Only "review" that matters, to be honest. Everything else from mainstream tech media is noise. Pretty impressed with the 9700X, can't wait for the 9950X.
The only review on youtube thats worth something. Thanks. When a new cpu is kind of crap just dont say its crap, overclock the crap out of it... like you did.
Those who understand the changes don't think this was a bad cpu. Major architecture changes leads to better memory handling, efficiency, and room to grow. Lowering power draw at base clocks leaves TONS of room for overclocking. SkatterBencher showing this case in point! I can't wait for the 9800x3d or 9950x3d with this kind of potential. Here's hoping the rumors that PBO and curve shaper all work on x3d are true!
Fantastic work! Only deBauer shown this aspect in his review about 9700x and I got a little disappointed by the other media outlets I was following. I'll be back here for the next ones.
These chips weren't pushed close to the red like previous ones. So ocing nets way more performance at the cost of power and thermals. Kinda makes oc'ing a thing to go for on amd again.
Good video! AMD released Zen 4 with temperature limits. People were confused. AMD released Zen 5 with PPT limits so it's efficiënt. People are confused. They will always be confused.
Are people not aware that the Ryzen 7 7700 NON-X exists? It has the same 65W TDP and 88W PPT as the Ryzen 7 9700x, but it has 98% of the Ryzen 7 7700X performance, while costing just 250$.
@@rattlehead999 98% ? 🤔My 7700X runs at 85W PPT 5490 Mhz all core. It used to boost to 6200 MHz on two of the cores with early Agesa (ASRock Live Mixer). Now they hard capped it at 5500 MHz. I speculate the 7700 non X is a lesser bin, so they can sell those parts. It costs less so is not a fraud. A genius move.
This is the best reviewer by far and also the best review of the 9700x on the internet. This is what bigger channels and reviewers have forgot, the tweak, learning curve, applying what you have learnt about the platform. Just running test with everything by default and talking about the results can even be done my IA. This analysis instead, can be only performed by people with the knowledge of the how to. Thanks for the review, detailed and with value.
Great video overall, but you should note in the beginning that 6400MT/s EXPO is putting you at 1xUCLK : 2xMEMCLK so you would have probably gotten way better performance at 6200MT/s or 6000 depending on what that memory controller can do :D Love your detailed long format Videos!
Yet again, the standard tech press fails on their reviews because they dont understand new things unless they're externally explained to them. Still AMD's fault, but they should've done their own homework regardless. Thanks for confirming my initial gut instinct that they were poorly tuned, because the numbers didnt make sense.
We don't need every CPU to default to a 95°C thermal limit control. Personally I'd like most processors to default to a conservative thermal requirement (like the 9700X) and have simple options ( inverse eco mode) to enable beast mode that require upgraded thermal solutions instead of that being the default.
@Garrus-w2h Yeah they did, they drove the insane power draws of modern CPUs by insisting OOTB 125W rated CPU's drawing 253W continuous is in spec and valid. AMD looked as if they were about to follow the same path with 7000 series setting that trend, thank your deity this generation has come back to sanity.
@Garrus-w2h 95°C is Tjmax, the hard thermal stop. AMD CPUs are designed to not go beyond that temp. TDPs are the min cooler rated spec for normal operation @21°C ambient. They've always been that, modern boost technologies have shifted CPUs from being a 1:1 power draw to TDP ratio in the ever increasing amounts. AMD have charted the power draw of various TDPs, it's not that hard to find. Reviewers tend to test with high end 360 AIO coolers so we don't often get to see many of the CPU's OOTB thermals when paired with a rated cooler. In AMD's own Wraith range, the stealth is rated 65W, Spire 90W and prism is 140W. The 7600 (65W) sold with wraith stealth and 7700 the wraith prism, both are 65W processors. Under a cinebench 1hour all core workload the 7600 + stealth hit 95.8°C and average power draw of 81.3W, while the 7700+ prism hit 78.3 and average power of 86.8W. That example shows that in 65W (88W peak) configuration a standard 4 pipe cooler (typically 140-160W rated) will keep modern 65W rated CPUs running smoothly with overhead for overclocking. It also bodes well for SFF builds where tower and water cooling aren't an option. Just because a CPU can run happily @ 95°C doesn't mean that it's ideal. Thermals spread not just from the heatsink, but through the socket heating other nearby components. Capacitor lifetime is directly affected by Temperature. You might have noticed that TDPs are described as default, it used to be static, now its infinitely configurable, default TDPs are a drop in convenience, as is eco modes. No one complains that memory defaults to the JDEC rating and you have to enable EXPO/XMP to get the rated spec. Even reviewers who claim to be reviewing OOTB experience, configure the systems memory away from defaults for performance reasons, but not CPUs. Intel based motherboards automatically switched PL2 to unlimited time just by enabling XMP. Effectively overclocking the CPU using Tjmax as the safety from time limited boost to average TDP, to always on boost essentially doubling TDP. Blame them and reviewers if you don't like the fact that TDP no longer is directly associated with power draw.
@Garrus-w2h There's no real reason to list max power draw as typically the average power draw will be significantly less. Currently my 65W CPU is pulling a little over 50W with a couple of background tasks and watching a youtube video. It's quite common to find power draw below the rated TDP in typical use cases, gaming rarely taxes the full potential of the CPU. TDP as I already stated is the min spec recommendation for cooling solutions. With AMD the sustained max power draw typically seen with all core workloads is known for TDP defaults. 65=88, 90=120, 105=140, 170=230. Feel free to set your own PPT, PBO can also underclock if you want it to, setting PPT to 65W instead of the default 88W is allowed, just as setting PPT to 140W will turn a 65W chip into a 105W chip effectively.
