Share your AIDA 64 cache and memory benchmark here

[The King]

What are you talking about?? "Mistaken" about what??

I stated the higher boost clock helps latency not read / write / copy.

I don't you you actually read what I posted..

Seriously try reading what I posted first.

so it's not the BCLK giving the advantage in Read / Copy in my case.

Sorry did I read this incorrectly?

I said the BCLK does increase AIDA64 numbers so it does indeed increase your read/write/copy numbers. You stated it does not?

 

[mrcardio33x]

Sorry did I miss read this?

I said the BCLK does increase AIDA64 numbers so it does indeed increase your read/write/copy numbers. You stated it does not?

Okay so my CPU (R7 7800X3D) is at BCLK 105 w/ a "boost limit" of (-) 175 MHz or 5125MHz so it's not running FULL boost of 5300MHz, because I would not be "stable" running this at LLC7 - Scaler X6, etc.

So the "latency" benefit my CPU is getting from BCLK 105 is not much beyond STOCK or 5050MHz for R7 7800X3D.

The benefit my CPU receives isn't just the BCLK 105 but the combination of LLC 7 - Scaler X6 and other BIOS settings.

Regardless, higher "boost clocks" such as the boost clocks achieved by the NON-X3D CPU's help with LATENCY not read / write / copy bandwidth.

This is why the NON-X3D CPU's achieve lower LATENCY results, higher boost clocks help here.

 

[The King]

Okay so my CPU is at BCLK 105 w/ a "boost limit" of (-) 175 MHz or 5125MHz so it's not running FULL boost of 5300MHz, because I would not be "stable" running this.

So the "latency" benefit my CPU is getting from BCLK 105 is not much beyond STOCK or 5050MHz for R7 7800X3D.

The benefit my CPU receives isn't just the BCLK 105 but the combination of LLC 7 - Scaler X6 and other BIOS settings.

Regardless, higher "boost clocks" such as the boost clocks achieved by the NON-X3D CPU's help with LATENCY not read / write / copy bandwidth.

The original question posted was about the 70K read/copy speed was not in regards to latency.

All i stated was increasing FSB to 105 will increase your AIDA64 read/copy speed. Latency was a separate issue.

Sorry if there was any misunderstanding. We went off topic enough. Enjoy the rest of your day.

 

[mrcardio33x]

The original question posted was about the 70K read/copy speed was not in regards to latency.

All i stated was increasing FSB to 105 will increase your AIDA64 read/copy speed. Latency was a separate issue.

Sorry if there was any misunderstanding. We went off topic enough. Enjoy the rest of your day.

No, I actually address the question and you simply didn't read / comprehend my post.

The member who ask the question is running an R5 7600 which has a boost clock of 5500MHz, mine (R7 7800X3D) boost to 5125MHz.

So this notion that my BCLK 105 w/ (-) 175 MHZ boost clock - 5125MHz is "helping" with Read / Write / Copy is absurd.

Next time try reading / comprehending what is being discussed before commenting / asking for others to "chime in" etc.. and stating "I know I'm right" But yeah have a nice day.

Higher Boost Clocks help with LATENCY - hence the non X3D CPU's achieve lower LATENCY numbers.

I provided some timing suggestions that may help with the members (R5 7600) Read / Copy bandwidth.

 

[InVasMani]

It depends what the MT/s was set at before applying BCLK and how far you push BCLK at the same time how relatively small or large a impact it has you effectively are overclocking memory MT/s higher unless accounting for that by reducing it a bit then bumping BCLK to closely match the same MT/s before and after pushing BCLK higher. If you don't reduce a bit you've just overclocked MT/s higher and in turn bumped up read/copy/write performance a bit in turn since MT/s impacts the most readily.

 

[mrcardio33x]

It depends what the MT/s was set at before applying BCLK and how far you push BCLK at the same time how relatively small or large a impact it has you effectively are overclocking memory MT/s higher unless accounting for that by reducing it a bit then bumping BCLK to closely match the same MT/s before and after pushing BCLK higher. If you don't reduce a bit you've just overclocked MT/s higher and in turn bumped up read/copy/write performance a bit in turn since MT/s impacts the most readily.

