EDIT: People asking how frequency/timings would scale on a CPU side, you can watch this video I made some months ago: th-cam.com/video/B5K8Rg-oDwU/w-d-xo.html Get your V-Color RAM Kits here: v-color.net/ And here we are again! This time testing Several RAM Frequencies with the Ryzen 7 8700G IGPU (Radeon 780M) to see how much performance we can get out of this APU :D
I'm beginning to wonder how DDR5 at 8600 MT/s will preform with this iGPU, I think that's the fastest yet. But I'm not sure. There may be some 8800 MT/s models out there
Just one note: 6000 CL38 might sound like the same latency as 8000 CL38, but it isn't. The CL is linked to the clock, basically the number of cycles it takes to get access. These CL values are more or less meaningless for themselves, it would be better to compare the actual latencies in nanoseconds. For some who might be confused a simple example: 4000 CL40 has the same latency in ns as 8000 CL80 The 6000 CL38 has a CAS latency of 12.67ns, whereas the 8000 CL38 has a CL of 9.5ns
also in general primary timings have prove to be rather insignificant as long as trcd&trp are under 13ns and CL under 10.67ns if the secundary timings were properly set up. for example the corsair DDR5 8000CL38-50-50-50-128-178 1.4V is worse than Gskills CL40-48-48-48-128-168 1.35V eventough the Cas latency is loweron the corsair kit. Im running DDR5 8000CL40-48-48-48-116-116 1.4V and it is better than my previous Corsair 6000CL30-36-36-36-76-112 1.4V altough in HUB's Zen4 scaling video Gskills 6000CL30-38-38-38-96-96 Kit turned out to be slightly better than corsairs CL30 Kit cause Gskills came with better secondary timings.
A quality of work that rivals Hardware Unboxed, you are proving the quality of your chanel, and it should be on par with other big youtube chanels, keep up the good work 👍
This is exactly what I wanted youtubers to test out. Desktop allows to achieve full speed of these APUs due to faster memories, but no one was using more than DDR5 6400 and then were disappointed that these end up slower than a budget GPU.
More dosn't make much sense to be honest, when you go over 6400mhz you gonna end up with a 1000$+ system since 8700G and AM5 mini itx mobos are prices insane.
I know you know this, but for folks who might not - DDR5 6000, 6400, 7200, 7600 and 8000 is in MegaTransfers per second (MT/s). People often use MHz here, because indeed it is directly tied to frequency, but if you go into BIOS and after applying DDR5 6000 profile you see that your RAM goes "only" to 3000 MHz - Don't Panic! That is how it's supposed to be. Real frequency of DDR is always exactly half of the number of MT/s. Because DDR means "Double Data Rate" - it means that operation (read, write, many other more complex stuff) can happen twice for every "frequency tick". It is important to know that, because timings will often be displayed in such a way (though BIOS-es have differences and they make mistakes as well) that it will be tied to number of "cycles" that are required to wait until next operation can be done. And when you count nanoseconds of latency, you need to know "frequency" not "MT/s". That might be the reason why sometimes calculations are off by 2, 8 or 16 times. You might have forgotten to divide by 2 for frequency or divide/multiply by 8 to go from bytes to bits and vice versa, or both.
@@AncientGameplays I mean, it's not a big thing, but if someone tries to calculate latencies this comes up the most. Though also some BIOS-es show number of MT/s as "MHz". Pretty much everyone, expert or not make this mistake - I catch myself doing it regularly and in most cases it doesn't matter. Except for people freaking out in CPU-Z 😉or as I said those that calculate latencies.
Thank you for doing the tests everyone should be doing with these iGPU's. Honestly i would like to have a few hundred euros to spare just to build a PC with a 780M just to see how far i could push it.