@Garrus-w2h Multi Core Enhancement was a motherboard enabling of time extended PL2 eventually becoming infinite. The trick of it was the disclaimer that PL2 boosts did not break TDP averages as PL2 was designed in Intel spec to be short lived, typically less than 8 seconds and so was in spec even if pulling more power for a short time. Initially MCE would simply apply max turbo across all cores for an extended period of time, basically long enough to gain a benefit in benchmarks. This motherboard initiated soft overclock could be denied by Intel as their doing, nudge nudge wink wink. MCE was often auto enabled by initiating XMP on memory. Various motherboards and M/B manufacturers had their own settings. AMD introduced XFR and precision boost (not overdrive) on Ryzen 2000, it was part of the CPU spec just as PBO and curve optimiser is today. All a soft overclock whilst maintaining the CPU's factory max settings triangle. (PPT, EDC, TDC). Effectively MCE auto enabled would be like testing with PBO turned on by default. The days of the MCE controversy are over, though we do still see intel processors able to pull 300W+ well over their rated max turbo power of 253W in benchmarks showing that there is still some additional MCE trickery in play. Intel no longer uses TDP since 12th gen they use "processor base power" and "maximum Turbo Power" in their specs. Sustained power draw on unlocked processors is user controllable, I agree that max turbo power draw (effectively sustainable with correct choice of cooler) should be part of the CPU spec sheet just for easy reference. Retail Intel processors behave exactly the same as review samples in retail motherboards. OEM prebuilts however can choose to stick within "official" spec (no mce and PL2 time limited to average to the base processor power ) and those may behave quite differently in sustained loads. Once you learn to see modern desktop processors as infinitely configurable, you can get past the chicanery in play and see them for what they are. I treat the various SKUs as a proxy for binning these days. Though in many instances you can get great results from any unlocked SKU, you're still playing the odds for getting exceptional silicon.
You, sir, are a bloody genius. This is about the 5th video on your channel that I've watched and I've learned heaps. Thank you so much. The theory behind all this is actually pretty simple and I'm beginning to get a grip on it. You should get a job with AMD and tell them how to fix their stuff. This is amazing.
I have read most comments about other YT tech channels and I do agree with the commenter's points. Although it is important to understand where the other YT channels are coming from, as it were. Most channels are (are not "complete" channels): 1. Gaming focused and maybe, just might, consider reviewing with things XMP/PBO, etc., only, if they even go that far (most don't) 2. Considering the average user, the non-overclocker/non-tuner/non-tweaker -Therefore- 3. Are not interested in doing things they way they are done in this video. They don't tweak or tune since the vast majority of people also will not and have no desire to do so. This leaves us to take what they offer with a grain salt. A very tiny, kind of incomplete grain of salt. Thank goodness for videos like this.
Agreed there are way too many clickbait clown TechTubers that just don't provide any useful information. JayZ2Braincells spends 20 minutes yapping about something that can be explained in 30 seconds. So glad we have no nonsense channels like SkatterBencher to watch instead of clickbait clowns
First time watcher, loved the methodology and explanations. It seems that reviewers who hated this processor, only consider the merits OOTB or with one click overclocking if they bother at all. Personally I like processors with conservative default OOTB settings with plenty of headroom if you choose to chase higher performance. Three one click TDP settings in the BIOS defaulting to the middle, 9700X should have 45W (eco) , 65W (default) and 105W (performance), this would be no more difficult than enabling EXPO/XMP for memory. The differentiation between SKUs would be one of binning (and possibly higher levels of TDP settings). It's not like we can't already dial in a preferred performance/power draw for the use case. It's not just everything set to over 9000 as the only PBO option.
Yep, you can tune this one they way you want. SFF build and only a small cooler? No worries. Big radiator and want that little bit extra, you got it. I love the granularity of this gen.
Interesting looking at this and seeing that a lot of other tech youtubers just dont really understand how or what they're doing with AMD CPU's, quite eye opening really
I love how you show what exactly you adjust in the BIOS every time, such a well done review ! Would have liked to see maybe a game or two like Cyberpunk 2077 to see how these changes would affect that.
This gives impressive results in terms of frequency. And it's a fantastic quality of demonstration, really great job. I suppose it's normal for this not to affect gaming performance much. It did boost some benchmarks quite a bit, bit most were closer to 10% or less, and immediately at double the power. Again, normal for overclocking. Thanks for that demonstration, really well done and quite cool how this technology works to deploy/configure the OC. But at the same time, I would not be willing to go so high in power for such little performance difference, me personally. I would maybe boost it a little and try to stay closer to +20W-50W over stock, hard upper limit, or else just give up and use stock if it's not possible to see perf gains in my personal real-world workloads of interest.
Only Der8uer scraped this, you went far deeper, well done.. it seems that effectively these CPUs were like purposely held back by AMD. I however do wonder long term reliability. I remember ryzen 3000s didn't fare well with 10x scalar back in the day.
@@SkatterBencher i try to go co -35 the same as you did in the video,but once i run r23, the computer freeze everytime. is that PSU or it is the cpu problem?
I've used curve optimiser on my 5600X and has been happy with the lower temps i'm getting with it (-30 on all cores, no errors on OCCT extreme AVX2, and no crash/BSOD for a month yet) (using the Curve Optimiser on Ryzen master as i don't have the options on my asrock MB and bios) But one thing i understood with the graph at 13:25 taht i didn't know at the time was that i can go higher on the frequency like that. Definitely useful to know and will explore that to see if i can boost higher !
