I'd rather have a stable CPU and a valid warranty than a 5% boost to performance any day. It was a different story back in the day when a Celeron 300 boosted to 450MHz and you got a 50% overclock that matched a more expensive CPU (minux the cache).
over the time, everybody does what you said, i prefer stability, a painless experience, but i used to play with values, push limits, but those times are over, i dont have time and patience for that
The Celeron 300A was actually (a lot of times) faster than the Pentium 2, because, unlike the Pentium 2 (which did have more cache, 512KB, but it was external - separate SRAM ICs outside of the die/core, on the Slot 1 PCB), the Celeron 300A had 128 KB of much faster on-die cache. So, effectively, when overclocked to 450MHz, it was faster than the Pentium 2. Good times, good times...
I have my 9800X3D at -30 all core stable and I'm running with lower temps and power consumption. No need for higher frequencies since I'm already maxing out my GPU.
@RangaAce That's due to your GPU more then likely having bad temps or inadequate power. Check your GPU thermal paste and check your power supply and or connections to the power supply and motherboard Also i hope you didn't pair a 9800x3d with a 3070
@@forog1 Why? I paired a 7800x3D with my old 2080, because I already plan to upgrade to 5080 5070 ti or 8800xt. Or even Battlemage, should it unexpectedly compete and be viable.
You are using the built in benchmark that is why. Others use a repeatable run in an area that is much more taxing and then you see a larger difference. This holds true for a lot of the game. The built in benchmark is not a very good indicator of total performance.
Running around in an area is not reliable anyway, there could be different NPCs, cars, time of day, viewing angle, etc. That's why built in benchmarks exist.
@@Mavis847 In case of Cyberpunk 2077 the whole Phantom Liberty DLC has upped the hardware requirements a lot. In the built-in benchmark I got 40 average FPS on a GTX1060. In Dogtown my FPS would dip into low 20s, and that made me finally upgrade my PC :-)
Hang on a second at 12:13 you have cinebench sccores, how in the world is the 9800XD with X3D Turbo enabled score very slightly below stock or the oc model? Thats not even remotely possible since X3D disables SMT which has a big impact on cinebench scores.. Even my system 9800x3d / 4090 gets close to 24k with +200mhz. Then when i use X3D Boost on motherboard, i lose a decent chunk of points there. (I am on the X670E Gigabyte Masters)
Mine does 5.4Ghz on 1.2V and it ripps. My brother have 5800x3d and that feels 2 gens behind in Pubg atleast. 6400Mhz cl30 you start to get rid of the infinity bottleneck latency wise. I get 370w average fps in Pubg with 200-280 in 1% lows. Ive had 10700k, 12700k, 14900k, 7800x3d, 9800x3d and buildt 2 systems with 5800x3d for others. All systems ended up maxed out overclocked. The 9800x3d overclocked and 140w power budget is the only cpu capabable of running Pubg well without stutters. After 6 years of trying to run that game flawless I'm now happy.
i advised a friend of mine almost 2 years to wait before upgrading his 6700/1060 pc knowing we simrace and he wants triple screens . he decided he didn't want to wait anymore just before the 9800x3d launch . i advised him get a 7800x3d and 4080super . seeing the benchmarks and knowing he games with 4k resolution . i'm glad he decided to not wait longer . knowing the new assetto game is coming in januari i'm still trying to convince him to go ultrawide vs triple screen . better to have 100fps or more in simracing vs having triple screen but only getting 60fps .
I have been waiting for this one, I had a feeling you were putting in allot of time on this CPU. Will be looking forward to the underclocking episode also. You always manage to find the sweet spot. Relax and enjoy what you are doing, it makes a difference to quite a few viewers.
So I compared the 2 in a video in 4k anyways, and my cp2077 settings were crazy, it actually showed the 9800x3d did 1% worse then the 7800x3d. The games where I got a massive (20% ) improvement was helldivers 2 and space marine 2 games that were cpu bound with the 7800x3d
@@forog1 Mine does -25 5.4Ghz on what equals to stock voltage. I think silicon quality varies a lot. I have Asrock Hdv and after tuning the ram it does 8000Mhz on a cheap motherboard. It also does 5.5Ghz but that makes the infinity clocks run outside optimimum speed ranges/dividers unfortunately. If you get a good cpu like mine you can oc a ton without it throttling. And the cpu runs extremely cool, Amd have made a massive effort to get the heat out of the chip this time around.
@@impuls60 Mine also does CO -25 all cores with a +200mhz boost, and 1x scalar. I am getting a score of 23750 in cinebench r23 multi at 85-89 temps. All of this with a cheap Gigabye B650 X AX that I bought for $140 back when I got my 7800x3d. But when I tried 5.45 all-core OC with 1.225v on the vcore, I throttled at 95c in cinebench almost instantly. My 7800x3d never even reached 70c with this same cooler, so my 9800x3d definitely runs A LOT hotter.
@@impuls60 good to know. I'm just waiting on the 9950x3d hopes it will be just as amazing at 9800x3d. Hope to God it would be 29800x3ds's is 1 package lol only can hope
it's always worth tuning to some extent, there is a lot of people saying they would rather have reliability but as soon as you use the pre baked expo or xmp ram settings your over stock any so it's always worth tuning to find lower voltages mainly. look as the soc voltage in the video of 1.25v that's pretty high for zen 5 and probably doesn't need to be above 1.05v for that ram speed.
This is for context not a shameless plug. When I put my 7800X 3D up against my 9800X 3D in Cyberpunk using the in game benchmark. Just like you. 1080p using Low settings with a 7900 XTX Nitro+. 311 for 7800X 3D.....317 for the 9800 X 3D. I posted this around 6 days ago.