Increasing the BCLK on the MEMORY can help sure ( Through 6/8 core CPU's top out at ~ 70K read / copy) but the discussion was about CPU BCLK impact, and this "impact" is going to be slightly lower latency not read / copy bandwidth. BCLK on the MEMORY can also be a nightmare to stabilize.

 

[The King]

Increasing the BCLK on the MEMORY can help sure ( Through 6/8 core CPU's top out at ~ 70K read / copy) but the discussion was about CPU BCLK impact, and this "impact" is going to be slightly lower latency not read / copy bandwidth. BCLK on the MEMORY can also be a nightmare to stabilize.

So you are saying that your BCLK bus clock is separated from your Dimm channels?
This would mean increasing your BCLK clock has no effect on your DIMM frequency?
This could be possible with higher end motherboards that have external clock gen.

I have asked for clarification on this from those who have more knowledge on the matter in OC forums.
On AM4 if you Overclock BCLK/FSB it overclocks both CPU and DIMM frequency together because they are linked. You get increased read/write/copy from the increase in dimm MT frequency as a result.

BCLK OC results in lower latency because it increases both the CPU clock frequnecy and Dimm MT this is a basic fact most Overclocker already know since Pentium days. lol

I clearly demonstrated in my previous post here how increasing BCLK increased read/write/copy values in AIDA64. If I increased to 105 BCLK my read/write/copy speeds would go up.

Share your AIDA 64 cache and memory benchmark here

New best: VDD: 1.55v VDDQ/IO: 1.30v VSOC: 1.24v Still working on closing the safe mode gap of 1.9ns via OS optimization. Goal is 55.xx - Seen others with just a .2ns gap normal v. safe mode.

www.techpowerup.com

If your board supports BCLK2 then you can increase CPU frequency only according to this post.
but if you overclock BCLK1 it will mostly definitely affect your AIDA64 numbers so this is most likely what caused the confusion.

AMD DDR5 OC And 24/7 Daily Memory Stability Thread

www.overclock.net

Thanks @Det0x

 

[mrcardio33x]

So you are saying that your BCLK bus clock is separated from your Dimm channels?
This would mean increasing your BCLK clock has no effect on your DIMM frequency?
This could be possible with higher end motherboards that have external clock gen.

I have asked for clarification on this from those who have more knowledge on the matter in OC forums.
On AM4 if you Overclock BCLK/FSB it overclocks both CPU and DIMM frequency together because they are linked. You get increased read/write/copy from the increase in dimm MT frequency as a result.

BCLK OC results in lower latency because it increases both the CPU clock frequnecy and Dimm MT this is a basic fact most Overclocker already know since Pentium days. lol

I clearly demonstrated in my previous post here how increasing BCLK increased read/write/copy values in AIDA64. If I increased to 105 BCLK my read/write/copy speeds would go up.

Share your AIDA 64 cache and memory benchmark here

New best: VDD: 1.55v VDDQ/IO: 1.30v VSOC: 1.24v Still working on closing the safe mode gap of 1.9ns via OS optimization. Goal is 55.xx - Seen others with just a .2ns gap normal v. safe mode.

www.techpowerup.com

If your board supports BCLK2 then you can increase CPU frequency only according to this post.
but if you overclock BCLK1 it will mostly definitely affect your AIDA64 numbers so this is most likely what caused the confusion.

AMD DDR5 OC And 24/7 Daily Memory Stability Thread

www.overclock.net

Thanks @Det0x

Well the reason I stated “CPU” BCLK is because this was the point of the discussion (from my POV) My screen shots show ONLY a “CPU” BCLK OC not MEMORY.

My MB (ASUS ROG Strix X670E-F) allows BCLK for JUST the “CPU” which has no baring on the MEMORY speeds.

So apparently, you thought I was talking about running a 105 BCLK on both the CPU / MEMORY… Well if this is the case I don’t know what to say lol.

Regardless, I highly doubt you’d be able to run a BCLK of 105 on the MEMORY and have it run stable, it likely won’t even post.

