Ryzen Owners Zen Garden

[Nordic]

Not necessarily. The 7800X3D will never reach its thermal limit with good enough cooling. It's only X-variants that have much higher clock/power limits that they could reach without ruining into thermal limits first with any reasonable cooler.

Power/clock/voltage/thermal limits regulate every modern CPU - whichever limit it reaches first is the deciding factor on your performance.

Maybe I could have said it better. I typed quickly on my phone. I was responding to someone saying they didn't understand the desire to delid.

Power/clock/voltage/thermal limits regulate every modern CPU. I want to maximize clock speed / performance. Power is more or less a non issue on a 7800x3d. With an aggressive but stable curve optimizer profile and overkill expensive water cooling, I am maximizing the performance of my 7800x3d as shown my by 19400 R23 score. Delidding would only make it easier to achieve maximum performance.

With a Scythe Mugen 5 cpu heatsink with max speed thermaltake T30's I could not achieve 19400 R23. I probably don't need as overkill a water cooling system as I had built, but it wasn't until I installed water cooling that I was able to achieve 19400 R23. All it did was gain me 200mhz which is so little.

I built my loop overkill so I could theoretically cool a 14900k and 4090 if I ever wanted to. I have no plans to purchase that kind of hardware but my water cooling loop could keep it cool. My cooling loop will last far longer than my current set of hardware.

I have mentioned it many times but I wish I bought a motherboard with an asynchronous clock engine. I am certain I could achieve 5400mhz if not more. I got a pretty damn good 7800x3d. Maybe I will swap motherboards.

 

[kapone32]

Maybe I could have said it better. I typed quickly on my phone. I was responding to someone saying they didn't understand the desire to delid.

Power/clock/voltage/thermal limits regulate every modern CPU. I want to maximize clock speed / performance. Power is more or less a non issue on a 7800x3d. With an aggressive but stable curve optimizer profile and overkill expensive water cooling, I am maximizing the performance of my 7800x3d as shown my by 19400 R23 score. Delidding would only make it easier to achieve maximum performance.

With a Scythe Mugen 5 cpu heatsink with max speed thermaltake T30's I could not achieve 19400 R23. I probably don't need as overkill a water cooling system as I had built, but it wasn't until I installed water cooling that I was able to achieve 19400 R23. All it did was gain me 200mhz which is so little.

I built my loop overkill so I could theoretically cool a 14900k and 4090 if I ever wanted to. I have no plans to purchase that kind of hardware but my water cooling loop could keep it cool. My cooling loop will last far longer than my current set of hardware.

I have mentioned it many times but I wish I bought a motherboard with an asynchronous clock engine. I am certain I could achieve 5400mhz if not more. I got a pretty damn good 7800x3d. Maybe I will swap motherboards.

You hit the nail on the head. If you can keep them cool Am5 X3D parts will sing and laugh at frame rates.

 

[tabascosauz]

That would be weird considering that my board's default with these settings is 1.3 V.

Mine defaults to 1.3V too, but running that high shoots your performance in the foot all around - think about how high your SOC power would be. Just because it can doesn't mean it's a good idea.

I don't have to do much to replicate both problems (VSOC too low for UCLK, or VSOC too high for FCLK). Not USB issues though. However, one of my old USB docks would have similar symptoms on my Impact.

 

[freeagent]

I had usb dropouts on b550 with two boards, but none on its bigger brother.

 

[AusWolf]

You sure you don't have a broken USB cable somewhere?

I don't think so - I only have this issue in games.

Mine defaults to 1.3V too, but running that high shoots your performance in the foot all around - think about how high your SOC power would be. Just because it can doesn't mean it's a good idea.

I don't have to do much to replicate both problems (VSOC too low for UCLK, or VSOC too high for FCLK). Not USB issues though. However, one of my old USB docks would have similar symptoms on my Impact.

So basically, I should try both higher and lower VSoC, and see which value makes the dropout issue go away. Gotcha.

Maybe I could have said it better. I typed quickly on my phone. I was responding to someone saying they didn't understand the desire to delid.

Power/clock/voltage/thermal limits regulate every modern CPU. I want to maximize clock speed / performance. Power is more or less a non issue on a 7800x3d. With an aggressive but stable curve optimizer profile and overkill expensive water cooling, I am maximizing the performance of my 7800x3d as shown my by 19400 R23 score. Delidding would only make it easier to achieve maximum performance.