So basically a nothing burger, from what I can tell, my 9800x3d is actually pretty poor silicon too, so my 7800x3d could be a better bin, hence the slight change.
Overclocking? Maybe if you got a good cooler. But I prefer undervolting with the curveoptimizer and now also curveshaper. Doing that I reduced the powerusage by 20% (-30W of 145W) under synthetic load while gaining 3% performance with 23465 points in CBR23 (my base was ~22800). The temps also got 10-15°C lower. I was able to use -20 on the optimizer and -10 on the shaper on high and max clocks. Other then that I didn't touch anything.
@@frommatorav1not really, for heavy overclocking. You can make your whole pc ambient temp for example, by using a big external copper radiator. I wouldnt say AIOs are particularly suited for heavy overclocking, custom loops are more for that job
Appreciate you, Bryan for your testing methodology. The silicon lottery could also play a factor. Maybe your 7800x3d is a golden sample, or maybe the 9800x3d is an oc dud. It's still the best 2 CPUs for gaming.
In my own testing 9800x3d is about 15-20% higher 1%lows in Pubg over my meh 7800x3d. -5 on 7800x3d and -25 9800x3d (-25 equals 0.1V equals 200Mhz for free)In games that uses all threads and wattage go through the roof the 9800x3d is a massive step forward in performance.
I've sold over 50 computers I built as a hobby, none have had a "golden sample" or "oc dud". The silicon they use for these is already near the center of the wafer, the variance will be minimal. I know people like to gamble and make everything into a lottery, but in this case it's a myth.
Just pulled 23400 with +75 boost, Auto Scalar, and -40 All cores. Double-checking extended stability now but damn this thing feels SNAPPY and STABLE! 1.105V max voltage and 5.3Ghz max clock. Thank you for the recommendations as my new build is using an open mesh case and needs to stay quiet. Phantom Spirit 120SE capped at 50% fan speed didn't breach 80C. This absolutely owns.
Good stuff Bryan. Honestly AMD has these parts and their boost algorithm pretty much maxxed out for gaming. I'm still glad to see the advances in X3D stacking though since the chips can run cooler. As others have noted, where PBO and overclocking seems to have the biggest benefit is where the 9800X3D is power limited in all-core productivity. Apps like Cinema4D, 7Zip, v-ray really stretch their legs with an undervolt and unlocked power.
Yeah, people seem to forget 400avg fps with 50 fps 1% lows still means shit gameplay. Average is irrelevant, you want to maximize the lows to hit your hz, which for 240 for example is pretty hard in some games
Im not sure on gaming performance when comparing with the 7800x3d, but with the 9800x3d, when it comes to Cinebench 23, im getting a 24.000 score on multicore (scaler 10, +200 boost, -25 on curve, FCLK 2200, RAM at 6000, cl30), while with 5800x3d i was getting 14.400. Thats a 66% improvement, so i am pretty happy about it.
Appreciate your comparison. Would it make more sense to compare Cyberpunk at very high settings with everything maxed out as this is the more likely use-case for someone considering an upgrade from 7800x3d to 9800x3d
Curve optimiser -offset seems to be the go for amd. With the inbuilt unchangable pb algorithm (not pbo) in the cpus, running at a lower voltage is less heat and it lets it hit higher frequencies.
I want to see a budget roundup of what the 8000 series can comfortably hit at 1:1 on stock/budget cooling and an A620M or budget B650 since it has a better node (4nm monolithic) I/O and memory controller. 8400F is now around $90 on AliXP and 8600G is retailing $20 under the 7600. They have less cache, but this could be overcome overall, or at least by relative value, if they easily take 6800/3400/2266 and the benchmarks bear fruit.
With my testing. I found that Memory tuning will give me more performance than overclocking. Although I was able to get my 9800x3d to 5.425ghz at -25. Stays nice and cool. Didn't hit the silicon lottery as I cannot get 6400mhz at a 1:1 ratio like I did with my 7800x3d.
now that the 3d v-cache memory is under the compute die , do u think that the memory will overheat in turn reducing its longivity. All this is done to make the cpu overclockable at the cost of its longivity maybe... initially i thought the the memory is more heat sensitive so its wise to keep it in direct contact with the ISH.....................
I just built my pc yesterday with the 9800x3d and 6400mhz cl30 ram and it's awesome. My stock clocks are 5388mhz in game which is decent I don't think basically like a OC STOCK hahaha
Been running a 5900x at 4.8ghz all cores since 2020 ...better performance than boost.. ill be trying a manual overclock on this cpu too. If it doesn't impress ill revert to pbo boost.
It's tough, I love traditional OC. But I only need a single CCD. Getting a single CCD CPU without 3D cache and OC it turns out slower than just getting a 3D part. My previous 5950X was great fun to OC and dial in. Those days are gone.
Brian, can you experiment with Curve Shaper? From what I know, Cure Optimiser changes the entire voltage range while Shaper can be tuned for specific voltage points?
yep. Depending on which workload you type want to boost, you can curve shape the other frequencies. For example, I'm playing around with having Curve Optimizer -20, then Low and mid frequencies at -5, but high and max frequencies I have like +10 ( except for high temp is +0 ) I'm just too lazy to go and benchmark every thing to compare but I tell myself I'm squeezing the most out of my gaming performance lol. I'm maxing at 5.35Ghz comfortably tho
In Ryzen 7 9800X3D any kind of latency issue compared to its previous gen? Please make a video about it.... like 10th gen to 12th gen processor latency video you made before. Thanks man.
I lowered mine 200 mhz and -40 on all cores and lost any were from 1-6 fps in average benchmarks but lowered temps by 6-10c and watts by 10-15 depending on the game I think the last 2-300mhz cause a heavy power draw for small gains. I would like to see what you come up with…
Lol the 7800X3D has MORE power draw than 9800X3D does in CP2077 which is an impossible scenario so yes, that confirms it, you are messing something up in the CP77 testing unfortunately. Thanks for the effort though.