Very nice video, love the effort and digging in to it, especially with the tweaked timings I'd love to see a completely stock 8700g vs highest tuned memory + oc'd 8700g chip as well, to see how much further you could boost performance if they would allow this amount of tweaking in handhelds or laptops, it would be crazy
kinda crazy when you think about how much faster and more stable ddr5 is going to get before this platform is dead, we're gonna have 8700gs going BRRRRRRRRRRR
I think this is a big reason we see less low end ~$200 cards in development because iGPUs will handle that part of the market vs a 1650-4050 now and in the near future
@@UmmerFarooq-wx4yo at what point would these same people begin to be held back by a cpu bottleneck and have to upgrade anyway ? A “5050” would be knocking on a 2080’s door power wise so you are most likely going to bottleneck. Also we’re starting to see these new low tier cards use less and less pcie lanes so these cards would restricted with older cpu/motherboards anyway. At that point the used market is a better solution than a new low end card
Probably by the end of the year we will see AMD's "Strix Halo" architecture which has a 256 bit (instead of 128 bit) memory interface. It is LP-DDR5x only, so no desktop variant, even a cut down version, is possible. So we should see about double this performance in handhelds soon enough.
@@AncientGameplays Wait so are you playing this on a smart fridge with a Ryzen 7 8700G, why would a smart fridge have that? Or was that a joke I didn't get hehe
100% stable is an extremly huge claim with just enabling xmp when most of the time 8000hz needs tweaks, good voltage, fans on ram, cpu lottery and high end motherboard. What is your testing methodology for stable ram? My bet is that you would get errors in 1 minute in karhu ram testing. Running high speed unstable ram just for benchmarks is not genuine since instabilities might ruin some peoples PCs..
@@sharky9056 my testing is these games tested, TM5 running for 5 hours without a single error and running these games smoothly. Ram voltage 1.4v, soc at 1.3v. Done
Games are not a good test for ram and TM5 ram test is laughable even on the hardest custom presets... only pi and karhu are good test for exposing errors. Reddit post in ram community incoming with your claims.
My friend Karhu sucks (sorry). Used it for several months and had a ram instability in the past due to trusting that same software, the minute I started TM5 with the 1usmus profile I had an issue in the first seconds. Everything I testing with TM5 in the last years, all my systems, are rock stable, so don't come to me telling me that Karhu is better than TM5 and that "games don't stress anything" on the RAM side. I've been there on the same boat as you and I know for a fact that's not true@@sharky9056
Not fully sure about the 7000 series, but for 5800X3D higher clocked RAM did not make that much difference. Simply because with the huge chunk of local L3 cache, less transfers to RAM were necessary. And the iGPU in the 7000 series CPUs is more or less a basic display adapter, it's too slow anyways.
@@PokeStoryHarry There are always differences depending on the engine. However if you are playing a game that is depending on RAM speed, it most likely does not benefit from the additional CPU cache (or to a limited extent). CPU frequency could also be a factor. That's why channels like this one or HUB are important, they check for such differences.
Thanks Fabio! I just bought a laptop with a Ryzen 7840U (which has 780M) on LP-DDR5x 6400Mhz, and was wondering how much I "lost" by not choosing another (much more expensive) vendor for a similar product with 7500MHz. This video gave me the overview I has been looking for.
Honestly for Integrated every single FPS matters when your performance is already pretty poor. For example 48 vs 58 fps in Spider Man is a huge difference going from laggy to almost smooth gameplay. Now obviously buying hella expensive 8000mhz kits wouldn't make sense but there is a GSKill Neo 6400/CL32 which is Hynix A-die and it can be manually pushed out to 8000. (My Corsair Dominator Titanium 6400/cl32 is A-die though but more expensive than the G.Skill kit and it's also way easier to find A-die Gskill kits as the part number on the back of the ram stick ends with the chip type.) Great video though.