You're balking at 120-180W when intel was pushing over 300W-400W to get their performance. OCing is never about efficiency but pushing performance and clocks. Edit: You can also use curve optimiser and still get an OC if you want lower voltage.
Nice video. As someone still on 3800X this was very helpful to see all the new OC functionality explained. Looks like a worthy upgrade for tweakers despite the somewhat negative reviews because of lower stock TDP. My only concern is 95C sounds toasty, can they take that kind of heat now? Anything over 80C gets me nervous.
95C is the designed max sustained temperature and is considered safe. Obviously, trying to keep it lower is better as long as that does not interfere with your performance goals. That said, my personal max sustained temp limit is 80C as well.
Can we get a manual OC result? The Prime95 SmallFFT AVX2 results with undervolting looks so good. And why don't AMD just give us a V-F table and a per-core boosting settings let us change that also they can implement something similar to Intel XTU which lets you decide the frequency on a core number basis for example 1/2/3 cores - 5.8G, 4 cores - 5.4G, 8 cores - 5.3G etc then I can use Prime95 Small FFT AVX2 frequency for 8 cores and SSE frequency for 1/2 cores.
could you also run R23? it's hard to find comparisons for other cpus with r24 but it is very easy to find r23 scores for practically every cpu made in the last 5 years. ty for conisderation
Great video...for context is that running hot even at 170w at 95c? On a 14900k at 170w would be more like 60-70c even 253w can stay around 90 with just a 360 AIO ? Not really comparing like that just seems super hot?
I haven't seen a single OC benchmark yet where 9700X would've gone above 89-90° at full load operating on around 5,4MHz, so I'm assuming yes, all of the AIOs should be dealing with it perfectly 👌🏻
@@s3rm0n56 the CPU operates on higher clocks that I've mentioned yet it's still firmly stable at +5° temps. Lower wattage at higher temps generates less heat than high wattage on mid-to-high temps, so I wouldn't be worried at all.
Cool, but how does this translate into gaming and app performance? Der8aur also did a rough, basic overclock his 9700X, removing all power limits, but the difference in gaming performance was negligible. If all this work doesn't translate into tangible performance improvements where it matters, then it doesn't save the CPU. Can you make some gaming benchmarks comparing your 9700X OC to its stock performance?
you don't buy a 9700x for gaming unless you can't afford the 9800x3d. in that case buy a cheaper 7800xd. 9700x is the all around bargain one, good enough for everything.
Most people spend way too much time worrying about gaming performance with CPUs when in reality it's generally not that important since most people are GPU limited when playing at ultra settings. It only matters for those who play competitive online and chase max frames.
@@SkatterBencher sorry, i cannot really explain it, it is beyond my knowledge. But it seems to me that 2:3 is still a somewhat 'rounded' ratio (= in 6 cycles one runs 2 times, the other 3 times, without fractions). Buildzoid told in his AM5 videos that according to his experience if the IF clock moves only 1 or 2 steps away/up from the 2:3 ratio then the performance is generally slightly weaker. (Do not forget that the 'recommended' 6000 Mt/s at IF 2000 MGHz also gives 2:3 ratio.) I was just curious if you have tried it and able to share your experience.
Do you think that pbo curve is going to be the new stability optimization this gen over individually tuning core curves? Looks really good that you got -35 on all
Curve Optimizer is still the fastest route to more performance, but I find Curve Shaper to also be a very powerful tool. This chip had a lot of undervolt margin, so CO was easy. Maybe with other chips that have less margin CS will be more impactful. For any OC Strategy with async eCLK, Curve Shaper is a paradigm shift. Total lifesaver.
@@SkatterBencher interesting I never got around to trying to do eclk with my 7950x3d I was interested since I delid it but started to doubt it would be worth the trouble
Shocking, the single-CCD parts are thermal limited. Just like they've been since Zen 2. Potentially encouraging for 99x0X parts; they'll likely be capable of swallowing Intelesque power but with relatively similar scaling to that seen here. Not that over double power consumption for 10% is "good" scaling, but hey!
super disappointed by the memory controller... they didn't do anything with it. and that was zen 4's only issue, memory complete instability issues. I CAN'T BELIEVE THEY DID NOTHING HERE!
I can not wait until AM6, it is so obvious how much even these 8 core CPU's are hold back by way to little memory bandwidth. We need DDR6 as soon as possible.
That's why AMD need X3D cache. If they can improve the I/O die, the processor will good at everything. The core already better than Intel, but not the I/O die.
There are rumours that zen 6 will have much lower i/o to ccd latency Of it's true it will dramatically improve gaming performance P.s. i thought they will improve infinity fabric in zen 5 :(
Do you have any idea how cold Ln2 is? You can get cpus to boost over 2ghz with it. If they can’t even hit 7 ghz with Ln2 I doubt you’ll be able to hit 6 ghz without removing the ihs and doing a custom loop. I’m getting an arctic 3 420aio for the 9950x and even if I used a push/pull config with 6 fans I doubt I’ll be able to hit 6 ghz.
@@SkatterBencher Nothing's stopping you from taking the undervolt values it gives you (per core btw) and applying them in the bios. That's what I did and it let me still do a hefty undervolt on my 5800x despite having 1 really bad core that can only do -12 while the rest can do -20 or better.
They should've given the 9700x 105w PPT instead of giving it 88w like the 9600x, seems to be not quite enough power per core vs the 170w and 200w for the 9900x and 9950x
I agree since it's leading to some reviews disappointed because it doesn't always beat the 7700X which makes it look bad, despite being a more sophisticated chip.