Hi Bryan/Brian lol. Do you have a link for any vids on OCing DDR5 on asrock Taichis please, you have so many vids I wouldn't know where to start looking haha. Ty for the vid by the way, think I'll stick with my 7 7800x3D. P.S DDR5 I have is 6400.
I'm struggling to decide on a new motherboard for my 9800x3d. Now that I'm aware OC is not something I require, it's made my selection worse, lol. I want nice aesthetics from the board also. The new system is going into Antec Flux Pro, 32gb GSkill Z5 Royal, Lian Li Hydroshift cooler, and will pair with an RTX 5080 when released. It's purely a gaming system, nothing more, but picking the right motherboard is proving to be excruciatingly painful. Black build not white.
I have been buying new ASRock boards for allot of new parts builds for about the last 5 years, never have had any problems. My main gaming rig is also the Antec Flux Pro. The glass panels have a fairly dark smoke so to have a great looking mobo is ok but you will not see it clearly. Steel Legend has always been my go to. Bryan put out a review on boards last month, check it out.
I like X670E-F/B650E-F/B650E-E. All under $300 and can often be found for around $200 or less. I have X670E-F, and it's decent. Running 6400c28 1:1 with GDM disabled.
@@User-ys7cb MSI has always given me the "I wish I was special" feeling. Swapping out my MSi board next weekend for a Steel Legend. ASRock has always served me well for stability if I Look at the Qvl.
I do have th 9800x3d and i also found that the good old tried +200 boost 10x scalar just doesn't do anything in terms of performance and really pushes temps as high as 93c. And power draw increases to an insane 170w. With respect to undervolting the CPU, I found that an aggressive undervolt performed worse than a mild undervolt. The CPU behaveds totally differently from the 7800x3d. My take away is that the 7800x3d OCed is actually superior. But with future BIOS updates and game optimizations I'm sure the 9800x3d will perform better.
I experience stability issue when I switch from 7500F to 9800x3D after updating Bios on my Asrock B650m. After I switch to MSI X670 & then problem solved. It could be just the VRM or Bios Update... Also, I found no FPS gain on RX 580 RX 6700 RTX 3050 until I switch to 4070 Super on 1080k
hello everyone very good video I bought the 9800x3d and I play in 4k resolution I wanted to know should I stay on windows 10 or is it better to upgrade to windows 11 thanks in advance
More of these CPU’s need to go on sale and honestly these 3rd Party resellers like EBay and Mercari should place a restriction of not being able to sell a product that’s only been released recently. Say nothing newer than 3 months. This would thwart a ton of scalpers due to them not being able to capitalize right away and allowing manufacturers to get products out to the consumers.
If running less voltage helps your performance then your hitting a limit somewhere. in your case it sounds like power as you stated the cooling is very good already. try playing with the power limits to find the sweet spot for your mobo bios rev. And running canned benchmarks wont very often show a cpu bottleneck very well. find any area that has lots of people / cars that causes frame dips where your gpu is not working very hard and then just do 10 loads to make sure you have a fair idea where the average fps really is. and maybe only use the worse case or best case of those 10 runs to give you the best comparison. end of the day this should at least give you a really nice ballpark of the difference in cpus. even if its not exactly perfect.
i love how the audio of the sponsor is not audible now about criticism, people do exxagerate when you make mistakes with names, people seems to be triggered by that those numbers on this video, 7800x3d is great too
it can happen that your manual overclock is not good :D just sayin' . Its like memory, you can have higher memory speed, but at some point it may be hitting your overall performance because, error correction that you didn't notice or just hitting the power envelope. How was that overclocking done? ECLK for example can also be a solution... depends on so many variables ... than the PBO, CO and Turbo Boost norm ...
CP2077 shows a lot of flat results if you use the canned benchmark, but its a pain in the butt to run custom saves and try and keep track of everything so I don't blame you.
Yes I use windows 10, though for the those results just simply ask others to match them, as they 7800x3d results im getting are much better than others in Cyberpunk.
Hello, i have x870e carbon and 9800x3d. From my test i assume the problem is the bios cuz i use the stock, i dont wanna try beta and wait stable one. Seems to be such problems on TDP motherboard if i let decide for me. Max i can reach with CO in PBO is -20 curve +200 override 1x scalar, and in some run on cinebench my 9800 reach 84* around 23600 scoore. I test a few things only for fun cuz in game with this set the cpu reach constant 5.415 (IDK WHY NOT 5.425). I try to set manually TDP and a really good way for me with this idiot bios is to set: 105w(142 110 170) and now the cpu for magic is stable with -30 curve. The constant boost on 5.415 is better and i reach 24200 on cinebench r23 with 78* max. I hope is usefull for someone. What you guys think about it? Maybe with stable improved bios is better?
As its all about the perception on selling a product where numbers and more is better (supposedly) per generation and seems they are squeezing everything out of the factory to be running at near full tilt anyway in order to achieve that. I haven't bothered overclocking CPUs for years. Not like the old days of drawing on it with a pencil to get a chunky benefit. Why bang out more heat and shorten the life of the part for no noticeable benefit. 4% is indistinguishable on anything but a bar graph especially when you are already talking about a 4090 which means that % will be even less to anyone on a lower tiered card anyway.
