Ryzen AM5 ; 7000

[Neilski]

I used TestLimit from sysinternals and allocated 25GB of my ram with it.
When choosing 30GB, Windows decided to shove 25 of it into the page file...

I then took a stopwatch with 5s pre-time to really hit the 0 when starting.
With only 6GB ram used, going to sleep took 4.5s, waking up took 5.2s.
With 31 GB used, going to sleep took 4.2s and waking up took 5.1s.

I keep forgetting how many useful tools are in the Sysinternals suite!
And yeah, it's hard to make sense of those results. Did you ask Testlimit to touch the RAM or just allocate it? There are a shedload of command-line options and I'm not familiar with the tool, so hard to be certain quite what it's doing. It's conceivable that if Windows "knows" the pages have been allocated but not written, it won't attempt to preserve their contents.
I realise that I'm also not quite sure what Windows does wrt hibernation anyway - e.g. if any pages have been written to pagefile.sys, they may not NEED to then be written to hiberfil.sys as well...

Now the question is if it always writes the full 32GB onto my nvme, which would explain the slightly more than 4s.

It's really hard to believe that it would do that, but yeah it's also compatible with your tests.

I guess I'll have to clone my boot+win partitions to one of the 2.5" laptop HDDs I still have lying around and repeat my testing

LOL, if you have the stomach for that, it'd be fun to see the outcome (Is there a way to slow down the NVMe drive tho? Make it PCI Gen 1 or something?)

 

[RasmusP]

I realise that I'm also not quite sure what Windows does wrt hibernation anyway - e.g. if any pages have been written to pagefile.sys, they may not NEED to then be written to hiberfil.sys as well...

Definitely not! hiberfil.sys files are either the size of the physical RAM or smaller, no matter how big you set the pagefile.sys.
At least from my experience...

LOL, if you have the stomach for that, it'd be fun to see the outcome (Is there a way to slow down the NVMe drive tho? Make it PCI Gen 1 or something?)

Sadly not.. I also already cloned the full nvme to an old HDD while working anyway but sadly I can't boot from it. I can select it in the bios but it will then always use the nvme anyway.
I also can't disable the M.2 slot, so the nvme would need to be taken out...

My PC sits beneath my glass desk, which has some "fancy V-leg", next to the wall of the room.
So I'd have to unplug everything and slide the case out. Doable, but sadly the nvme sits below the GPU and the CPU cooler so I'd have to take out the GPU.
Still doable, but sadly the GPU needs an extra stabilizer arm, which uses 3 slot-screws.
And that I'm not willing to do just for this experiment

With the stock 5600X cooler before I modified the Thermalright frame to fit AM5:

How it looks now:

 

[Neilski]

How it looks now:

Wow, that's a tight squeeze
And no worries on the testing - you've already gone above and beyond, so it will just have to remain a mystery for a little while longer

 

[RasmusP]

Wow, that's a tight squeeze

Nah not really, it's just the perspective! The gap between GPU backplate and CPU cooler is about 2.5 cm, the CPU cooler has just more height.
I could unscrew the nvme but there's no chance I would get it out from there or back in

 

[Trebormoore84]

AMD Ryzen 7000X3D series coming February, 16-core Ryzen 9 7950X3D features 144MB cache - VideoCardz.com

AMD announces its upgraded Ryzen 7000 desktop CPUs Today AMD unveils its plans for Ryzen 7000 CPU series featuring brand new 7000 non-X and X3D series. AMD Ryzen 7000 series will be updated with 6 new CPUs. The company’s CEO Dr. Lisa Su confirmed today that Ryzen 7000 non-X will include Ryzen 9...

videocardz.com

https://www.pcgamer.com/amds-new-3d-v-cache-cpus-are-hella-quick-but-theres-no-clear-pick-for-gamers/

So the 3D stacked chips come out February, but no idea on price yet.

Also the 16 and 12 core parts are using a hybrid design, including one chiplet with V-cache and the other a normal zen 4 chiplet design that boosts higher. Sounds like a nightmare to get windows to schedule for that tbh. But for a workstation chip this is very interesting as it should overcome the problem of using slow memory in large quantities, which is a problem on both AMD and Intel at the moment if you need more than 64gb.

However the interview that PC World did with an AMD rep talked about how they are enabling the curve optimiser on these 3D chips this time which should allow better overclocks. This was disabled on the 5800x3D.

I am very interested in seeing how say a 7800x3D with a standard 6000mhz ram kit performs and what the cost is. I'm sure it won't beat a tuned 13900k with an expensive 7800-8000mhz ram kit. But if it's like half the cost and within a few percent who cares!