 

[The King]

Well the reason I stated “CPU” BCLK is because this was the point of the discussion (from my POV) My screen shots show ONLY a “CPU” BCLK OC not MEMORY.

My MB (ASUS ROG Strix X670E-F) allows BCLK for JUST the “CPU” which has no baring on the MEMORY speeds.

So apparently, you thought I was talking about running a 105 BCLK on both the CPU / MEMORY… Well if this is the case I don’t know what to say lol.

Regardless, I highly doubt you’d be able to run a BCLK of 105 on the MEMORY and have it run stable, it likely won’t even post.

I have used BCLK OC extensively back in the day with my LGA1366 X5670 Xeon on my Rampage III extreme was able to go up to 108 on the FSB then.
Current limit on my B450 is just under 102 on the FSB but It can lead to data corruption or early component failure in some cases so I generally leave it at stock 100.

 

[lolek86]

tRAS: 30
tRC: 67
tRRDS: 4
tRRDL: 8
tFAW: 20
tRTP: 12

Thanks. Yeah I will tighten up those

Did you guys heard anything if those low values for tRAS and tRC still apply to M-die kits?. By "Low" i mean lower than "the rules". Asking because i had some weird, hard to repeat instability with those set up low. On the other hand that was fresh 6200 set up and it also could be voltages not pinpointed exactly.

 

[67Elco]

Would something like this be useful when trying different things with memory? Prevent corrupting your pc's drive? I assume you can install some type of OS on it. https://www.amazon.com/Satechi-M-2-...81377f39234eb&language=en_US&ref_=as_li_ss_tl

Perhaps a external SSD just for testing?

 

[67Elco]

No clue what I'm doing, but I did manage to get four sticks to run at 6000MHz without EXPO or XMP.

 

[iRmt]

Maybe someone will find these timings useful. Ryzen 5700X 4x8GB Samsung b-die vipers. No crazy voltages and just under 55ns. Enjoy! If your motherboard could do tRCD 14, you're looking at 53.5ns or so. With GDM enabled and CMD 1T I get slightly less performance, so try it for yourself. Also ProcODT should go to 36.9, haven't tested yet. Tested with OCCT as well.

 

[Tatili]

Maybe someone will find these timings useful. Ryzen 5700X 4x8GB Samsung b-die vipers. No crazy voltages and just under 55ns. Enjoy! If your motherboard could do tRCD 14, you're looking at 53.5ns or so. With GDM enabled and CMD 1T I get slightly less performance, so try it for yourself. Also ProcODT should go to 36.9, haven't tested yet. Tested with OCCT as well.View attachment 336229

Hi enthusiast, nice work out there with your Vipers beside X570 way slower in mem trace than my B550, now 1T OFF should be your goal
Guess you are not dealing with dirty run L1 & L2 Prefetcher OFF in Bios setup, you'll find better down latency 'ns at 1T than my actual BCLK profile on X3D, well known as lazier cpu indeed

Found your timmings close to mine's at 4000mhz GearON but lands in Wheas territory in hard workloads and back off, guess I can live with 3954mhz Wheas free, so happy me

 

[iRmt]

Hi enthusiast, nice work out there with your Vipers beside X570 way slower in mem trace than my B550, now 1T OFF should be your goal
Guess you are not dealing with dirty run L1 & L2 Prefetcher OFF in Bios setup, you'll find better down latency 'ns at 1T than my actual BCLK profile on X3D, well known as lazier cpu indeed
View attachment 336368
Found your timmings close to mine's at 4000mhz GearON but lands in Wheas territory in hard workloads and back off, guess I can live with 3954mhz Wheas free, so happy me
View attachment 336369