8-cores is ideal for using the x3d cpus for gaming anyway so no point waiting for the 9950X3D when the 9800X3D is the same performance. The only thing you'll gain with the 9950X3D is worse latency and losing money
Tested settings on my system, temp increases 10 degree for %3.8 performance increase. USELESS. NO PBO OC but -25 NC no fr. drop and temp dropped 5 degrees with same results.Memory oc is best solution.
yeah that differs by generation, on the 9800x3d the boost algorithm has a vid limit of around 1.35 V, it will never request more on older parts pbo and the scalar was always somewhat broken
something is wrong.. My 9800x3d goes any CB at 5.42GHz, using: pbo +200, scalar x2, co -25 Even in prime95 small fft goes around 5.25m occt avx512 5.2.
In this day and age overclocking is a joke. 30 years ago it could really make a useful difference in your framerate and I did a lot of OC back then. Today it's nothing more then a silly hobby. Sorry TYC
I tried putting boost clock modifier too and it also resulted in less 1% low and less average. Just keeping PBO on and everything on AUTO for now until maybe a new AGESA comes out that can maybe fix something
Ur numbers are spot on, but the cpu you have is not the best at -15Pbo or loadline on that mb is very hot. Mine does -25 and it equals 0.1V!!! Its a massive difference. Intels cpu should be run at max 1.5V and brakes at 1.6V. Minus Pbo -25 is the same difference and I just wanted to put 0.1V in context. X3d have never been Bigger numbers are faster, its all about keeping average clocks high/not fluctuating. In Pubg running 370fps I found that an -25 with 200Pbo I needed an Riva fps lock to avoid downclocking, and that gives very good dipps. Its insanly smooth. The 1% lows is now a massive 200-280fps range. Lows are almost as fast as my previous Intel rigs average in Pubg. Pubg uses all the threads it can lay its hands on so its a good tester.. The 9800X3d only shines at a bigger power budget, not surprising since 5nm>4nm isnt a big difference in node shrinkage. 120-150w on water is really nice spot to aim for. The cpu runs cool and feels like it has a decent headroom for ocing, atleast with a good piece of silicon. I'm much more happy with my 9800x3d than my 7800x3d that could only do -5, so better silicon and a huge powerbudget makes for a massive fps uptick in esport gaming. The lows are just so much higher.
I think its time for some new games, that utilize more CPU usage for benchmarks - Although they dont have cookie cutter benchmarks, try Space Marine 2 and Hell Divers.
I'd rather have a stable CPU and a valid warranty than a 5% boost to performance any day. It was a different story back in the day when a Celeron 300 boosted to 450MHz and you got a 50% overclock that matched a more expensive CPU (minux the cache).
over the time, everybody does what you said, i prefer stability, a painless experience, but i used to play with values, push limits, but those times are over, i dont have time and patience for that
The Celeron 300A was actually (a lot of times) faster than the Pentium 2, because, unlike the Pentium 2 (which did have more cache, 512KB, but it was external - separate SRAM ICs outside of the die/core, on the Slot 1 PCB), the Celeron 300A had 128 KB of much faster on-die cache. So, effectively, when overclocked to 450MHz, it was faster than the Pentium 2. Good times, good times...
Curve Optimiser is fine as you are actually bringing temps down. Overclocking…..No chance.
I & I Second Dat!!!
S T A B I L I T Y
Yea like amd gonna know you broke cpu with pb9
I have my 9800X3D at -30 all core stable and I'm running with lower temps and power consumption. No need for higher frequencies since I'm already maxing out my GPU.
Yeah, I feel like all samples should be able to do -20. I have 3 samples doing 30 easy.
My 3070 has shut down 3 times from cpu pushing it sohard
@@RangaAce that shouldn't be happening, something is up
@RangaAce That's due to your GPU more then likely having bad temps or inadequate power.
Check your GPU thermal paste and check your power supply and or connections to the power supply and motherboard
Also i hope you didn't pair a 9800x3d with a 3070
@@forog1 Why? I paired a 7800x3D with my old 2080, because I already plan to upgrade to 5080 5070 ti or 8800xt. Or even Battlemage, should it unexpectedly compete and be viable.
You are using the built in benchmark that is why. Others use a repeatable run in an area that is much more taxing and then you see a larger difference. This holds true for a lot of the game. The built in benchmark is not a very good indicator of total performance.
If you can link me the area or show me a place where I should be testing that would be great!
@@techyescityi bet he didnt expect this answer from you , very mature. love it! just earned yourself a subscriber
@@techyescityDogtown is actually the most taxing area, the lowest fps you get are near The Moth bar where you nawet Alex.
Running around in an area is not reliable anyway, there could be different NPCs, cars, time of day, viewing angle, etc. That's why built in benchmarks exist.
@@Mavis847 In case of Cyberpunk 2077 the whole Phantom Liberty DLC has upped the hardware requirements a lot. In the built-in benchmark I got 40 average FPS on a GTX1060. In Dogtown my FPS would dip into low 20s, and that made me finally upgrade my PC :-)
Hang on a second at 12:13 you have cinebench sccores, how in the world is the 9800XD with X3D Turbo enabled score very slightly below stock or the oc model? Thats not even remotely possible since X3D disables SMT which has a big impact on cinebench scores..
Even my system 9800x3d / 4090 gets close to 24k with +200mhz. Then when i use X3D Boost on motherboard, i lose a decent chunk of points there. (I am on the X670E Gigabyte Masters)
I will double check the X870E Aorus Master. It was a final test I forgot to double check, so might have had it disabled accidentally.
Other results don't make much sense too...
@@techyescity Mine also scores around 24000 with a +200 and a negative curve
Was able to get my hands on a 9800x3d at micro center just before the sell out. Got Super Lucky. I plan on Benchmark versus this with my 5800x3d.