 

[Bob Gnarley]

AMD Motherboard Partners Remove AMD AGESA 1.0.0.4 BIOS Firmware Due To Major Issue With Ryzen 7000 CPUs​

https://wccftech.com/msi-asrock-roll-back-amd-agesa-1-0-0-4-bios-firmware-as-it-disables-cores-on-ryzen-cpus/

 

[blekenbleu]

AMD 3DV cache asymmetry without hardware support for task scheduling
7900 and 7950 X3D CPUs have stacked cache on only 1 (of 2) chiplets,
requiring (Windows 11) OS support to appropriately schedule threads
benefitting from massive cache vs threads benefitting more from clock boosting.
Microsoft support commitment is unclear.
(Windows 10 hold-outs should presumably consider only 7800X3D).

 

[Novopaine]

AMD Motherboard Partners Remove AMD AGESA 1.0.0.4 BIOS Firmware Due To Major Issue With Ryzen 7000 CPUs​

https://wccftech.com/msi-asrock-roll-back-amd-agesa-1-0-0-4-bios-firmware-as-it-disables-cores-on-ryzen-cpus/

That's unfortunate. If this was a pre-release BIOS for all the new chips I hope it doesn't delay the launch.

I also noticed that WCCF is still plagued with trolls in the comment section. I stopped visiting their site years ago because of it. It looks like nothing there has changed. Very unprofessionally run. No accountability there.

 

[Spinelli]

7000x3d is destroying 13900Ks in leaked benchmarks. With the 13900KS probably only being 0-3 % faster in games than the 13900K, it looks like Intel will have no answer.

 

[Rebel_488]

@RasmusP which ram slots are u using? And are u still at 4800, iv tested 5600 and 6000 in expo, curious to know what u r using these days?
If I select 4800 it just says ddr5 4800, whereas for 5600/6000 I could choose xmp or expo. Cheers

 

[RasmusP]

Hi, I'm using the "recommended" Ram Slots A2 + B2, which are the ones further away from the CPU:

And these were my BIOS settings up until a few days ago:

Loaded the EXPO1 profile of my 5600CL16 sticks but raised the ram voltage to 1.30V! That's quite important, wasn't stable at 1.25V! These SK Hynix sticks want 1.3-1.35V for 5600-6000 MHz from what I've read!

I then enabled the two little tuning settings "Low Latency Support" and "High Bandwidth Support". They lower 2-3 timings, nothing fancy, but runs stable.

Then it's quite important to set the ode to UCLK=MEMCLK. It's basically "Gear1" or "1:1" mode instead of "1:2" mode.

And then I had the Inifity Fabric set to "auto", which sets itself to 2000 MHz for some reason... That works well for 6000CL30 => 6000 -> 3000 -> 2000. The 3 and 2 fit into the 6.

However I had my ram at 5600 to still have Sleepmode working!

But now I've read about needing to have the Infinity Fabric synced with the ram clock.
2000 doesn't fit into 5600 at all. It would need to be set to 1866.66 MHz, which isn't available in my BIOS.

Soooo, I've lowered the RAM to 5400 MHz and set the Infinity clock to 1800 MHz.

Looks like this in Ryzen Master now (I only use it to read out, not to set anything! Otherwise I need to do a full bios-reset to get rid of the settings...):

 

[Bob Gnarley]

Then it's quite important to set the ode to UCLK=MEMCLK. It's basically "Gear1" or "1:1" mode instead of "1:2" mode.

This^^. Higher frequencies can be achieved with Command rate at 2T but the latency penalty leaves you worse off.

Just wondering Rasmus, now that the 7800X3D has been announced for April, have your hands started itching yet? You'd get a good trade in with the 7600X...

 

[Rebel_488]

Hi, I'm using the "recommended" Ram Slots A2 + B2, which are the ones further away from the CPU:

View attachment 637242
View attachment 637243


And these were my BIOS settings up until a few days ago:
View attachment 637244

Loaded the EXPO1 profile of my 5600CL16 sticks but raised the ram voltage to 1.30V! That's quite important, wasn't stable at 1.25V! These SK Hynix sticks want 1.3-1.35V for 5600-6000 MHz from what I've read!

I then enabled the two little tuning settings "Low Latency Support" and "High Bandwidth Support". They lower 2-3 timings, nothing fancy, but runs stable.

Then it's quite important to set the ode to UCLK=MEMCLK. It's basically "Gear1" or "1:1" mode instead of "1:2" mode.