These CPUs are pretty much squeezed out at stock. The era of 30-40% extra juice is long gone. I've spent 3 weeks messing with 7800x3d just to get 7% out of memory and CO per core offset. That is on top of just xmp/PBO. The funny part was that clearing the crap out of 22H2 win 10 custom already cut down iso gave me another 10% in Night raid and Cyberpunk lowest settings QHD. That took me like 20 minutes with Shutup O&O. For older games LTSC 1809 may be a bit faster, atleast it was side by side in entry presets 3dmark11.
Weird of AMD to restrict an 8-core "X" part to 65W when it can achieve a lot more when it's more unconstrained. At 65W it barely beats the 2 years old 7700x which is a bit embarrassing. Good to know that at least it has a good OC potential to get back the performance.
Finally! A review worth watching
As usual, SkatterBencher writes the BIBLE OF OVERCLOCKING AMD!
Amazing job!
Only "review" that matters, to be honest. Everything else from mainstream tech media is noise. Pretty impressed with the 9700X, can't wait for the 9950X.
finally 100% review of cpu not 1% like almost all others on yt. Hats off
The only review on youtube thats worth something. Thanks. When a new cpu is kind of crap just dont say its crap, overclock the crap out of it... like you did.
Those who understand the changes don't think this was a bad cpu. Major architecture changes leads to better memory handling, efficiency, and room to grow. Lowering power draw at base clocks leaves TONS of room for overclocking. SkatterBencher showing this case in point!
I can't wait for the 9800x3d or 9950x3d with this kind of potential. Here's hoping the rumors that PBO and curve shaper all work on x3d are true!
and you have no idea what you are talking about. I feel bad for this creator making such a great video and thats what you got out of it.
An OC is an OC you're running it out of spec
If something happens to that chip they can deny you warranty..
the silicon lottery is also a thing
@@jittertn they cannot prove that you overclocked it
@@Celatra Except they can. There are flags (tiny fuses) within CPU's that trip under certain OC conditions.
Ah, the BEST overclocking guide in all of TH-cam as always - keep pushing it !
Massive improvements on this cpu while not using extreme cooling. Impressive!
Fantastic work! Only deBauer shown this aspect in his review about 9700x and I got a little disappointed by the other media outlets I was following. I'll be back here for the next ones.
These chips weren't pushed close to the red like previous ones. So ocing nets way more performance at the cost of power and thermals. Kinda makes oc'ing a thing to go for on amd again.
It's opposite to Intel which drives their processor balls to the walls, borderline crazy and degrading itself.
@@KarrasBastomi Intel specs don't degrade CPU, motherboards applying crazy voltage do. It would happen the same to AMD CPUs aswell
@@Mostwest lol Intel admitted that their microcode was broken and it is cooking all the 13-14th gen CPUs do you live under a rock or something
@@Mostwestintel specs are just recommendation
Yeah but Amd does not like users to OC, is there a model that they will support OC on? Fooz could void warranties
Good video!
AMD released Zen 4 with temperature limits. People were confused.
AMD released Zen 5 with PPT limits so it's efficiënt. People are confused.
They will always be confused.
Glad there are people who can describe the failure of amd in such a funny way, IMAO.
never give up, even when you're being fed bullshit
AMD is confused since they didn't give it 105w at default and kneecapped it with 88w PPT vs the 9900x 170w and 9950x 200w
Are people not aware that the Ryzen 7 7700 NON-X exists? It has the same 65W TDP and 88W PPT as the Ryzen 7 9700x, but it has 98% of the Ryzen 7 7700X performance, while costing just 250$.
@@rattlehead999 98% ? 🤔My 7700X runs at 85W PPT 5490 Mhz all core. It used to boost to 6200 MHz on two of the cores with early Agesa (ASRock Live Mixer). Now they hard capped it at 5500 MHz. I speculate the 7700 non X is a lesser bin, so they can sell those parts. It costs less so is not a fraud. A genius move.
This is the new Sandy-Bridge, It's a Overclocking beast. Great OC Review-Guide.
Why I discover this channel only now?
Great content, never stop what you are doing.
This is the best reviewer by far and also the best review of the 9700x on the internet.
This is what bigger channels and reviewers have forgot, the tweak, learning curve, applying what you have learnt about the platform.
Just running test with everything by default and talking about the results can even be done my IA. This analysis instead, can be only performed by people with the knowledge of the how to.
Thanks for the review, detailed and with value.
This is the best overclocking guide I have seen in my entire life!
Great video overall, but you should note in the beginning that 6400MT/s EXPO is putting you at 1xUCLK : 2xMEMCLK so you would have probably gotten way better performance at 6200MT/s or 6000 depending on what that memory controller can do :D Love your detailed long format Videos!
Yet again, the standard tech press fails on their reviews because they dont understand new things unless they're externally explained to them.
Still AMD's fault, but they should've done their own homework regardless.
Thanks for confirming my initial gut instinct that they were poorly tuned, because the numbers didnt make sense.
We don't need every CPU to default to a 95°C thermal limit control. Personally I'd like most processors to default to a conservative thermal requirement (like the 9700X) and have simple options ( inverse eco mode) to enable beast mode that require upgraded thermal solutions instead of that being the default.
@Garrus-w2h Yeah they did, they drove the insane power draws of modern CPUs by insisting OOTB 125W rated CPU's drawing 253W continuous is in spec and valid.