Mine does 5.4Ghz on 1.2V and it ripps. My brother have 5800x3d and that feels 2 gens behind in Pubg atleast. 6400Mhz cl30 you start to get rid of the infinity bottleneck latency wise. I get 370w average fps in Pubg with 200-280 in 1% lows. Ive had 10700k, 12700k, 14900k, 7800x3d, 9800x3d and buildt 2 systems with 5800x3d for others. All systems ended up maxed out overclocked. The 9800x3d overclocked and 140w power budget is the only cpu capabable of running Pubg well without stutters. After 6 years of trying to run that game flawless I'm now happy.
Thanks for this! I always follow your undervolting guides and it's always served me very well!
i advised a friend of mine almost 2 years to wait before upgrading his 6700/1060 pc knowing we simrace and he wants triple screens . he decided he didn't want to wait anymore just before the 9800x3d launch . i advised him get a 7800x3d and 4080super . seeing the benchmarks and knowing he games with 4k resolution . i'm glad he decided to not wait longer . knowing the new assetto game is coming in januari i'm still trying to convince him to go ultrawide vs triple screen . better to have 100fps or more in simracing vs having triple screen but only getting 60fps .
Love your work. Keep it up 👍
I would tune the IF/RAM side for better lows.
Ignore those trolls.....your cred is good with us supporters.
oh lord, tribalism
I'm running +200 -45 offset, and 2x scalar. 8000mhz ram, on a Asus Hero E870 mobo.
1.13v under load, 60 degrees. Very nice.
I have been waiting for this one, I had a feeling you were putting in allot of time on this CPU. Will be looking forward to the underclocking episode also. You always manage to find the sweet spot. Relax and enjoy what you are doing, it makes a difference to quite a few viewers.
So I compared the 2 in a video in 4k anyways, and my cp2077 settings were crazy, it actually showed the 9800x3d did 1% worse then the 7800x3d. The games where I got a massive (20% ) improvement was helldivers 2 and space marine 2 games that were cpu bound with the 7800x3d
1% difference is just measuring error. Essentially no difference
Great video as always. Would be interested in memory scaling of the CPU also.
Welcome back Bryan ❤
You are one of the best reviewers i follow. I bought my first GPU (a rx570) thanks to you and it was a really good choice for me at that time.
It is hard to ignore some of the negative comments, for sure. I really enjoy your channel. Thanks for all the time you put into it. cheers.
Good. I'll leave it at stock values. Can even save a good sum of $$$ by skipping the overkill highend boards.
Well, I'd leave it at stock clocks but undervolt it with curve optimizer for better temps and it will boost itself if need be up to 5.2ghz
@@forog1 Mine does -25 5.4Ghz on what equals to stock voltage. I think silicon quality varies a lot. I have Asrock Hdv and after tuning the ram it does 8000Mhz on a cheap motherboard. It also does 5.5Ghz but that makes the infinity clocks run outside optimimum speed ranges/dividers unfortunately. If you get a good cpu like mine you can oc a ton without it throttling. And the cpu runs extremely cool, Amd have made a massive effort to get the heat out of the chip this time around.
@@impuls60 Mine also does CO -25 all cores with a +200mhz boost, and 1x scalar. I am getting a score of 23750 in cinebench r23 multi at 85-89 temps. All of this with a cheap Gigabye B650 X AX that I bought for $140 back when I got my 7800x3d. But when I tried 5.45 all-core OC with 1.225v on the vcore, I throttled at 95c in cinebench almost instantly. My 7800x3d never even reached 70c with this same cooler, so my 9800x3d definitely runs A LOT hotter.
@@impuls60 good to know. I'm just waiting on the 9950x3d hopes it will be just as amazing at 9800x3d. Hope to God it would be 29800x3ds's is 1 package lol only can hope
@@xblur17It doesn’t run hotter. It is hotter in that situation because you’re overclocking it compared to stock/PBO 7800X3D
it's always worth tuning to some extent, there is a lot of people saying they would rather have reliability but as soon as you use the pre baked expo or xmp ram settings your over stock any so it's always worth tuning to find lower voltages mainly.
look as the soc voltage in the video of 1.25v that's pretty high for zen 5 and probably doesn't need to be above 1.05v for that ram speed.
Never forget most people are clueless
my SoC wasnt able to go lower than 1.25v. But then again, im using a 64GB 6400 kit
1.25 soc is safe and not unusual if you’re oc’ing lol
This is for context not a shameless plug. When I put my 7800X 3D up against my 9800X 3D in Cyberpunk using the in game benchmark.
Just like you. 1080p using Low settings with a 7900 XTX Nitro+. 311 for 7800X 3D.....317 for the 9800 X 3D. I posted this around 6 days ago.
Looks gpu limited.
So basically a nothing burger, from what I can tell, my 9800x3d is actually pretty poor silicon too, so my 7800x3d could be a better bin, hence the slight change.
Cyberpunk isn’t heavy on the CPU except the populated city areas that Hardware Unboxed tests in
Overclocking? Maybe if you got a good cooler.
But I prefer undervolting with the curveoptimizer and now also curveshaper. Doing that I reduced the powerusage by 20% (-30W of 145W) under synthetic load while gaining 3% performance with 23465 points in CBR23 (my base was ~22800). The temps also got 10-15°C lower.
I was able to use -20 on the optimizer and -10 on the shaper on high and max clocks. Other then that I didn't touch anything.
He has a 420 AIO. Not good enough?
@@frommatorav1not really, for heavy overclocking. You can make your whole pc ambient temp for example, by using a big external copper radiator. I wouldnt say AIOs are particularly suited for heavy overclocking, custom loops are more for that job
@@ppeez hey bro, we were clocking 50% more GHz on Celerons and Core2Duos while on tower coolers. All OC WR's are set with LN2
Appreciate you, Bryan for your testing methodology. The silicon lottery could also play a factor. Maybe your 7800x3d is a golden sample, or maybe the 9800x3d is an oc dud. It's still the best 2 CPUs for gaming.
