No Manual Overclocking for AMD Ryzen 7000X3D Series CPUs

Tsing

The FPS Review
Staff member
Joined
May 6, 2019
Messages
12,634
Points
113
Ryzen users who prefer to take the more intricate and/or challenging route of unlocking the maximum performance of their processors may be mildly disappointed with the new Ryzen 7000X3D Series, according to AMD, which told members of the press that while the Ryzen 7 7800X3D and other new CPUs can be automatically overclocked with Precision Boost Overdrive (PBO) and tweaked further via Curve Optimizer, manual overclocking remains off the table.

See full article...
 
yeah yeah yeah ..... said the same about the 5800X3D....we all know that can be overclocked.

I'll be honest though. I have lost all interest in overclocking. I do run custom loop cooling to keep temps and noise down. But CPU's and GPU's are nearly pushed to their limits out of the box. I see no point in spending hours on my time chasing an extra 5%.

The days of gaining 20-40% from overclocking are over.
 
yeah yeah yeah ..... said the same about the 5800X3D....we all know that can be overclocked.

I'll be honest though. I have lost all interest in overclocking. I do run custom loop cooling to keep temps and noise down. But CPU's and GPU's are nearly pushed to their limits out of the box. I see no point in spending hours on my time chasing an extra 5%.

The days of gaining 20-40% from overclocking are over.

I overclock GPU's these days, but CPU's seem almost pointless.

There is so little to be gained, and most of the time I am GPU limited anyway.

I still find the loss of this ability a little disappointing.
 
I overclock GPU's these days, but CPU's seem almost pointless.

There is so little to be gained, and most of the time I am GPU limited anyway.

I still find the loss of this ability a little disappointing.

I don't see the point with GPU's either. A couple hundred megahertz here and there gets you, what, an extra 5-10 fps? If my card is suffering at playing a game where 5-10 FPS matters it's time to buy a new card.
 
Overclocking is dead, there I said it.

PBO is the way to Go.

There isn't much to elaborate on, I feel CPU overclocking is on the way out, the trend is gone, and so is the headroom. CPUs are now coming maxed out, at their limits, and there just isn't any headroom. With things like AMD PBO, and Intel's dynamic clocking these CPUs today can eek out that last little bit, automatically. It's just best to let them dynamically clock to begin with. Manual overclocking isn't worth it anymore, dynamic clocking is the future.
 
Overclocking is dead, there I said it.

PBO is the way to Go.

There isn't much to elaborate on, I feel CPU overclocking is on the way out, the trend is gone, and so is the headroom. CPUs are now coming maxed out, at their limits, and there just isn't any headroom. With things like AMD PBO, and Intel's dynamic clocking these CPUs today can eke out that last little bit, automatically. It's just best to let them dynamically clock to begin with. Manual overclocking isn't worth it anymore, dynamic clocking is the future.
Agreed, but that doesn’t mean there isn’t anything for a tweaker to do. PBO offsets and cooling still make a difference - my 5950 is much, much better on an oversized custom loop that it was with a D15. The CPU is so bad I could probably RMA it for crashing under stock configuration when running folding at home on a D15, but I haven’t gotten around to ripping apart the loop.
 
Overclocking is dead, there I said it.
I'm so happy I just saw this from someone (maybe the only one) I truly respect. I've felt that way even before these modern processors. Felt that trying to squeeze just a bit more with the heat and power draw just weren't worth the effort to me if it meant a crash during a game (or really anything else.) I'm on a 7700k now and these just might be the ticket to upgrade. I have no desire to OC. And, as you've as much said, setting everything to auto is the way to go anyway.
 
I met with AMD today and we chatted a bit about overclocking (or lack thereof) on the X3D's. They did emphasize that the curve optimizer function was the way to go, but to them, the victory would be more along the lines of undervolting to achieve the same performance at a lower power/heat profile - basically taking away the built in factory margin for this.

The other thing on these X3D chips is that they do not seem quite as sturdy as the non X3D's - the 120W TDP vs even a 135W TDP would cause you to hit other (i.e. thermal) limitations before you ever hit that 135W level. I get the sense they've gotten most everything you can get out of the chip.