Hello to you to Thanks. Man your vipers are better binned than mine. Should've had these instead of 4400c19 ;D anyways... I play games, so my goal is to OC for better gaming xp. I can boot up to 4000 but with Wheas all over and my mobo is a cheap x570. Another thing might suggest something about you OC. Since we have similar timings, but your frequency is higher and I get better latency. That might suggest that you're out of your sweet spot I have a question for you, how do you know this: "beside X570 way slower in mem trace than my B550"
I recently oc'd ram on 5600g so i reached 4800mt/s-2400 infinity fabric and i noticed one thing that going from 4000cl18 to 4733cl20 gives basically the same latency(somwhere in the middle there are frequencies that gives better performance than the one before or one after will give). But at 4800cl20 latency improved by around 3ns. This was truly a satisfying oc I know that I took a picture only a minute after boot. But PC was on for half an hour whea free. Ripjaws 4000c18 ram, but better mobo than my x570

 

[Tatili]

Hello to you to Thanks. Man your vipers are better binned than mine. Should've had these instead of 4400c19 ;D anyways... I play games, so my goal is to OC for better gaming xp. I can boot up to 4000 but with Wheas all over and my mobo is a cheap x570. Another thing might suggest something about you OC. Since we have similar timings, but your frequency is higher and I get better latency. That might suggest that you're out of your sweet spot I have a question for you, how do you know this: "beside X570 way slower in mem trace than my B550"
I recently oc'd ram on 5600g so i reached 4800mt/s-2400 infinity fabric and i noticed one thing that going from 4000cl18 to 4733cl20 gives basically the same latency. But at 4800cl20 latency improved by around 3ns. You can find a photo here : https://www.techpowerup.com/forums/...y-benchmark-here.186338/page-112#post-5203080

Hi thx, to answer your question, YES I know it from all over the OC's forums member OCN, HardwareluXX, reddit can confirm better mem trace OC for B550 than X570
And secondary keep in mind not to compare in latency 'ns 5700X vs 5800X3D which does not much care about, only for higher mem freq & bandwidth for his internal huge Vcache, I'am only playing this 2 games, here my Rog Swift 1440/144hz All max. out preset+RayTracing CB

 

[iRmt]

Hi thx, to answer your question, YES I know it from all over the OC's forums member OCN, HardwareluXX, reddit can confirm better mem trace OC for B550 than X570
And secondary keep in mind not to compare in latency 'ns 5700X vs 5800X3D which does not much care about, only for higher mem freq & bandwidth for his internal huge Vcache, I'am only playing this 2 games, here my Rog Swift 1440/144hz All max. out preset+RayTracing CBView attachment 336377
View attachment 336379

Oh i didn't knew that x3d needs more bandwidth than lower latency. Then in that case all makes sense

 

[Tatili]

Oh i didn't knew that x3d needs more bandwidth than lower latency. Then in that case all makes sense

All fine, please take your time and check worldwide Zen RAM OC Leaderboards here cheers

 

[iRmt]

All fine, please take your time and check worldwide Zen RAM OC Leaderboards here

If you have any suggestions on improving my timings further, I would be happy to hear that.

 

[Gica]

Pressed by mistake "1.23" instead of "1.33" at the RAM voltage and ... it works!
Apparently stable.

 

[mrcardio33x]

Pressed by mistake "1.23" instead of "1.33" at the RAM voltage and ... it works!
Apparently stable.

View attachment 336453

Looks like pretty decent RAM needing just 1.23v VDD for 10ns first word latency.

Probably a good kit to OC.

I have pretty good DDR5 that needs no more than 1.55v VDD for 6200 MT/s CL26 - 8.38ns

 

[Zyll Goliat]

Ok I managed to tweak my memory a bit more so now work and it's fully stable on 2800Mhz also I tighten timings a bit so it is clearly faster 'tho it wasn't that easy because this is reg ecc that is 2133Mhz on stock speed......

Old 2666Mhz memory speed in attached file bellow.......

 

[67Elco]

 

[mrcardio33x]

View attachment 336538

Any reason to run tRAS so high? You could probably set tRAS to 28-30..

Here's my best at 6000MT/s CL28: - With a Non X3D CPU your latency should be a bit lower due to higher boost clocks. I'm only boosting to 5125MHz.
Hynix Adie 2X16GB

 

[67Elco]

Any reason to run tRAS so high? You could probably set tRAS to 28-30..

I just got this pc back together and threw this memory in to see if it would even boot. I'll fiddle with it soon.