In my own testing 9800x3d is about 15-20% higher 1%lows in Pubg over my meh 7800x3d. -5 on 7800x3d and -25 9800x3d (-25 equals 0.1V equals 200Mhz for free)In games that uses all threads and wattage go through the roof the 9800x3d is a massive step forward in performance.
This time around from reading the comments, it is exactly what I think is happening!
I've sold over 50 computers I built as a hobby, none have had a "golden sample" or "oc dud". The silicon they use for these is already near the center of the wafer, the variance will be minimal. I know people like to gamble and make everything into a lottery, but in this case it's a myth.
Which is the best for multitasking and gaming, video editing and content creator
Just pulled 23400 with +75 boost, Auto Scalar, and -40 All cores. Double-checking extended stability now but damn this thing feels SNAPPY and STABLE! 1.105V max voltage and 5.3Ghz max clock. Thank you for the recommendations as my new build is using an open mesh case and needs to stay quiet. Phantom Spirit 120SE capped at 50% fan speed didn't breach 80C. This absolutely owns.
Bring back the "I wish that I was a madman" theme. This is like a Star Wars movie without the 20th Century Fox promo.
Good stuff Bryan. Honestly AMD has these parts and their boost algorithm pretty much maxxed out for gaming. I'm still glad to see the advances in X3D stacking though since the chips can run cooler.
As others have noted, where PBO and overclocking seems to have the biggest benefit is where the 9800X3D is power limited in all-core productivity. Apps like Cinema4D, 7Zip, v-ray really stretch their legs with an undervolt and unlocked power.
Why do the 1% lows seem to shit the bed with the OC ?
I’d rather have steady lows than a couple over all frames. Just me.
There is only one plausible reason: It's not stable.
Yeah, people seem to forget 400avg fps with 50 fps 1% lows still means shit gameplay. Average is irrelevant, you want to maximize the lows to hit your hz, which for 240 for example is pretty hard in some games
Im not sure on gaming performance when comparing with the 7800x3d, but with the 9800x3d, when it comes to Cinebench 23, im getting a 24.000 score on multicore (scaler 10, +200 boost, -25 on curve, FCLK 2200, RAM at 6000, cl30), while with 5800x3d i was getting 14.400. Thats a 66% improvement, so i am pretty happy about it.
Undervolting and tight ram timings is the way 2 go 🤘🤘🤘
Appreciate your comparison. Would it make more sense to compare Cyberpunk at very high settings with everything maxed out as this is the more likely use-case for someone considering an upgrade from 7800x3d to 9800x3d
Curve optimiser -offset seems to be the go for amd. With the inbuilt unchangable pb algorithm (not pbo) in the cpus, running at a lower voltage is less heat and it lets it hit higher frequencies.
Interesting topic. I remember in the Skylake days being able to push more with OC. Now I'm not sure I want to FAFO as much.
What CPU would you choose for the "most efficient"? I'm building a traveling gaming PC and I want the most frames per watt. 1080p
I am running 200 and -45, x10 scalar temp always below 70 , power always below 100w , so far so good
your mic sound is very good!
0:06 the thermal paste on the pointing finger says a lot. 👍
I want to see a budget roundup of what the 8000 series can comfortably hit at 1:1 on stock/budget cooling and an A620M or budget B650 since it has a better node (4nm monolithic) I/O and memory controller. 8400F is now around $90 on AliXP and 8600G is retailing $20 under the 7600.
They have less cache, but this could be overcome overall, or at least by relative value, if they easily take 6800/3400/2266 and the benchmarks bear fruit.
With my testing. I found that Memory tuning will give me more performance than overclocking. Although I was able to get my 9800x3d to 5.425ghz at -25. Stays nice and cool. Didn't hit the silicon lottery as I cannot get 6400mhz at a 1:1 ratio like I did with my 7800x3d.
Raise the voltages
If your memory can do it try for 8000mhz. It's faster than 6400 1:1.
i can do it but SoC voltage needs like 1.3 for stability...
6400mhz at 1:1 is kind of a lottery, which is why I went for 6000mhz RAM.
now that the 3d v-cache memory is under the compute die , do u think that the memory will overheat in turn reducing its longivity. All this is done to make the cpu overclockable at the cost of its longivity maybe... initially i thought the the memory is more heat sensitive so its wise to keep it in direct contact with the ISH.....................
Can I ignore pbo and just use asus ai automatic overclock instead ? My 9800x3d is already boosting to 5.4 this way
I just built my pc yesterday with the 9800x3d and 6400mhz cl30 ram and it's awesome. My stock clocks are 5388mhz in game which is decent I don't think basically like a OC STOCK hahaha
How do I stop my 9800x3d stop thermal throttling after 1 hour playing rust, 420mm cooler, running performance settings.
Why am I even watching it when I will ride my i9 9900k till that day it doesn't play my games anymore? Because I like the presentation!
Been running a 5900x at 4.8ghz all cores since 2020 ...better performance than boost.. ill be trying a manual overclock on this cpu too. If it doesn't impress ill revert to pbo boost.
What would be much more helpful are the actual steps you took in the Aorus master MB to achieve best performance.
Hello is just doing an undervolt worth it from your testing? with curve optimizer?
It's tough, I love traditional OC. But I only need a single CCD. Getting a single CCD CPU without 3D cache and OC it turns out slower than just getting a 3D part. My previous 5950X was great fun to OC and dial in. Those days are gone.
Brian, can you experiment with Curve Shaper? From what I know, Cure Optimiser changes the entire voltage range while Shaper can be tuned for specific voltage points?
yep. Depending on which workload you type want to boost, you can curve shape the other frequencies.