I'm going to write this up in an actual article, but to preview for you - something that wasn't clearly articulated in the keynote is how the 79xxX3D chips are organized - the 7950x3d will have 1 CCD with 3D cache, and one without (8 cores per CCD), and the 7900X3d the same, but with 6 cores per CCD. The 7800X3D will have a single CCD. They will use the their preferred core technology to direct the workload to the cores best suited for it - i.e. games to the X3D CCD or single threaded productivity apps to the non-X3D CCD. This should make the 79xxX3D chips close the gap a bit in non-gaming applications with its non-X3D counterpart, but still, due to the lower TDP, things like Cinebench will run faster on the non-3D cache chips.
 
Agreed, but that doesn’t mean there isn’t anything for a tweaker to do. PBO offsets and cooling still make a difference - my 5950 is much, much better on an oversized custom loop that it was with a D15. The CPU is so bad I could probably RMA it for crashing under stock configuration when running folding at home on a D15, but I haven’t gotten around to ripping apart the loop.
Heh, I have the opposite problem. Running F@H or WCG keeps my 5800x from crashing. If something isn't hitting multiple cores pretty hard the CPU likes to have stability issues. Using CO makes things even worse because no matter how dialed in it seems to be, it's never stable. Multiple times have I run different programs to get a starting point for CO settings and I get vastly different numbers. At one point early on I thought I had it dialed in and the CO offsets were very conservative (since my 5800x was a turd from the beginning) but it didn't last more than a few months before instability started. Using the +50-200mhz extra boost is also a no go. I know the CPU is bad but I need to find a repeatable way to show it before I try an RMA. There's a program called Core Cycler which only hits one core at a time with Prime95 and it shows at least three of the cores will fail even at pure stock settings. OCCT's burn in for cores has also shown the same cores as bad under the same conditions but it's not repeatable.

Anyway, most everyone has it right that CPU overclocking is basically dead for the foreseeable future. Intel and AMD are already pushing the limits on the CPUs from the factory. As much as I've enjoyed tweaking and overlocking in the past, it's a good thing they do this now. It saves a lot of time and effort for very little gain with the way CPUs are now. The binning difference is usually just the number of cores which is far different from the days of CPUs being identical with the exception of binning and clockspeeds. There was good reason to overclock back then because the performance difference could be very large. 50%+ overclocks on Q6600s comes to mind. A 3%-5% overclock on most current CPUs doesn't amount to anything noticeable and the increased power and heat usually isn't worth the small gain you might get. Even for that small gain it usually comes with the downside of having worse single core or multi-core performance, probably making whatever gains moot in the long run.
 
I'm going to write this up in an actual article, but to preview for you - something that wasn't clearly articulated in the keynote is how the 79xxX3D chips are organized - the 7950x3d will have 1 CCD with 3D cache, and one without (8 cores per CCD), and the 7900X3d the same, but with 6 cores per CCD. The 7800X3D will have a single CCD. They will use the their preferred core technology to direct the workload to the cores best suited for it - i.e. games to the X3D CCD or single threaded productivity apps to the non-X3D CCD. This should make the 79xxX3D chips close the gap a bit in non-gaming applications with its non-X3D counterpart, but still, due to the lower TDP, things like Cinebench will run faster on the non-3D cache chips.
Oh dang, well hhhhhmmmmm...
 
Agreed, but that doesn’t mean there isn’t anything for a tweaker to do.
"Tweaking" is the new overclocking. Focusing on freeing clock headroom for the built-in boost algorithms, either by lowering voltage at particular clockspeeds without compromising stability, or by increasing cooling performance.

Kind of wish we had better tools for direct-die cooling.
 
I can say that after having had a 5800X3D for almost a year now that it's been a great experience. Power wise it has mostly held 50-60W usage during most games while some that are more CPU intensive will push it to 65-75W. As I don't really use it for much anything else CPU benchmark tools are the only things that I've seen push it to 85W. In terms of cooling the 360 AIO, I got (per David's review) has been a champ. I mainly keep it at stock and it purrs along all day long at 4450 MHz but under heavy loads might drop into the 4300 MHz range if I keep the AIO settings at stock but can easily crank it up to keep it at 4450 MHz.