With a Scythe Mugen 5 cpu heatsink with max speed thermaltake T30's I could not achieve 19400 R23. I probably don't need as overkill a water cooling system as I had built, but it wasn't until I installed water cooling that I was able to achieve 19400 R23. All it did was gain me 200mhz which is so little.

I built my loop overkill so I could theoretically cool a 14900k and 4090 if I ever wanted to. I have no plans to purchase that kind of hardware but my water cooling loop could keep it cool. My cooling loop will last far longer than my current set of hardware.

I have mentioned it many times but I wish I bought a motherboard with an asynchronous clock engine. I am certain I could achieve 5400mhz if not more. I got a pretty damn good 7800x3d. Maybe I will swap motherboards.

Ah I see.

What clocks does your X3D run at with your aggressive CO? Mine does 4800 MHz all-core with no CO, but it's miles away from any kind of limit, that's why I thought that it's the hard clock limit of this CPU.

 

[tabascosauz]

Not that it makes much difference, but here's a related but probably obvious observation: Since there is an apparently valid CCD1 L3 temperature value on the single-CCD 7800X3D - valid in value, not valid in that it's actually measuring anything other than the substrate temperature at the spot - the thermal diode measuring that temperature is presumably in the packaging substrate under the CCD footprint, below where the non-3D L3 would be (or have been).

And a question for the knowledgeable: Do non-X3D parts get as much temperature delta between the cores and the CCD hotspot, ~10 celsius here, especially at idle, now that it seemed that the L3 temperature is not actually the hotspot temperature inside the V-Cache stack?

FWIW mine is manually set to the stock 1.1 V at the same clocks, and seemed stable in stress tests with no memory errors, or errant USB disconnect sounds during those. It won't take any negative CO without failing those same tests, though. Maybe it's something else?

The X3Ds even on AM4 have never reported a unique L3 temp value in HWInfo. It pretty much behaves the exact same way as any of the other CPUs, which seems to suggest that it's always been the same sensor in the CCD itself.

There are some reports that 2CCD 7800X3D is prevalent. Of course, it's easy to tell now, because it's possible to look under the heatspreader without delidding

The AM5 core temps are kinda odd, sometimes being lower than ambient temp in the mornings leads me to believe that it's AM3+ all over again and the temp sensors are inaccurate below a certain temp. Either that or HWInfo isn't keeping up with the times. AM4 didn't have this phenomenon.

I don't think so - I only have this issue in games.


So basically, I should try both higher and lower VSoC, and see which value makes the dropout issue go away. Gotcha.


Ah I see.

What clocks does your X3D run at with your aggressive CO? Mine does 4800 MHz all-core with no CO, but it's miles away from any kind of limit, that's why I thought that it's the hard clock limit of this CPU.

There are others too like VDD_MISC which all the Fabric is supposed to run on now. And VDDGs and VDDP are still around. But I don't remember any rules regarding those, too lazy to read through pages and pages on OCN

Just remember to run some TM5 as you're adjusting VSOC. It is very sensitive to UCLK stability now

 

[Nordic]

What clocks does your X3D run at with your aggressive CO? Mine does 4800 MHz all-core with no CO, but it's miles away from any kind of limit, that's why I thought that it's the hard clock limit of this CPU.

I am hitting the max boost limit in most workloads. My recent real gaming load tests showed 5040 mhz average clocks. I have more thermal headroom than the chip can boost.

 

[lexluthermiester]

And a question for the knowledgeable: Do non-X3D parts get as much temperature delta between the cores and the CCD hotspot, ~10 celsius here, especially at idle, now that it seemed that the L3 temperature is not actually the hotspot temperature inside the V-Cache stack?

The differences between 3DCache models and non-3DCache models isn't too great. Depending on silicon lottery, you can get one were the differential has a 15C spread, and some the spread is dead even. Hotspots don't matter as much as people like to think, especially considering that AMD is soldering the IHS to the dies. So as long as your cooling solution is solid and you use a quality TIM, there shouldn't be any issues with die temp differentials.

 

[JWNoctis]

The X3Ds even on AM4 have never reported a unique L3 temp value in HWInfo. It pretty much behaves the exact same way as any of the other CPUs, which seems to suggest that it's always been the same sensor in the CCD itself.

