Ryzen Owners Zen Garden

[TheLostSwede]

Strange ITX

Asus' ROG X670E-I Has a Chipset on a PCIe Stick

Asus places the Promontory 21 chipset on a PCIe card.

www.tomshardware.com

Nothing strange, they just wasted the second chipset so they could charge people more for the board.
The only thing it seems to be connected to is the second M.2 slot, which clearly could've been done just fine, without that extra chipset.

What do you think my odds are of getting a used 5800x3d first week of March when the 7900x3d starts landing on peoples doorsteps? Come on folks! I know one of you out there is itching to upgrade!

Yes, to a 5800X3D...

 

[AusWolf]

Strange ITX

Asus' ROG X670E-I Has a Chipset on a PCIe Stick

Asus places the Promontory 21 chipset on a PCIe card.

www.tomshardware.com

It's a B650 and X670 2-in-1!

 

[kapone32]

Strange ITX

Asus' ROG X670E-I Has a Chipset on a PCIe Stick

Asus places the Promontory 21 chipset on a PCIe card.

www.tomshardware.com

Didn't As Rock released one like that add in card with a Chipset.

 

[Veseleil]

It's just a beginning. Modularity is the future.

 

[Nordic]

I have been running my 5950x at 65w lately, unless I am doing something demanding. This mostly effects my MT performance.

In CB23 I get approximately the following scores:
65w - ~17k MT, ~1600 ST
95w - ~24k MT, ~1600 ST
225w - ~29k MT, ~1640 ST

Although it is only 2.5% more performance, is there a way to get 1640 ST or more performance at 65w?

 

[tabascosauz]

I have been running my 5950x at 65w lately, unless I am doing something demanding. This mostly effects my MT performance.

In CB23 I get approximately the following scores:
65w - 17k MT, 1600 ST
95w - 24k MT, 1600 ST
225w - 29k MT, 1640 ST

Although it is only 2.5% more performance, is there a way to get 1640 ST or more performance at 65w?

65W PPT is low enough to affect ST performance consistently. 2CCD is not very efficient for the job and regularly draws a lot of power for single-threaded load. 95W you may get away with if you run low SOC power and don't play very CPU-demanding games.

You shouldn't need to go to 225W to ensure ST score. Anything above 100W should be more than plenty. Run some corecycler or ycruncher limited to 1 core and see whereabouts your package power lands.

If you just ran R23 once or twice at each power setting, you need to be doing more runs and cleaning up your background processes if you want consistent scores. ST test is long and prone to being affected by every little background thing that goes on.

The rest of your limits (EDC) will affect your ST score. You can't just change only PPT and expect consistent performance and wildly different power. You need to be tweaking all 3.

iirc on 5900X I settled on 130/95/120, which with CO config applied would give me comparable MT and better ST scores than bone stock 142W.

 

[Nordic]

65W PPT is low enough to affect ST performance consistently. 2CCD is not very efficient for the job and regularly draws a lot of power for single-threaded load. 95W you may get away with if you run low SOC power and don't play very CPU-demanding games.

You shouldn't need to go to 225W to ensure ST score. Anything above 100W should be more than plenty. Run some corecycler or ycruncher limited to 1 core and see whereabouts your package power lands.

If you just ran R23 once or twice at each power setting, you need to be doing more runs and cleaning up your background processes if you want consistent scores. ST test is long and prone to being affected by every little background thing that goes on.

The rest of your limits (EDC) will affect your ST score. You can't just change only PPT and expect consistent performance and wildly different power. You need to be tweaking all 3.

iirc on 5900X I settled on 130/95/120, which with CO config applied would give me comparable MT and better ST scores than bone stock 142W.

I have a few dozen benchmark runs at different settings recorded in a table. Those were just examples.

Note: CO is applied to all cores. Each test was ran with bare bones background processes after a reset.

So theoretically if I used similar settings from the last run shown and limited the PPT, I could keep the higher ST scores with less power used. Just tune the PPT is what your saying.