AMD looked as if they were about to follow the same path with 7000 series setting that trend, thank your deity this generation has come back to sanity.
@Garrus-w2h
95°C is Tjmax, the hard thermal stop. AMD CPUs are designed to not go beyond that temp.
TDPs are the min cooler rated spec for normal operation @21°C ambient. They've always been that, modern boost technologies have shifted CPUs from being a 1:1 power draw to TDP ratio in the ever increasing amounts. AMD have charted the power draw of various TDPs, it's not that hard to find.
Reviewers tend to test with high end 360 AIO coolers so we don't often get to see many of the CPU's OOTB thermals when paired with a rated cooler. In AMD's own Wraith range, the stealth is rated 65W, Spire 90W and prism is 140W.
The 7600 (65W) sold with wraith stealth and 7700 the wraith prism, both are 65W processors. Under a cinebench 1hour all core workload the 7600 + stealth hit 95.8°C and average power draw of 81.3W, while the 7700+ prism hit 78.3 and average power of 86.8W.
That example shows that in 65W (88W peak) configuration a standard 4 pipe cooler (typically 140-160W rated) will keep modern 65W rated CPUs running smoothly with overhead for overclocking. It also bodes well for SFF builds where tower and water cooling aren't an option.
Just because a CPU can run happily @ 95°C doesn't mean that it's ideal. Thermals spread not just from the heatsink, but through the socket heating other nearby components. Capacitor lifetime is directly affected by Temperature. You might have noticed that TDPs are described as default, it used to be static, now its infinitely configurable, default TDPs are a drop in convenience, as is eco modes. No one complains that memory defaults to the JDEC rating and you have to enable EXPO/XMP to get the rated spec. Even reviewers who claim to be reviewing OOTB experience, configure the systems memory away from defaults for performance reasons, but not CPUs.
Intel based motherboards automatically switched PL2 to unlimited time just by enabling XMP. Effectively overclocking the CPU using Tjmax as the safety from time limited boost to average TDP, to always on boost essentially doubling TDP. Blame them and reviewers if you don't like the fact that TDP no longer is directly associated with power draw.
@Garrus-w2h
There's no real reason to list max power draw as typically the average power draw will be significantly less. Currently my 65W CPU is pulling a little over 50W with a couple of background tasks and watching a youtube video. It's quite common to find power draw below the rated TDP in typical use cases, gaming rarely taxes the full potential of the CPU. TDP as I already stated is the min spec recommendation for cooling solutions. With AMD the sustained max power draw typically seen with all core workloads is known for TDP defaults. 65=88, 90=120, 105=140, 170=230. Feel free to set your own PPT, PBO can also underclock if you want it to, setting PPT to 65W instead of the default 88W is allowed, just as setting PPT to 140W will turn a 65W chip into a 105W chip effectively.
@Garrus-w2h
Multi Core Enhancement was a motherboard enabling of time extended PL2 eventually becoming infinite. The trick of it was the disclaimer that PL2 boosts did not break TDP averages as PL2 was designed in Intel spec to be short lived, typically less than 8 seconds and so was in spec even if pulling more power for a short time.
Initially MCE would simply apply max turbo across all cores for an extended period of time, basically long enough to gain a benefit in benchmarks. This motherboard initiated soft overclock could be denied by Intel as their doing, nudge nudge wink wink.
MCE was often auto enabled by initiating XMP on memory. Various motherboards and M/B manufacturers had their own settings. AMD introduced XFR and precision boost (not overdrive) on Ryzen 2000, it was part of the CPU spec just as PBO and curve optimiser is today. All a soft overclock whilst maintaining the CPU's factory max settings triangle. (PPT, EDC, TDC).
Effectively MCE auto enabled would be like testing with PBO turned on by default. The days of the MCE controversy are over, though we do still see intel processors able to pull 300W+ well over their rated max turbo power of 253W in benchmarks showing that there is still some additional MCE trickery in play. Intel no longer uses TDP since 12th gen they use "processor base power" and "maximum Turbo Power" in their specs.
Sustained power draw on unlocked processors is user controllable, I agree that max turbo power draw (effectively sustainable with correct choice of cooler) should be part of the CPU spec sheet just for easy reference.
Retail Intel processors behave exactly the same as review samples in retail motherboards. OEM prebuilts however can choose to stick within "official" spec (no mce and PL2 time limited to average to the base processor power ) and those may behave quite differently in sustained loads.
Once you learn to see modern desktop processors as infinitely configurable, you can get past the chicanery in play and see them for what they are. I treat the various SKUs as a proxy for binning these days. Though in many instances you can get great results from any unlocked SKU, you're still playing the odds for getting exceptional silicon.
You, sir, are a bloody genius. This is about the 5th video on your channel that I've watched and I've learned heaps. Thank you so much. The theory behind all this is actually pretty simple and I'm beginning to get a grip on it.
You should get a job with AMD and tell them how to fix their stuff. This is amazing.
I have read most comments about other YT tech channels and I do agree with the commenter's points. Although it is important to understand where the other YT channels are coming from, as it were. Most channels are (are not "complete" channels):
1. Gaming focused and maybe, just might, consider reviewing with things XMP/PBO, etc., only, if they even go that far (most don't)
2. Considering the average user, the non-overclocker/non-tuner/non-tweaker
-Therefore-
3. Are not interested in doing things they way they are done in this video. They don't tweak or tune since the vast majority of people also will not and have no desire to do so.
This leaves us to take what they offer with a grain salt. A very tiny, kind of incomplete grain of salt.
Thank goodness for videos like this.