@@heinzletzte.6385 Even if the architecture is Zen 4 on both, the whole die design is different. Ryzen 7000 is chiplet, while 8000 is monolithic which means the mem controller and infinity fabric also work very differently. 8000 series design is also meant to support faster lpddr5 for low power devices so that could also help stability at higher frequencies. People are pushing these new chips up to 10000MT/s, while Ryzen 7000 is still stuck at 8000MT/s
@AncientGameplays, Fabio, I'm sure you have this already lined up but, just in case you don't yet, can you please do an episode on the new drivers - 24.3.1? In particular I'm sure many will be interested in the supposed unlock of overclocking for the memory on the 7900 GRE. I've found this to be disappointing personally, since I can only reach around 2376, about a 60 Hz increase on the memory clock. Beyond that they don't seem to be stable at all. I'm wondering if AMD will release updated drivers in the future that address this? Anyway, just a suggestion. Keep up the excellent work you're doing and thank you.
@@AncientGameplays As we say in the UK, you are a scholar and a gentleman Sir. Much appreciated. 😊 As a side note here, I help with a medium sized channel, about 27, 000 subscribers, which is done by my lovely wife and it would be great, when you have some time, if you could come on her show and chat a bit about your enjoyment of computers and technology. I think you'd be a great guest. Valeu por tudo, Fabio.
Hey brother, I just wanted to let you know, that you can manually get your FCLK and UCLK up to 2400mhz with this chip without decoupling the frequencies. I was able to get mine to 2400 FCLK for 7200mhz, and 2467 for 7400mhz. I haven't seen anyone try this on TH-cam. I was able to do this on a B650m MSI mortar wifi nothing special. I used 1.4v for ram. I use this chip with my 4070, and haven't tried the APU yet but without being decoupled it should give even more performance. If you have time it would be cool to see if there is any benefit to raising your FCLK that high. Love the channel!!!
Yes Hynix chips OCd and tuned at 7200-7400 is the optimal for 8000G APUs, not 6400, not even 8000, cause many games like the lower latency, so generally 7200 tuned is the best for these APUs :)
@@DragonOfTheMortalKombat absolutely sir. Fabio did not include what the ram configuration was(that I remember?) for that video. You could open both videos side by side and compare results on game to game and see the % difference and assume that vs the rx 580 etc and give you an idea. Report back what ram configuration you think that the "vs rx 580" video was using. It'll be a lil game for you. 😜
@@christophermullins7163 before making a fool of yourself, did you go and watch the video ? In the RX 580 vs 780M ddr5 6000 cl 36 was used. I am talking about ddr5 8000 with manually tuned memory timings. I never said that he didn't mention what memory was used. Lmao, you think you I wouldn't bother verifying kid 🤪
Be careful, the "Tweaked" versions are not the same for everyone, if you buy a cheaper kit at lower frequencies you won't be able to tweak it as much. Still stands that you'll match or be better than more expensive without tweaking but something to take in mind.
I'm looking at DDR 7200, CL34 for my Ryzen Zen 5 build. I won't use IGP, but 7200 DDR 5 gives you a nice mix between high bandwidth and latency at CL34.
It's well documented that decoupling your CPU clock and memory clock leads to _decreased_ performance, not improved performance. It may be beneficial in this niche case that's memory starved, but will hurt overall performance everywhere else. Memory speed matters, but it's also important to be in sync with the CPU's memory controller, and we still have no idea if Zen5 will come with an improved memory controller that doesn't slam into a brick wall at 6000mt/s.
I know your content is mostly focused on gaming, but can you test if the IGPU helps improve performance in productivity applications like premiere pro, after effects or davinci resolve? and how it compares against an 7700X? I know the Intels APUs have an impact on programs that decode video, but I found anything about the AMD APUs doing the same.
I have the Ryzen 7 5700g CPU at 1.32V 4.5Ghz all cores i think :D paired with 6700XT tho. And actually its Good and very effectively with low voltage in big loads :) so the Ryzen 7 8700g must be a Beast.