Every few months, I want to play with my overclock. It wasn't until the last binge that I started recording everything in a table. I spend a day maybe two fiddling with some settings and seeing what scores I get and then get bored. Trying to optimize my CO per core was a tedious task that did not give useful results. I must be feeling the need again if I am asking questions in the Zen Garden.

 

[JrRacinFan]

Didn't As Rock released one like that add in card with a Chipset.

AFAIK for one person and it was a B650 LiveMixer. I forget who but a quick YouTube search you should be able to find it easily.

 

[tabascosauz]

I have a few dozen benchmark runs at different settings recorded in a table. Those were just examples.

Note: CO is applied to all cores. Each test was ran with bare bones background processes after a reset.

So theoretically if I used similar settings from the last run shown and limited the PPT, I could keep the higher ST scores with less power used. Just tune the PPT is what your saying.
So theo

No, what I mean is that you should work on your EDC as well, not just PPT. Affects clocks more than one would think.

I also remember the 5900X being very picky about temps during ST. I would blast fans at full speed to get higher ST score, even though temps were like 55C, and it would make a very tangible difference.

Looks like neither of us got very lucky with our CPUs. Preferred cores on my 5900X ran out of steam at about 4.93 @ 1640 ST.

 

[Nordic]

No, what I mean is that you should work on your EDC as well, not just PPT. Affects clocks more than one would think.

I also remember the 5900X being very picky about temps during ST. I would blast fans at full speed to get higher ST score, even though temps were like 55C, and it would make a very tangible difference.

Looks like neither of us got very lucky with our CPUs. Preferred cores on my 5900X ran out of steam at about 4.93 @ 1640 ST.

Swapping to custom water was a game changer. 65c under full load dramatically increased my MT scores and showed me that my -10 CO on all cores was not stable. I still wasn't able to get that ST over ~1640. I will see what I can get this time.

What is the latest methods for optimizing CO per core? I thought Core Cycler had problems. Is Y cruncher preferred now?

 

[amd64skater]

Ok so I have a 5700x. I was wondering if I can get some help with overclocking it. I have been using Ryzen Master and I'm getting a max of 4850Mhz. That is with the curve optimizer off. My system is in my spec page if that helps. Board is a Asus Rog Strix B550 f gaming non wifi.

 

[freeagent]

I just give my 5600X 200/140/180 -29 AC and it slams into the wall at 4850 on each core. But its only core cycler stable at 4800.

It doesn't use all of that power, it cant.. but I lost my paper that had my notes on it so I just give it what I used to give my 5900X

 

[tabascosauz]

Swapping to custom water was a game changer. 65c under full load dramatically increased my MT scores and showed me that my -10 CO on all cores was not stable. I still wasn't able to get that ST over ~1640. I will see what I can get this time.

What is the latest methods for optimizing CO per core? I thought Core Cycler had problems. Is Y cruncher preferred now?

Yeah, R23 is pretty definitive. If the core in question can't go any further then it just can't go any further. Perhaps you still have more CO headroom on that core, perhaps not.

Corecycler is fine as it's honestly still the most convenient way of getting an overview of all the cores (ie. run it as-is). It should at least get you to a point of not crashing under normal workloads. If you want to make sure you're 100% stable there are many options but they all will take exponentially more time and effort. Ycruncher is good but there's config involved as well. You have to figure out what tests are most effective for you.

Ok so I have a 5700x. I was wondering if I can get some help with overclocking it. I have been using Ryzen Master and I'm getting a max of 4850Mhz. That is with the curve optimizer off. My system is in my spec page if that helps. Board is a Asus Rog Strix B550 f gaming non wifi.

You're already at max. +200 limit for 5700X is 4850MHz.

Anything more will just be making boost more consistent with Curve Optimizer (bringing effective clocks up if they are not already 4850Mhz), or increasing all-core clocks by increasing power limits. Or adding a bit of BCLK on top if you're feeling adventurous.

 

[Nordic]

Used Ryzen Masters Core Optimizer test to get starting per core values.

CB23 29990 MT , 1647 ST. Looks like I did eek out a little more on that core. Ryzen master says Core 1 is my fastest but Core 4 was able to run 75mhz faster.