Agreed there are way too many clickbait clown TechTubers that just don't provide any useful information. JayZ2Braincells spends 20 minutes yapping about something that can be explained in 30 seconds.
So glad we have no nonsense channels like SkatterBencher to watch instead of clickbait clowns
First video of you I have seen.
Dude! Epic stuff here I am very impressed. Instant sub.
You make all other reviewers look lazy.
Wow.. you just give me clear explanation about curve shaper.. awesome feature.. many Thanks..
Compare tuned/untuned 7700x vs 9700x perf / power scaling curves. To see if the 50% efficiency claims from some outlets are accurate.
Extremely interesting and superb granularity. Love your reviews. Even a cheaper board w/o ext clk generator use would fare nicely. Thks again.
First time watcher, loved the methodology and explanations.
It seems that reviewers who hated this processor, only consider the merits OOTB or with one click overclocking if they bother at all.
Personally I like processors with conservative default OOTB settings with plenty of headroom if you choose to chase higher performance.
Three one click TDP settings in the BIOS defaulting to the middle, 9700X should have 45W (eco) , 65W (default) and 105W (performance), this would be no more difficult than enabling EXPO/XMP for memory.
The differentiation between SKUs would be one of binning (and possibly higher levels of TDP settings). It's not like we can't already dial in a preferred performance/power draw for the use case. It's not just everything set to over 9000 as the only PBO option.
Yep, you can tune this one they way you want. SFF build and only a small cooler? No worries. Big radiator and want that little bit extra, you got it. I love the granularity of this gen.
Zen 5 honestly looks like an absolute banger! More efficient than Zen 4, massive OC headroom! I'm waiting for the 9950X, the 9950X3D or the 9800X3D.
Interesting looking at this and seeing that a lot of other tech youtubers just dont really understand how or what they're doing with AMD CPU's, quite eye opening really
Try pointing that out to the Steve stans, brainwashed puppets are fanatical 😅😂
I love how you show what exactly you adjust in the BIOS every time, such a well done review ! Would have liked to see maybe a game or two like Cyberpunk 2077 to see how these changes would affect that.
Epic video bro, one of the best I've seen in terms of OC and general tech
Thanks for elaborate job!
Christmas came early! Looks like delidding will be popular on the 9000 series.
Glad to see you got this up right away good job!
Looks good that simple all core co and pbo is enough for this
I cannot wait for the 9950X video!
This gives impressive results in terms of frequency. And it's a fantastic quality of demonstration, really great job.
I suppose it's normal for this not to affect gaming performance much. It did boost some benchmarks quite a bit, bit most were closer to 10% or less, and immediately at double the power. Again, normal for overclocking. Thanks for that demonstration, really well done and quite cool how this technology works to deploy/configure the OC. But at the same time, I would not be willing to go so high in power for such little performance difference, me personally. I would maybe boost it a little and try to stay closer to +20W-50W over stock, hard upper limit, or else just give up and use stock if it's not possible to see perf gains in my personal real-world workloads of interest.
Wow this thing is a little beast when undervolted :D
Just wow! Great video!!
Only Der8uer scraped this, you went far deeper, well done.. it seems that effectively these CPUs were like purposely held back by AMD.
I however do wonder long term reliability. I remember ryzen 3000s didn't fare well with 10x scalar back in the day.
10X scaler doesn't seem to add more than 25mV headroom in with any kind of load, so I'm not too worried about the long-term effects
@@SkatterBencher I see, so it's pretty much contained
I love that AMD allowed more customisation and made overclocking/undervolting more interesting.
I can't wait to buy 9950x and delid!
You sicko 😂
So cool, thank you
Que buen trabajo realizas, a la espera de tus resultados para el 9950x la siguiente semana, de mientras te ganaste un fan!
these cpus are really good. the price is high right now at launch but they are pretty amazing nonetheless.
under appreciated review. Is that a 24/7 stable result?
Yes, all my OC Strategies are intended to be stable configurations. That's why I always include AVX and SSE stability tests as well
@@SkatterBencher i try to go co -35 the same as you did in the video,but once i run r23, the computer freeze everytime. is that PSU or it is the cpu problem?
that was a really interesting video, thank you.
I've used curve optimiser on my 5600X and has been happy with the lower temps i'm getting with it (-30 on all cores, no errors on OCCT extreme AVX2, and no crash/BSOD for a month yet) (using the Curve Optimiser on Ryzen master as i don't have the options on my asrock MB and bios)
But one thing i understood with the graph at 13:25 taht i didn't know at the time was that i can go higher on the frequency like that. Definitely useful to know and will explore that to see if i can boost higher !
U got my sub btw
Thanks, this is what I've been wait for
Great video! Very well explained and good to follow
However, double or even triple the power draw is a hard pill to swallow
You're balking at 120-180W when intel was pushing over 300W-400W to get their performance. OCing is never about efficiency but pushing performance and clocks.
Edit: You can also use curve optimiser and still get an OC if you want lower voltage.
This is very good. I am still deciding if I should get the Ryzen 7 9700X or the ryzen 9 7900 non-X.
great post, i have a question, how much VRAM is there on your 2080Ti
TWEAKERS DREAM CPU!
How will people prepare for Hurricane Debby when AMD used up the sandbags for the 9700X?
Really helpful. Thanks!
Amazing work. I hope you will try 8000 mhz ram speed overclock.
It's on my to-do list ... just need to find a kit that can run 8000 :)
Nice video. As someone still on 3800X this was very helpful to see all the new OC functionality explained. Looks like a worthy upgrade for tweakers despite the somewhat negative reviews because of lower stock TDP.