Thank you. Great video. I thought 5600 was standard DDR5 speed for AM5. I imagine uplift from that is even better, you get like 20-25% better result. I have only miniPC with 7840HS that can not be really tweaked (BIOS version shows 0.01 ?!! )
:| i have a 10100f+gt1030+8gb ddr4 2999... So i ll probably wait for the upcoming apu, which to have strix point igpu. 😅 I am not a laptop guy so, I'm not talking about the 8850hs. I'm talking about the 9700g (speculation) :| i know it's going to take around 2 years for the 9700g to hit the market. Pray for the upcoming glory 😑🙏🐣 (780m is not so good) Also: rdna 3.5 is actually a refresh of rdna 3. We may be able to see a refreshed processors like 8750g (😑🙏 praying for it to come out asap.) 😑🙏 Praying for the upcoming strix halo as well (which has 40cu igpu) Please be cheaper (and be a desktop gpu... Not the laptop) 😑🙏 guys pray.
Dragon Dogma 2 in 4 hours, so people are going to be really testing that game out to see what exactly really messes up the CPU usage... But I think their game engine is the issue, basically can't handle the heavy workload they ambitiously put on it.
Hello , excellent video , the reason I went with 8700G and 6000mhz , 30CL Expo from Kingston.I set up only Expo profile , I am satisfied with results but I would like to optimize the timings as you did.I am a bit of noob for this but I cannot find how to set/optimize DRAM timings and all other configuration? Is there a separate video for this?Thanks in advance :)
Why did you do that ? Why choose unrealistic timings ? Nobody runs 6000@cl38. Why not start with conventional timings, 5600cl36 / 6000cl30 / 6400cl36 ?
@@AncientGameplays yes! 4070 super is more expensive; the model is Msi rtx 4070 super gaming x, as for the Rx 7900 gre its the xfx rx 7900 gre qick 319 bulk edition
Hi, just a thought. What if you ran the same exact test, but add in a dedicated GPU. Would the higher freq tweaked memory scale the same? At 1080p of course to test the CPU limit. Sounds like a good test to do since this CPU is seen as "buy this system now and save for a dGPU later" kind of system. Hope you see this, great video nonetheless 👌
Do you have a video to tweak the ram? I have 6000 MHz CL30 and I wouldn’t mind tweaking a bit, I play at 1440p and not all games would benefit but it is fun to try and get some more fps in some games, I have a 7950x and 7900 xtx, I use no upscaling, all native
I bought this APU Ryzen 7 8700G and I'll use just the iGPU Radeon 780M + 32GB of RAM 2x16GB at 7200Mhz because of, the motherboard Asrock A620I Lightning WIFI support (7200mhz maximum frequency) the case is a InWin Chopin MAX :)
still waiting for strix point halo (codename sarlak) apu. dunno if it's mobile only or not but 40cu apu sounds crazy af. i hope it comes to desktop too
I'm interested in this as well as Lunar Lake. It will be interesting to see how they overcome memory bandwidth issues on such powerful iGPUs. I'd assume they almost HAVE to park some GDDR6 right next to it to fully utilize that many CUs. AFAIK these are only intended for laptops so they have a little more leeway in dictating the motherboard setup. Putting these into a desktop socket would absolutely bottleneck the iGPU _hard_ on memory bandwidth.
maybe the same quality But with less ghosting. More performance off course. But even perfomance looks good on 4k I mean dlss perfomance fsr 2 performance i don't know
Yeah I prefer dlss performance most of the time in most demanding games except tlou. For some reason dlss performance implementation on that game seems terrible,and generally unbearable imo.
I expected some new info about how to tweak the memory but if it's the same method from previous video then my latency will stay around 60( Well anyway great video❤
hi Fabio, I know it's off-topic, but I have a RX7800XT paired with a R5 5600X, as I am playing at FHD I noticed I am CPU bottlenecked, especcialy in some games like helldivers 2, do you reccomend me to go for the R7 5800X3D, or should I wait until the Ryzen 9000 series? Side note: I found a nice kit of ddr5 2x16gb at 7800mhz cl36, would it be a great pair with the am5 platform? or should I stick with 6000mhz and tight timings? (the difference in price is 10 euros)
Problem is, I suspect that having a dgpu is still more value for money since high high speed ddr5 is gotta be pretty expensive. And getting amd to work with high-speed ram is not an easy task, closer to rocket science with a bit of lottery depending on which chips are in your ram.