For example, I'm playing around with having Curve Optimizer -20, then Low and mid frequencies at -5, but high and max frequencies I have like +10 ( except for high temp is +0 )
I'm just too lazy to go and benchmark every thing to compare but I tell myself I'm squeezing the most out of my gaming performance lol. I'm maxing at 5.35Ghz comfortably tho
In Ryzen 7 9800X3D any kind of latency issue compared to its previous gen? Please make a video about it.... like 10th gen to 12th gen processor latency video you made before. Thanks man.
I lowered mine 200 mhz and -40 on all cores and lost any were from 1-6 fps in average benchmarks but lowered temps by 6-10c and watts by 10-15 depending on the game I think the last 2-300mhz cause a heavy power draw for small gains. I would like to see what you come up with…
Just had a random question, do you know if the 9800x3d is compatible with the B650i Jginyue Night Devil?
Lol the 7800X3D has MORE power draw than 9800X3D does in CP2077 which is an impossible scenario so yes, that confirms it, you are messing something up in the CP77 testing unfortunately.
Thanks for the effort though.
Very likely as most of his results are off.
@@darreno1450The lower 1% lows indicate instability.
Hi Bryan/Brian lol. Do you have a link for any vids on OCing DDR5 on asrock Taichis please, you have so many vids I wouldn't know where to start looking haha. Ty for the vid by the way, think I'll stick with my 7 7800x3D. P.S DDR5 I have is 6400.
I'm struggling to decide on a new motherboard for my 9800x3d. Now that I'm aware OC is not something I require, it's made my selection worse, lol.
I want nice aesthetics from the board also.
The new system is going into Antec Flux Pro, 32gb GSkill Z5 Royal, Lian Li Hydroshift cooler, and will pair with an RTX 5080 when released. It's purely a gaming system, nothing more, but picking the right motherboard is proving to be excruciatingly painful. Black build not white.
I have been buying new ASRock boards for allot of new parts builds for about the last 5 years, never have had any problems. My main gaming rig is also the Antec Flux Pro. The glass panels have a fairly dark smoke so to have a great looking mobo is ok but you will not see it clearly. Steel Legend has always been my go to. Bryan put out a review on boards last month, check it out.
any B650 that has the io you require make isn't really that important.
I like X670E-F/B650E-F/B650E-E. All under $300 and can often be found for around $200 or less. I have X670E-F, and it's decent. Running 6400c28 1:1 with GDM disabled.
My MSI X870E Carbon seems decent so far. ASRock should also be fine.
@@User-ys7cb MSI has always given me the "I wish I was special" feeling. Swapping out my MSi board next weekend for a Steel Legend. ASRock has always served me well for stability if I Look at the Qvl.
can you do a optimization/UV video? Just like with the 7800X3D and 5800X3D
I do have th 9800x3d and i also found that the good old tried +200 boost 10x scalar just doesn't do anything in terms of performance and really pushes temps as high as 93c. And power draw increases to an insane 170w.
With respect to undervolting the CPU, I found that an aggressive undervolt performed worse than a mild undervolt. The CPU behaveds totally differently from the 7800x3d.
My take away is that the 7800x3d OCed is actually superior. But with future BIOS updates and game optimizations I'm sure the 9800x3d will perform better.
Hello, I saw that you talk a lot about the x58 platform, what is the best GPU for a xeon x5650 Overclocked and its strongest variants?
Did you test only with AsRock or did you also did test with the Gigabyte?
I experience stability issue when I switch from 7500F to 9800x3D after updating Bios on my Asrock B650m. After I switch to MSI X670 & then problem solved. It could be just the VRM or Bios Update... Also, I found no FPS gain on RX 580 RX 6700 RTX 3050 until I switch to 4070 Super on 1080k
What happened to the S4 of the $100 flip up challenge?
Hey mate, what motherboard are you using?
hello everyone very good video I bought the 9800x3d and I play in 4k resolution I wanted to know should I stay on windows 10 or is it better to upgrade to windows 11 thanks in advance
You testing it on win 11 or win 10
i think the cpu is out of the box at its limit. and when you overclock the clockstreching begins. or maybe its cachethrottling becaus of the temps.
More of these CPU’s need to go on sale and honestly these 3rd Party resellers like EBay and Mercari should place a restriction of not being able to sell a product that’s only been released recently. Say nothing newer than 3 months. This would thwart a ton of scalpers due to them not being able to capitalize right away and allowing manufacturers to get products out to the consumers.
If running less voltage helps your performance then your hitting a limit somewhere. in your case it sounds like power as you stated the cooling is very good already. try playing with the power limits to find the sweet spot for your mobo bios rev. And running canned benchmarks wont very often show a cpu bottleneck very well. find any area that has lots of people / cars that causes frame dips where your gpu is not working very hard and then just do 10 loads to make sure you have a fair idea where the average fps really is. and maybe only use the worse case or best case of those 10 runs to give you the best comparison. end of the day this should at least give you a really nice ballpark of the difference in cpus. even if its not exactly perfect.
I am wondering if it is the motherboard\memory Brian is using for the cyber punk benchmark?
If all I care about is temps, should I just go w/ a -30 undervolt all core & tune ram?
If it runs stable, yes. Mine doesn't run stable at -30 all core.
i love how the audio of the sponsor is not audible now
about criticism, people do exxagerate when you make mistakes with names, people seems to be triggered by that
those numbers on this video, 7800x3d is great too
it can happen that your manual overclock is not good :D just sayin' . Its like memory, you can have higher memory speed, but at some point it may be hitting your overall performance because, error correction that you didn't notice or just hitting the power envelope. How was that overclocking done? ECLK for example can also be a solution... depends on so many variables ... than the PBO, CO and Turbo Boost norm ...
Overclocking is dead. Undervolting is in.