And then I had the Inifity Fabric set to "auto", which sets itself to 2000 MHz for some reason... That works well for 6000CL30 => 6000 -> 3000 -> 2000. The 3 and 2 fit into the 6.

However I had my ram at 5600 to still have Sleepmode working!

But now I've read about needing to have the Infinity Fabric synced with the ram clock.
2000 doesn't fit into 5600 at all. It would need to be set to 1866.66 MHz, which isn't available in my BIOS.

Soooo, I've lowered the RAM to 5400 MHz and set the Infinity clock to 1800 MHz.

Looks like this in Ryzen Master now (I only use it to read out, not to set anything! Otherwise I need to do a full bios-reset to get rid of the settings...):
View attachment 637254

Thanks man, great reply as always. I’m using the same slots as per the mobo manual. I just wanted to be sure. I need to find a better setting, iv upgraded my fans and put 9 bequiet fans in so temps are much better, iv went with pbo enabled and -28 with temp capped at 85 and 6000 expo which is stable but obv not the best settings as per my cinebench score. Ram timings changes scared me as did the voltage changes so I will have a better look now.

Iv always had a few issues loading rf2 so I know my settings are not stable enough

 

[Rebel_488]

I then enabled the two little tuning settings "Low Latency Support" and "High Bandwidth Support". They lower 2-3 timings, nothing fancy, but runs stable.

I don’t see this option anywhere in my bios? (Asrock x670)?

 

[RasmusP]

Just wondering Rasmus, now that the 7800X3D has been announced for April, have your hands started itching yet? You'd get a good trade in with the 7600X...

Nah not really. The 7600X is as big of a jump from my 10600k, as the 10600k was from my old i7 2600k!
So I have plenty of headroom right now
But I will probably get a 8800X3D in 2-4 years

pbo enabled and -28

You're running PBO enabled but applied -28 for the curve optimizer?
That's a lot! My 7600X wasn't stable at -20!

Ram timings changes scared me as did the voltage changes so I will have a better look now.

Just put the RAM voltage to 1.35V. Is it SK Hynix? Then it should run stable at 1.35V and 5600-6000 MHz.
Everything up to 1.35V is totally fine for all RAM sticks

I then enabled the two little tuning settings "Low Latency Support" and "High Bandwidth Support". They lower 2-3 timings, nothing fancy, but runs stable.

I don’t see this option anywhere in my bios? (Asrock x670)?

Maybe AsRock doesn't have these settings. They don't really matter anyway.. It's a tiny difference really!

 

[Rebel_488]

You're running PBO enabled but applied -28 for the curve optimizer?
That's a lot! My 7600X wasn't stable at -20!

Just put the RAM voltage to 1.35V. Is it SK Hynix? Then it should run stable at 1.35V and 5600-6000 MHz.
Everything up to 1.35V is totally fine for all RAM sticks


Maybe AsRock doesn't have these settings. They don't really matter anyway.. It's a tiny difference really!

Yip, lol, I was at -30 but I got random crashes so dialed it back to -28. Seems to hold but iv been watching skatterbencher all day, (I think that’s his yt name) and he was at -12. I know it’s chip dependant so maybe I’m lucky (ish). Or I need to go lower to score better in benchmarks?

When I expo 6000 it puts the voltage to 1.35 so that’s ok. Yeah Hynix, ram is 2x32g tyson fury. I’m not sure 6000 is the correct way to run them but again it seems stable?

Sorry dude iv came from a 3600x that I never even updated the bios on so this is all new to me, terrifies me tbh lol, but I’m gettin braver. After updating the bios I grew brave anyway haha.

I’m running the ram in aggressive. My other options were agesa or competitive, iv no idea what the best choice there would be?

I don’t expect the x3d gains to be as big this gen so upgrading is not in my mind either, a better cpu cooler I think is the final touch and perhaps a higher wattage psu as even with an 850w iv hit the power limit the odd time.

Edit have u manually changed the timings, I assume that’s not for a novice like me?

 

[RasmusP]

Seems to hold but iv been watching skatterbencher all day, (I think that’s his yt name) and he was at -12. I know it’s chip dependant so maybe I’m lucky (ish). Or I need to go lower to score better in benchmarks?

But you mentioned that it isn't stable when loading rF2?
Don't use benchmarks. Only benchmark that would be useful:
Cinebench r15 with 3 threads.
Single thread is useless as EVERYTHING in reality might be only using 1 big thread, but always a few more background threads.
Full multithreading is useless too though, since nothing apart from rendering videos or similar will use all cores.

