely lower tRAS to 68. However, through some testing and benchmarking, the instability wasn't worth the minor performance increase. And I mean fine tooth minor increase. In fact I would say the efficiency actually dropped off during longer benchmark memory intensive tasks like 7zip and Cruncher 5b and higher. Might had been the memory temperature also, FreeAgent you definitely have thermal advantages for most times of the year. From what I can tell during 3D benching, I don't need to actually run 8000mt/s But if using 7Zip and Cruncher, that is important place to be. But this doesn't really always show up in gaming.

Besides, OP is utilizing a 9800X3D Cpu. He could run pretty much any memory configuration he wants, I don't think it's going to show up in most games. Maybe some select few AAA titles, but whats a single or two FPS? Margins of errors that day?

View attachment 403547View attachment 403544

Could it be that the 68 TRAS is more stable for him because he left his ram at CL36?

 

[freeagent]

It was mainly because I was locked into a dead-end socket, with a DDR4 motherboard. If I'd had a DDR5 board I would have just picked up a 14700k because it has a larger level 3 cache than the 12700k and more performance in general. Would have been an energy hog but beastly for productivity as well as competitive in gaming.

And for what it's worth, the new setup does seem to have smoothed out Helldivers 2 somewhat.


Could it be that the 68 TRAS is more stable for him because he left his ram at CL36?

tRFC is at 512 as well, big bump there.

 

[ShrimpBrime]

Could it be that the 68 TRAS is more stable for him because he left his ram at CL36?

Its possible, but I was trying to show the latency difference all of .8ns. 1Ns = billionth of a second.

 

[Pets4Ever]

Its possible, but I was trying to show the latency difference all of .8ns. 1Ns = billionth of a second.

1Ns aka, my reaction time

(I wish lol)

 

[JustBenching]

I love the forum calculator! Dig this dude. I looked your rig details up and set them in with mine for comparison. Though mine is not an AMD, with this sort of comparison, it's the memory only details. These latencies are swayed obviously via different overclock settings between each user, for say in instance comparing benchmarks.

Free Agent 8000mt/s settings vs Shrimp 6800mt/s settings. Obviously by the frequency, that is direct bandwidth description, his should just be higher despite memory timing differences. If it where to be lower, the memory wouldn't be set up properly!


I can definitely lower tRAS to 68. However, through some testing and benchmarking, the instability wasn't worth the minor performance increase. And I mean fine tooth minor increase. In fact I would say the efficiency actually dropped off during longer benchmark memory intensive tasks like 7zip and Cruncher 5b and higher. Might had been the memory temperature also, FreeAgent you definitely have thermal advantages for most times of the year. From what I can tell during 3D benching, I don't need to actually run 8000mt/s But if using 7Zip and Cruncher, that is important place to be. But this doesn't really always show up in gaming.

Besides, OP is utilizing a 9800X3D Cpu. He could run pretty much any memory configuration he wants, I don't think it's going to show up in most games. Maybe some select few AAA titles, but whats a single or two FPS? Margins of errors that day?

View attachment 403547View attachment 403544

Both of your TRASes are wrong,no?

First case should be 82 + Trtp, second case should be 66+Trtp.

 

[ShrimpBrime]

Both of your TRASes are wrong,no?

First case should be 82 + Trtp, second case should be 66+Trtp.

Wrong? Maybe by opinion strictly. There's stable or unstable. I could bring it down to 82, probably stable. But I got that complex everything should end at the same number haha. So in this case, I ended everything with an 8.

 

[freeagent]

68 is stock on my sticks, I just left it..

 

[JustBenching]

That's weird, Tras is TRCD + Tcas + Trtp. You can probably drop it to a very low value and still be stable but mobos either ignore the number you put there and do their own thing or you are just skipping the refreshes. Im running mine at 28 - even though it should be 84 lol

 

[ir_cow]

RCD+RP = RAS . Anything lower will change the performance by 0%. You can run into stability issues matching it though. Not all IC are made equal and some require more voltage for the same values as another.

 

[ShrimpBrime]

I also don't run a really high Trfc either. On auto, it's around 6650, I typically don't go over 11,000. ummm well..... I should really say, this depends on what I'm doing and the benchmark specifically. Many all respond quite differently, and also a persuasion on stability with certain clocks and timings. But yes, ranging up to 65,000. HOWEVER, to Note: my board only "supports" (And totally not stable without sever loosening of the timings.) 7200mt/s. So to do 8000mt/s+, (which I have) I literally need to be at Jedec Timing specs for that rated speed and looser HAHA!!! That's the difference between a TUF and an ROG board. So for me personally, I have to do things others wont do in order to achieve at least decent results, fair to say good results. Still not more bandwidth though. Just good latency and efficiency is what I desire mostly. CL28 at 6800 costs 1.45v SA (yeah, sorry intel.) and 1.60 vdimm. Temps are max'ish 45-50c with long process times or burn in.

So I use that as my daily, I'm webbing and gaming (SupCom FA, Entropy Zero and some Physx games) as well as using this for Benchmarking video cards. So, I'm pretty fairly stable. Not happy with current NV Gpu drivers at this time, but I'm sure they are working on it. Eventually it'll be better or I'm putting the 4070 Super Duper back in...... lol

 

[freeagent]

I am bothered that this board is not stable a ball hair over 8000 with two different sets of ram..

Also bothered that Asus released a new bios saying 4x dimms is awesome, have at it boys.. when it is not.

Ugh.

 

[ir_cow]

Let me know when 4x DIMM is ever a good choice outside Epyc or Xeon.

 

[SyWillCry]

I'm on AM5, 32GB DDR5-6000 CL28. If I were to overclock the ram to 6200, would that push my memory controller into Gear 2?

Anything over 6000 will automatically have the motherboard set UCLK DIV1 mode to UCLK = MCLK/2. You can try tune it manually if you so wish. To note, higher Vsoc means a higher chance of working but, it can affect FCLK stability. Ideally you either follow the 3:2:3 (MCLK:FCLK:UCLK) ratio, or you just send FCLK as high as it'll go.

 

[c2DDragon]

You're using an X3D chip, stop trying to get 3% more out of your RAM because the X3D chip doesn't care in the slightest, doesn't need more bandwidth, and doesn't need lower latency.

The only thing that comes from pushing your memory controller to the limits on an X3D chip is increased idle power draw and potential instability that doesn't show up very easily thanks to DDR5's inherent single-bit ECC.

Essentially, you're gaining almost nothing, definitely worsening your power consumption and temperatures, and pushing your memory controller from further from the guaranteed speeds its rated for. It's like tuning an engine to within an inch of its life when all it actually needs to do it drive two miles to the shops in heavy, slow-moving traffic. Look at articles that show DDR5 scaling on 7800X3D and 9800X3D. It's damn-near pointless so the risk-reward is basically all risk, zero reward; Tune literally anything else in your system first.

You could have said this in general, not just for X3D chips (AMD)
I used to overclock RAM with Intel and it's the same, you just earn better benchmarks scores with no real benefits.
You could be happy to have a temporary world record in RAM OC. You could sell the RAM kit for better than you paid, that's it.

I guess I should go back to my 12700k and gtx 1070. You know, to save electricity.

I thought trolling wasn't allowed in here. Time flies it seems.