I'm pretty sure that I'm aiming at getting the 7900X3D but still too early to tell and there's a good chance I'll be looking for a 420mm for it (I'm sure a 360 would suffice but I want to ensure it's as quiet and maxed out as possible). I agree that the new approach for CPUs (at least AMD) is to let the motherboard do its thing and keep 'em cool and feed them whatever they need (ram+power) to hit their maximum potential. Overclocking as we once knew it is no longer needed. It was a fun ride though.
 
From reviews of the 7950x (non-X3D part), setting a lower power target/profile seemed like the winning move to me. Much less energy/heat for almost the same level of performance, which sounds quite similar to what @David_Schroth relayed from his conversation with AMD in his post above.

AMD, Intel, and motherboard manufacturers seem pretty aggressive with their defaults these days. It's tempting to say too aggressive.
 
Last edited:
From reviews of the 7950x (non-X3D part), setting a lower power target/profile seemed like the winning move to me. Much less energy/heat for almost the same level of performance — essentially what @David_Schroth relayed from his conversation with AMD in his post above.

AMD, Intel, and motherboard manufacturers seem pretty aggressive with their defaults these days. It's tempting to say too aggressive.

Power is usually kept a bit on the higher side for stability reasons. We still play the silicon lottery when it comes to being able to under volt and maintain clocks.

But, yeah, undervolting to free up some power/temp limit can open up some extra megahertz.

Personally, I don't care to even try. Turn on PBO and move on with my life.
 
I think the last serious CPU I overclocked was the 2600k. Haven't really bothered much with overclocking since then. The last GPU I overclocked is my 3090.
 
Power is usually kept a bit on the higher side for stability reasons. We still play the silicon lottery when it comes to being able to under volt and maintain clocks.

But, yeah, undervolting to free up some power/temp limit can open up some extra megahertz.

Personally, I don't care to even try. Turn on PBO and move on with my life.
I'm not a fond of inputting arbitrary values, either. The settings I was referring to are controllable through PBO. Scroll down/find the heading that begins with "Ryzen 9 7950X at 65 W (ECO Mode)" on the following page:

Edit: I may have misunderstood what David was hinting at from his conversation with AMD. I'm not familiar with AMD's "curve optimizer" other than by name, nor its interaction with other settings — a consequence of having been on Intel since forever. What I was thinking of were essentially predefined profiles that didn't require manual tweaking.
 
Last edited:
The key is going to be Windows directing games to the correct CCD with the cache. If it misses, then all of a sudden you won't see a benefit from the cache, so better hope it gets it right every time.

In addition, I have a feeling the boost frequencies will be different based on which CCD is being used, the one without cache is the one that will boost higher, to the "up to" quoted frequency for 1T. However, what I'm curious to see is how high will the CCD be with the cache boost while gaming, compared to a non-X3D CPU of the same. Will boost clocks while gaming, be different between a 7950X vs 7950X3D, THAT is what I want to test and get to the bottom of.

BTW, PC World also has a video explaining this CCD cache difference, and why only one CCD has it, and not the other with an interview with Scott Stankard

 
In addition, I have a feeling the boost frequencies will be different based on which CCD is being used, the one without cache is the one that will boost higher, to the "up to" quoted frequency for 1T. However, what I'm curious to see is how high will the CCD be with the cache boost while gaming, compared to a non-X3D CPU of the same. Will boost clocks while gaming, be different between a 7950X vs 7950X3D, THAT is what I want to test and get to the bottom of.

I met with Scott as well. The boost frequencies will be different on each CCD, with the 3D CCD being slower.

The way he explained it to me is that for light threaded workloads the non-3D side of the 7950X3D should perform the same as the 7950X. Of course, this assumes the threads get to the right cores. Under a full cinebench load, performance will be lower than the 7950X, however, the margin of difference will not be as large as the 7800X3D vs 7800X.

He also mentioned some games would still be better off getting assigned to the non3D side based upon performance characteristics.
 
Last edited:
Become a Patron!
Back
Top