Cinebench r15 doesn't use AVX instructions, so it's closest to our simracing titles.

Lowering the curve optimizer will give you more headroom regarding the power limit, but simracing really shouldn't hit the power limit...

When I expo 6000 it puts the voltage to 1.35 so that’s ok. Yeah Hynix, ram is 2x32g tyson fury. I’m not sure 6000 is the correct way to run them but again it seems stable?

Seems correct! I didn't read much about 32GB sticks though. They seem to be slightly less stable than the 16GB sticks. The expo profile should be working at 1.35V though

I’m running the ram in aggressive. My other options were agesa or competitive, iv no idea what the best choice there would be?

Good question.. No idea..

I don’t expect the x3d gains to be as big this gen so upgrading is not in my mind either, a better cpu cooler I think is the final touch and perhaps a higher wattage psu as even with an 850w iv hit the power limit the odd time.

Your power supply has no idea about the power limit of parts and the parts have no idea about your PSU.
If the PSU reaches its limit, your PC will SHUT OFF INSTANTLY.

The power limit for the CPU is set by the motherboard, in the BIOS.

No one knows how big the 3D gain will be. It might be just as massive as with the 5800X3D!

Edit have u manually changed the timings, I assume that’s not for a novice like me?

I manually changed them, yes. I used this site for it:

The big Ryzen 7000 Memory and OC Tuning Guide - Infinity Fabric, EXPO, Dual-Rank, Samsung and Hynix DDR5 in Practice test with Benchmarks and Recommendations | igor´sLAB

AMD’s new Ryzen 7000 desktop CPUs, based on the Zen4 micro-architecture, still use the same chiplet design as their predecessors, with a few small but not negligible changes. Igor already gave details…

www.igorslab.de

I first set everything via Ryzen Master but lost the sleep mode due to going beyond 5600 MHz.

So I wanted to get back to the expo profile but couldn't.
Ryzen Master can't load expo profiles and the BIOS can't overwrite what Ryzen Master set, lol.

So I did a BIOS reset and everything was back at defaults.
I then "translated" all the Ryzen Master timings into what I could find in the BIOS.
Some things I had no idea and had to "guess". Booted into Windows, opened Ryzen Master and checked if I guessed correctly (that felt risky, lol, but I first only changed the suspicious settings by only 1 digit).
After having them all figured out, I adjusted them to Igor's 6000CL30 timings with Infinity Fabric at 2000 MHz and stored them in a BIOS profile.
Then loaded the expo profile, lowered the RAM to 5400 MHz and set the Infinity Fabric to 1800 MHz.


I can send you my Bios photos later tonight. Not on my PC right now

 

[Rebel_488]

Sorry I meant 2x16 ram. Cheers for the link and in advance for the bios pic, yes please. Good info about the power supply/power limit and guide about timings. A lot to look into.

 

[Bob Gnarley]

Yip, lol, I was at -30 but I got random crashes so dialed it back to -28. Seems to hold but iv been watching skatterbencher all day, (I think that’s his yt name) and he was at -12. I know it’s chip dependant so maybe I’m lucky (ish). Or I need to go lower to score better in benchmarks?

Setting too aggressive a curve for your chip will likely result in clock stretching.

Taken from SkatterBencher's PBO2 guide:

"Clock Stretching is a safety feature that is built into all AMD Ryzen CPUs. When the CPU thinks the actual voltage is too low to sustain a stable system at a given frequency, it will reduce the clock period until the voltage is back at the acceptable level.

The result of clock stretching is that while the system will report the Core Frequency as usual, the effective clock will be lower. So, since a lower amount of clock cycles occur the performance will also be lower.

Clock Stretching has been a major topic among AMD Ryzen enthusiasts. Typically, users would find that setting a higher frequency would result in lower performance. The best way to check if clock stretching is happening is by checking the Effective Clock measurement in HWinfo. If the effective clock is significantly lower than the configured core frequency when the system is under load, then it is likely that clock stretching is happening on your system."

You may have already been using it Rebel but here is the full guide for anyone that might find it useful.

AMD Precision Boost Overdrive 2 - SkatterBencher

Precision Boost Overdrive 2, is AMD's overclocking toolkit to adjust the parameters governing the Precision Boost 2 algorithm.

skatterbencher.com

 

[Rebel_488]

No and again, thanks. If I’m this clueless and there’s no settings I can just copy, I probably shouldn’t be messing to much with this stuff until I learn how to test it out properly.