My only concern is 95C sounds toasty, can they take that kind of heat now? Anything over 80C gets me nervous.
95C is the designed max sustained temperature and is considered safe. Obviously, trying to keep it lower is better as long as that does not interfere with your performance goals.
That said, my personal max sustained temp limit is 80C as well.
belo trabalho man, acompanhando
AMAZING!!!!
Damn this is great. will they do this for 9950x?
Crazy overclocking potential! How much can the peak 1-4 core clocks be pushed with better cooling?
Outstanding!
Spectacular!
The 9950x is 170w and so is the 7950x. Is it safe to assume stock performance of the 9950x won't have the same issues as stock 9700x?
I would be surprised if they did
@Jarrod When can we expect your review of amd strix point?
Absolute Cinema
Well done! as a gamer with 99% of my PC time it seems the only things I would want to do is undervolt and use EXPO and tune my ram. wonderful video!
You're the best!!!!!!
Can we get a manual OC result? The Prime95 SmallFFT AVX2 results with undervolting looks so good.
And why don't AMD just give us a V-F table and a per-core boosting settings let us change that
also they can implement something similar to Intel XTU which lets you decide the frequency on a core number basis
for example 1/2/3 cores - 5.8G, 4 cores - 5.4G, 8 cores - 5.3G etc
then I can use Prime95 Small FFT AVX2 frequency for 8 cores and SSE frequency for 1/2 cores.
could you also run R23? it's hard to find comparisons for other cpus with r24 but it is very easy to find r23 scores for practically every cpu made in the last 5 years. ty for conisderation
So is it better to use option 1 or 2 and why?
thank you
does the new bios with 105w tdp help any for better OC
This is really useful
I am having high idle temps around 50C, from what i have read, there is more ppl with this issue, maybe its even normal for this chip?
X870e taichi, 9800x3d, 6400cl32 ram.
Which strategy should I follow? Not too much power consumption, but decent gainz
Strategy #3 should be fine. Curve Optimizer will boost the frequency and reduce the power.
@SkatterBencher thanks a lot
Great video...for context is that running hot even at 170w at 95c? On a 14900k at 170w would be more like 60-70c even 253w can stay around 90 with just a 360 AIO ? Not really comparing like that just seems super hot?
I haven't seen a single OC benchmark yet where 9700X would've gone above 89-90° at full load operating on around 5,4MHz, so I'm assuming yes, all of the AIOs should be dealing with it perfectly 👌🏻
@@F1stzz what about the temps in this video?
@@s3rm0n56 the CPU operates on higher clocks that I've mentioned yet it's still firmly stable at +5° temps. Lower wattage at higher temps generates less heat than high wattage on mid-to-high temps, so I wouldn't be worried at all.
@@F1stzz i dunno man only 5ghz at 1.155v at 168w maxing 95c seems pretty toasty
A chiller could have better results on temperature, what is your point?
Damn i learn so much that i just pushed my 5600x to 4.85ghz single core and 4.6mhz occt , pbo max ,+6 co , 200mhz offset 😊 not stable with -ve CO ,
How is it for gaming? Dont feel like dealing with RMA from intel.
Cool, but how does this translate into gaming and app performance? Der8aur also did a rough, basic overclock his 9700X, removing all power limits, but the difference in gaming performance was negligible. If all this work doesn't translate into tangible performance improvements where it matters, then it doesn't save the CPU. Can you make some gaming benchmarks comparing your 9700X OC to its stock performance?
you don't buy a 9700x for gaming unless you can't afford the 9800x3d. in that case buy a cheaper 7800xd. 9700x is the all around bargain one, good enough for everything.
@@JohnSmith-ro8hkhe wants to know how 9700x performs all around
Most people spend way too much time worrying about gaming performance with CPUs when in reality it's generally not that important since most people are GPU limited when playing at ultra settings. It only matters for those who play competitive online and chase max frames.
Any gaming benchmarks
Would not be better to set the infinity clock to 2133 MHz when the RAM runs at 6400 Mt/s? That would give 2:3 ratio between the IF and memory clock.
I've seen people suggestion 2:3 ratio. Is there any reference I can look at that explains why it's better?
@@SkatterBencher
sorry, i cannot really explain it, it is beyond my knowledge.
But it seems to me that 2:3 is still a somewhat 'rounded' ratio (= in 6 cycles one runs 2 times, the other 3 times, without fractions).
Buildzoid told in his AM5 videos that according to his experience if the IF clock moves only 1 or 2 steps away/up from the 2:3 ratio then the performance is generally slightly weaker. (Do not forget that the 'recommended' 6000 Mt/s at IF 2000 MGHz also gives 2:3 ratio.)
I was just curious if you have tried it and able to share your experience.
Hi mate! I wish you attached some gaming benchmarks as well
I did include a couple of 3DMarks, Final Fantasy, Returnal, and Tomb Raider as game or game equivalent benchmarks
Do you think that pbo curve is going to be the new stability optimization this gen over individually tuning core curves?
Looks really good that you got -35 on all
Curve Optimizer is still the fastest route to more performance, but I find Curve Shaper to also be a very powerful tool. This chip had a lot of undervolt margin, so CO was easy. Maybe with other chips that have less margin CS will be more impactful.
For any OC Strategy with async eCLK, Curve Shaper is a paradigm shift. Total lifesaver.
@@SkatterBencher interesting I never got around to trying to do eclk with my 7950x3d I was interested since I delid it but started to doubt it would be worth the trouble
Shocking, the single-CCD parts are thermal limited.