 

[tabascosauz]

Used Ryzen Masters Core Optimizer test to get starting per core values.

CB23 29990 MT , 1647 ST. Looks like I did eek out a little more on that core. Ryzen master says Core 1 is my fastest but Core 4 was able to run 75mhz faster.

Be careful around RM's "core optimizer test" and the numbers it spits out. A quick run through corecycler should verify whether those values are at least good enough to trust for now. -29 across 13 cores is......optimistic regardless of sample quality.

Relatively low CO offset for the Preferred Cores is normal and to be expected. Also not a surprise that a non-Preferred core can beat a CPPC Preferred core after a CO offset, but sadly it won't be of much use especially if you're on Windows 10 as the scheduler will continue to follow the hardware order.

 

[Nordic]

Those are just starting values to test with. I plan to run core optimizer over night through the next day to test. I plan to swap to Windows 11 soonish.

Benchmark results are good though. I broke 30k in MT. It ran at a smooth 4600mhz during the MT benchmark.

 

[freeagent]

1000PPT.. watts up with that?

 

[Nordic]

1000PPT.. watts up with that?

It is the "motherboard" set limits. It doesn't seem to go higher than actually using 225 PPT according to ryzen master. I can and do set the limit lower for daily use. I don't even know if I can really trust the "cpu power" value shown in my table as reported by HWinfo.

 

[freeagent]

It is the "motherboard" set limits. It doesn't seem to go higher than actually using 225 PPT according to ryzen master. I can and do set the limit lower for daily use.

I have mine set to 240/160/190

 

[tabascosauz]

Those are just starting values to test with. I plan to run core optimizer over night through the next day to test. I plan to swap to Windows 11 soonish.

Benchmark results are good though. I broke 30k in MT. It ran at a smooth 4600mhz during the MT benchmark.

Don't get me wrong, I wasn't saying that you should go to Win 11 just pointing out that Windows 10 is VERY strict about CPPC order, so in 1- or 2-core gaming loads for example they will always be on the Preferred Cores. Whereas Win 11 is a lot more loosey goosey about it.

I used to keep a spreadsheet for CO on 5900X and 5700G, but stopped doing it for the 5800X3D because, well...5800X3D

When you achieve higher clocks on a non-Preferred core, it's the biggest effing tease, like dangling a steak just out of reach of a hungry lion

 

[Mussels]

Honestly, I don't know. I don't really know what the fuck I'm doing, so I'm guessing, but still trying to be careful at the same time.

When you lock the VID (requested voltage) the CPU will not progress past the clock speed associated with it when boosting

Strange ITX

Asus' ROG X670E-I Has a Chipset on a PCIe Stick

Asus places the Promontory 21 chipset on a PCIe card.

www.tomshardware.com

I love the way this is working out with these expandable IO cards - i hope AMD does this for future boards and series where all the boards are basically B650 with a 4x 5.0 slot and optional upgrade cards exist for users to buy

AMD sells more chipsets, board makers get optional extras to sell - and end users get some crazy ass expansion cards with high end USB-4 ports, 10Gb ethernet, etc etc

 

[Blaeza]

At these settings I keep getting crashes in MW2. As in the game hangs and the reboots. What have I done wrong? I can watch youtube/twitch for hours but I try to play COD and it just hangs.

 

[Chomiq]

AFAIK for one person and it was a B650 LiveMixer. I forget who but a quick YouTube search you should be able to find it easily.

For Wendell from Level1Techs.

 

[Mussels]

View attachment 283575
At these settings I keep getting crashes in MW2. As in the game hangs and the reboots. What have I done wrong? I can watch youtube/twitch for hours but I try to play COD and it just hangs.

as said in the other thread, you turn that +125Mhz off and see if it changes anything because it's very likely its something else

 

[Shou Miko]

For Wendell from Level1Techs.

I think this one is a prototype and not for sale to test the waters but this is what AsRock got known for like AGP and PCI-E on the same motherboard back in the days.