Try turning off GPU Power data before running benchmark, it seems that is causing an issue, and seems to be a bigger issue on 9800x3d.
I swapped back to the 7700X, been great so far for the 4090
If you overclock to 5600mhz. Do you also set undervolt to -15?
CP2077 shows a lot of flat results if you use the canned benchmark, but its a pain in the butt to run custom saves and try and keep track of everything so I don't blame you.
I really want to figure out what the best ram is for that.I need sixty four gigs
Why does everyone not mention there is another x3d coming?
So OC is dead , do UV at stock settings, not worth on 9800X3D
Still testing on Win10? If so, I wonder if your CP0277 result would change on Win11 24H2.
Yes I use windows 10, though for the those results just simply ask others to match them, as they 7800x3d results im getting are much better than others in Cyberpunk.
cant wait undervolting! sub given
Hello, i have x870e carbon and 9800x3d. From my test i assume the problem is the bios cuz i use the stock, i dont wanna try beta and wait stable one. Seems to be such problems on TDP motherboard if i let decide for me. Max i can reach with CO in PBO is -20 curve +200 override 1x scalar, and in some run on cinebench my 9800 reach 84* around 23600 scoore. I test a few things only for fun cuz in game with this set the cpu reach constant 5.415 (IDK WHY NOT 5.425). I try to set manually TDP and a really good way for me with this idiot bios is to set: 105w(142 110 170) and now the cpu for magic is stable with -30 curve. The constant boost on 5.415 is better and i reach 24200 on cinebench r23 with 78* max. I hope is usefull for someone. What you guys think about it? Maybe with stable improved bios is better?
As its all about the perception on selling a product where numbers and more is better (supposedly) per generation and seems they are squeezing everything out of the factory to be running at near full tilt anyway in order to achieve that. I haven't bothered overclocking CPUs for years. Not like the old days of drawing on it with a pencil to get a chunky benefit. Why bang out more heat and shorten the life of the part for no noticeable benefit. 4% is indistinguishable on anything but a bar graph especially when you are already talking about a 4090 which means that % will be even less to anyone on a lower tiered card anyway.
Using a 7900XTX (OC'd) upgrading from a 5900X, would I better off just getting a 7950X3D over waiting on a 9950X3D? I game at 1440P at this time.
8-cores is ideal for using the x3d cpus for gaming anyway so no point waiting for the 9950X3D when the 9800X3D is the same performance.
The only thing you'll gain with the 9950X3D is worse latency and losing money
I would love to see enough of a undervolt to bring it down to same power draw of 7800xrd and then compare them.
Why disable soc uncore?
Tested settings on my system, temp increases 10 degree for %3.8 performance increase. USELESS. NO PBO OC but -25 NC no fr. drop and temp dropped 5 degrees with same results.Memory oc is best solution.
the scalar is scary cause at 9x i saw my VID go upto 1.500V on my 5800X with llc set at level 5
yeah that differs by generation, on the 9800x3d the boost algorithm has a vid limit of around 1.35 V, it will never request more
on older parts pbo and the scalar was always somewhat broken
something is wrong..
My 9800x3d goes any CB at 5.42GHz, using: pbo +200, scalar x2, co -25
Even in prime95 small fft goes around 5.25m occt avx512 5.2.
AGESA 1.2.0.2b might sway the gaming results a bit.
Tech YEAH!
it has headroom but do you really need it - maybe not until the next fastest cpu comes out in a couple of years
my $10 scrapyard X470 boards and aliexpress 5700X3D makes it not worth it yet.
7950x3d vs 9800x3d
i had a 7950x3d and bought the 9800x3d was this a downgrade
maybe ask before next time lol
Just test it and sell the other one
9800X3D is the better gaming CPU obviously.
In this day and age overclocking is a joke. 30 years ago it could really make a useful difference in your framerate and I did a lot of OC back then. Today it's nothing more then a silly hobby. Sorry TYC
I tried putting boost clock modifier too and it also resulted in less 1% low and less average. Just keeping PBO on and everything on AUTO for now until maybe a new AGESA comes out that can maybe fix something
curious exploration, but for overclock to be useful it needs to be same (and stable) in all tests.
I would be more interested in watching a video on how to Undervolt the 9800 X3D than overclocking personally.
Ur numbers are spot on, but the cpu you have is not the best at -15Pbo or loadline on that mb is very hot. Mine does -25 and it equals 0.1V!!! Its a massive difference. Intels cpu should be run at max 1.5V and brakes at 1.6V. Minus Pbo -25 is the same difference and I just wanted to put 0.1V in context. X3d have never been Bigger numbers are faster, its all about keeping average clocks high/not fluctuating. In Pubg running 370fps I found that an -25 with 200Pbo I needed an Riva fps lock to avoid downclocking, and that gives very good dipps. Its insanly smooth. The 1% lows is now a massive 200-280fps range. Lows are almost as fast as my previous Intel rigs average in Pubg. Pubg uses all the threads it can lay its hands on so its a good tester.. The 9800X3d only shines at a bigger power budget, not surprising since 5nm>4nm isnt a big difference in node shrinkage. 120-150w on water is really nice spot to aim for. The cpu runs cool and feels like it has a decent headroom for ocing, atleast with a good piece of silicon. I'm much more happy with my 9800x3d than my 7800x3d that could only do -5, so better silicon and a huge powerbudget makes for a massive fps uptick in esport gaming. The lows are just so much higher.
tried setting all c2077 all low, native, 80 fov, around 322 fps average 10 runs. 9800x3d, 32gb 6000mhz, 4090
Wait what? default is better than PBO?
I think its time for some new games, that utilize more CPU usage for benchmarks - Although they dont have cookie cutter benchmarks, try Space Marine 2 and Hell Divers.