There are some reports that 2CCD 7800X3D is prevalent. Of course, it's easy to tell now, because it's possible to look under the heatspreader without delidding

The AM5 core temps are kinda odd, sometimes being lower than ambient temp in the mornings leads me to believe that it's AM3+ all over again and the temp sensors are inaccurate below a certain temp. Either that or HWInfo isn't keeping up with the times. AM4 didn't have this phenomenon.

Makes me want to pull my own out and have a look. Maybe next cleaning.

It would be weird though, if the disabled CCD would report and only report the temperature of its otherwise very dead and very inert L3 cache. I don't remember finding any other value or reference regarding the second CCD on HWInfo on my current system, though I might have missed something, and it is still plausible.

At least none of the readings on mine are obviously offset or garbage, from ambient temp to thermal throttle. I did lower the thermal limit a bit - Just in case they systematically read low? Maybe it'd live longer that way, and the performance penalty is minimal even in benchmarks.

 

[AusWolf]

Mine defaults to 1.3V too, but running that high shoots your performance in the foot all around - think about how high your SOC power would be. Just because it can doesn't mean it's a good idea.

I don't have to do much to replicate both problems (VSOC too low for UCLK, or VSOC too high for FCLK). Not USB issues though. However, one of my old USB docks would have similar symptoms on my Impact.

I had a roughly 2-hour run in Hogwarts Legacy with 1.3 VSOC. No USB dropout sounds so far. Fingers crossed the problem was that simple to solve, and it stays this way.

It looks like the IO on my 7700X isn't as good of a sample as on my 7800X3D. Oh well, we've survived worse.

 

[Ruru]

Gotta hate that during summer, temps are way higher than otherwise. Just gotta remember that those higher ones aren't "normal" temperatures.

 

[freeagent]

Dial nothing back for summer. If temps too high, cooling too bad

 

[Ruru]

Dial nothing back for summer. If temps too high, cooling too bad

Already running a full custom loop. I need to try if I can undervolt more.

 

[freeagent]

Already running a full custom loop

That doesn’t really mean anything these days, at least with AMD..

 

[lexluthermiester]

Already running a full custom loop. I need to try if I can undervolt more.

What temps are you getting? I have yet to see any consumer Ryzen that get so hot that a loop isn't working well. Threadripper, sure that can happen, but consumer Ryzen's?

 

[freeagent]

I would be surprised if his CPU is allowed to get ~200w from PBO

 

[lexluthermiester]

I would be surprised if his CPU is allowed to get ~200w from PBO

That's what I was thinking. If we're talking about the 5800X in their specs, then a custom loop should be overkill for cooling on that CPU, even if it's massively OC'd and OV'd. That's why I'm asking about temps. Perhaps they are worried that temps that are normal might be worrisome.

 

[Ruru]

What temps are you getting? I have yet to see any consumer Ryzen that get so hot that a loop isn't working well. Threadripper, sure that can happen, but consumer Ryzen's?

About ~mid 80s

I would be surprised if his CPU is allowed to get ~200w from PBO

Oh hell no, PPT is set to 110W

 

[thesmokingman]

Already running a full custom loop. I need to try if I can undervolt more.

Psst, always have to be aware of the season that you're building a loop in. This becomes an issue when a loop doesn't have extra radiator padding.

 

[Ruru]

Psst, always have to be aware of the season that you're building a loop in. This becomes an issue when a loop doesn't have extra radiator padding.

Exactly, the temps have risen due to increased ambient temps.

 

[freeagent]

Oh hell no, PPT is set to 110W

Dude.. mid 80s with 110w is terrible..

 

[Ruru]

Dude.. mid 80s with 110w is terrible..

I still guess it's reasonable with ambient ~29c

 

[freeagent]

I still guess it's reasonable with ambient ~29c

Since you put it like that

 

[Nordic]

Already running a full custom loop. I need to try if I can undervolt more.

My loop can theoretically cool 1800w which is excessive for only cooling a 7800x3d. The challenge is getting heat from the cpu to the cooling loop.

My cpu will get to high 80's when boosting to 5050mhz. High summer ambient temperature hasn't decreased my performance. If the cpu gets too hot it will throttle down.

What kind of loop are you running? Are you seeing thermal throttling or are you simply concerned about a big number?

 

[R-T-B]

Aorus X670 elite AX blackscreens if you try to disable the iGPU on latest stable bios, F30. Really Gigabyte?

True at least with 7000 series cpus, anyways.