780m is great, but not great enough. It doesn't even have L3 cache like an RX6400 and only has 2MB L2 cache. I don't think the best iGPU will beat the lowest dGPU of same generation.
Obviously. The problem with iGPUs always is, and always will be, that they share memory bandwidth with the CPU. On a 128bit bus. The 780M is RDNA3 and roughly on par with an RX 470 at default LP-DDR5x 6400 spec. The lowest RNDA3 dGPU is the RX 7600 which is about 2.6x faster.
Yup, on AM5 it was never a problem with the IMC itself. Just that at launch the 2:1 mode equivalent to Intel Gear 2 just doesn't work. AMD fixed that a few months ago. The APUs can do 8000+ stable much easier than even Intel.
@@AncientGameplaysI know, i use 6000cl28 and other subtimings maximaly tuned on A-Die. I wanted to see how "that" compares with more clocked mem you done.
EDIT: People asking how frequency/timings would scale on a CPU side, you can watch this video I made some months ago: th-cam.com/video/B5K8Rg-oDwU/w-d-xo.html
Get your V-Color RAM Kits here: v-color.net/
And here we are again! This time testing Several RAM Frequencies with the Ryzen 7 8700G IGPU (Radeon 780M) to see how much performance we can get out of this APU :D
I think new driver update have some performance drop on 7900xtx ?
Btw, could we see how the 780M fares with Elden Ring and Monster Hunter World?
if you can pump water through it i dont put it in my rig lol
I like SARfield. Good game ;D
I'm beginning to wonder how DDR5 at 8600 MT/s will preform with this iGPU, I think that's the fastest yet. But I'm not sure. There may be some 8800 MT/s models out there
Just one note: 6000 CL38 might sound like the same latency as 8000 CL38, but it isn't. The CL is linked to the clock, basically the number of cycles it takes to get access.
These CL values are more or less meaningless for themselves, it would be better to compare the actual latencies in nanoseconds.
For some who might be confused a simple example: 4000 CL40 has the same latency in ns as 8000 CL80
The 6000 CL38 has a CAS latency of 12.67ns, whereas the 8000 CL38 has a CL of 9.5ns
TY❤
also in general primary timings have prove to be rather insignificant as long as trcd&trp are under 13ns and CL under 10.67ns if the secundary timings were properly set up. for example the corsair DDR5 8000CL38-50-50-50-128-178 1.4V is worse than Gskills CL40-48-48-48-128-168 1.35V eventough the Cas latency is loweron the corsair kit.
Im running DDR5 8000CL40-48-48-48-116-116 1.4V and it is better than my previous Corsair 6000CL30-36-36-36-76-112 1.4V altough in HUB's Zen4 scaling video Gskills 6000CL30-38-38-38-96-96 Kit turned out to be slightly better than corsairs CL30 Kit cause Gskills came with better secondary timings.
Bravo Fabio, not even Gamers Nexus or Hardware Unboxed care to dive into APU performance like you have. This video needs to be spread everywhere.
A quality of work that rivals Hardware Unboxed, you are proving the quality of your chanel, and it should be on par with other big youtube chanels, keep up the good work 👍
Thank you for the kind words!
Frankly I think this channel leaves the one mentioned in the dust in some respects.
This is exactly what I wanted youtubers to test out. Desktop allows to achieve full speed of these APUs due to faster memories, but no one was using more than DDR5 6400 and then were disappointed that these end up slower than a budget GPU.
Yeah. The reason is that we don't get past that because of the 1:1 ratio.