Just like they've been since Zen 2. Potentially encouraging for 99x0X parts; they'll likely be capable of swallowing Intelesque power but with relatively similar scaling to that seen here. Not that over double power consumption for 10% is "good" scaling, but hey!
TDP is now also 105w for the 9700x, now what?
I can see the 10000 series will only have some silicon optimization and factory overclocked 9000 series cpu.
Can you do r7 2700?
This eight core 9700X should've been launched with a TDP of at least 80 W (=110 W PPT).
A TDP of 65 W (=88 W PPT) clearly limits it too much.
super disappointed by the memory controller... they didn't do anything with it. and that was zen 4's only issue, memory complete instability issues. I CAN'T BELIEVE THEY DID NOTHING HERE!
I can not wait until AM6, it is so obvious how much even these 8 core CPU's are hold back by way to little memory bandwidth.
We need DDR6 as soon as possible.
That's why AMD need X3D cache. If they can improve the I/O die, the processor will good at everything. The core already better than Intel, but not the I/O die.
There are rumours that zen 6 will have much lower i/o to ccd latency
Of it's true it will dramatically improve gaming performance
P.s. i thought they will improve infinity fabric in zen 5 :(
Apparently newer BIOS revisions actually let these chips do over 8000MT/s :D
Are you showing us a highly binned chip?
Not binned as far as I know. Maybe lucky random sample :)
This for benchmarking but running the cpu on 97 C , 24/7 will degrad the cpu faster than it is speed 😬
Interesting, so most likely 9950x can do around 6 ghz with some tuning.
😂
Actually the 9950X according to Gamers Nexus can clock to 6.50 GHZ with LN2, I suspect an easy 6 GHZ to 6.20 GHZ with standard liquid cooling.
Do you have any idea how cold Ln2 is? You can get cpus to boost over 2ghz with it. If they can’t even hit 7 ghz with Ln2 I doubt you’ll be able to hit 6 ghz without removing the ihs and doing a custom loop. I’m getting an arctic 3 420aio for the 9950x and even if I used a push/pull config with 6 fans I doubt I’ll be able to hit 6 ghz.
@@JacobBrown-pc2bd Lets hope its able to hit it.
Why it not possible at 7600x and 7700x?
7700X guide: th-cam.com/video/0E0Z4ieo7nw/w-d-xo.html
7600X guide: th-cam.com/video/sEWYcpcZ2Zo/w-d-xo.html
Why not just used Hydra?
I prefer overclocking from BIOS since the configuration is set at boot and I can benefit from the motherboard vendor auto-rules
@@SkatterBencher Nothing's stopping you from taking the undervolt values it gives you (per core btw) and applying them in the bios. That's what I did and it let me still do a hefty undervolt on my 5800x despite having 1 really bad core that can only do -12 while the rest can do -20 or better.
king!
Ryzen 7 9800X3D Can have this same stock clock as 7 9700x
The ones who knows, knows...
Will it degrade? this its like pushing it beyond the extreme....
Doesn't even go beyond 1.4V. Intel's degrading chips hit 1.6V.
@@The_Noticer. previous gen chip when it went over 1.35 it blew up the memory controller, and literally the CPU blow-up but AMD and Asus fix it
@@eilegz That was SoC voltage...
They should've given the 9700x 105w PPT instead of giving it 88w like the 9600x, seems to be not quite enough power per core vs the 170w and 200w for the 9900x and 9950x
I agree since it's leading to some reviews disappointed because it doesn't always beat the 7700X which makes it look bad, despite being a more sophisticated chip.
👍
A light in the dark tunnel! This was a disaster launch. But OC saves the day?
Nope, my 13600k oced to 5.8ghz still curbstomps this cpu 😂😂😂. Don’t go near Ryzen 9000.
@@ProVishGaming cool story bro.
@@ProVishGaming your cpu is subject to degradation courtesy of intel though
@@PSXman9 cinebench video on my channel. Facts don't care about your feelings.
@@vespa7961 actually no cause I don't run it 1.5v out of the box like most people do. It runs at a stable 1.28v. Nice try tho
But...is it healthy for the CPU?
As long as voltage stays under 1.4v you are probably safe. Most people wont be able to get over 1.3v due to thermals
6-12 mo. from now we'll know the answer to that question. That's when you buy.
These CPUs are pretty much squeezed out at stock. The era of 30-40% extra juice is long gone. I've spent 3 weeks messing with 7800x3d just to get 7% out of memory and CO per core offset. That is on top of just xmp/PBO. The funny part was that clearing the crap out of 22H2 win 10 custom already cut down iso gave me another 10% in Night raid and Cyberpunk lowest settings QHD. That took me like 20 minutes with Shutup O&O. For older games LTSC 1809 may be a bit faster, atleast it was side by side in entry presets 3dmark11.
This chips is not actually for consumers is more for servers and database applications.
There is like a improvement of 100%
Ryzen cpus are not for servers lol
😘
Weird of AMD to restrict an 8-core "X" part to 65W when it can achieve a lot more when it's more unconstrained. At 65W it barely beats the 2 years old 7700x which is a bit embarrassing. Good to know that at least it has a good OC potential to get back the performance.
How is it embarrassing to achieve the same performance while using 30-40% less power? Thats pretty incredible if you ask me.
@@FlorianskiHUB compares it with a 7700 non-X 65w and the 9700X is faster by only 7%.
Could be a lot better, for sure.
@@nvnh319Steve is a clickbait drama queen and quite ignorant. That is why you are here, to actually learn something.