I agree visibility of this improvement should be done.
More dosn't make much sense to be honest, when you go over 6400mhz you gonna end up with a 1000$+ system since 8700G and AM5 mini itx mobos are prices insane.
24 is the year, 3 is the month! 😎
hehehehe
1 is the revision....
We meet once again my friend
@@Kavan_Maneuntil it becomes a meme by itself!
Thank you, these are the numbers everyone is interested in but no one but you was willing to put the work in.
Love the memes at the start of your videos. Great info as always Fabio.
Thank you
I've been looking for this exact benchmark and i found it nowhere
Its amazing that you make these really informative videos
Glad I could help!
Nice video. I hope the next APUs will feature some kind of game cache.
thank you bro this is so important for apus and no one else is doing it
I know you know this, but for folks who might not - DDR5 6000, 6400, 7200, 7600 and 8000 is in MegaTransfers per second (MT/s). People often use MHz here, because indeed it is directly tied to frequency, but if you go into BIOS and after applying DDR5 6000 profile you see that your RAM goes "only" to 3000 MHz - Don't Panic! That is how it's supposed to be. Real frequency of DDR is always exactly half of the number of MT/s. Because DDR means "Double Data Rate" - it means that operation (read, write, many other more complex stuff) can happen twice for every "frequency tick". It is important to know that, because timings will often be displayed in such a way (though BIOS-es have differences and they make mistakes as well) that it will be tied to number of "cycles" that are required to wait until next operation can be done. And when you count nanoseconds of latency, you need to know "frequency" not "MT/s". That might be the reason why sometimes calculations are off by 2, 8 or 16 times. You might have forgotten to divide by 2 for frequency or divide/multiply by 8 to go from bytes to bits and vice versa, or both.
Yes indeed, even in this video I state that for some seconds haha
@@AncientGameplays I mean, it's not a big thing, but if someone tries to calculate latencies this comes up the most. Though also some BIOS-es show number of MT/s as "MHz". Pretty much everyone, expert or not make this mistake - I catch myself doing it regularly and in most cases it doesn't matter. Except for people freaking out in CPU-Z 😉or as I said those that calculate latencies.
No my bios says 6000mhz so youre wrong by default
Very interesting, thanks for making these niche videos.
This shows why GPD devices with their 7500MHz ram beat other handhelds with the same APU.
A man of his word. Obrigado, Fabio 👍🏽
thank you! Ive always wanted to make a mini itx gpu-less PC. This helped alot when I bought my RAM
Thank you so much for this incredibly amazing info.
Now i won't sleep searching how to overclock RAM to the max.
Hahahaha, thank you for watching
The best benchmarks and tech reviews! The quality data, the humor, voice and accent just...chefs kiss 😍🤩
This iGpu is monster especially with 8000mhz tweaked. Good job there.
BEEN WAITING FOR THIS!!
Good
Exactly what I searched for that’s worth a subscribe
Thank you for doing the tests everyone should be doing with these iGPU's.
Honestly i would like to have a few hundred euros to spare just to build a PC with a 780M just to see how far i could push it.
I feel you
all those heavy gpus you have gotten recently and been messing/moving around seems to have made you more muscular 💪
the APU physical limits is at 8600-8800mhz with 2200FCLK(cant run stable at 2400 if high mem clock)
Very nice video, love the effort and digging in to it, especially with the tweaked timings
I'd love to see a completely stock 8700g vs highest tuned memory + oc'd 8700g chip as well, to see how much further you could boost performance
if they would allow this amount of tweaking in handhelds or laptops, it would be crazy
Be nice to see this same test using the 7800X3D CPU 🤔
kinda crazy when you think about how much faster and more stable ddr5 is going to get before this platform is dead, we're gonna have 8700gs going BRRRRRRRRRRR
Thank You! 🤩
6400 to 7200 is just a 12% clock increase. i'd say a 6% uplift from just memory clocks seems fair for that amount of extra clock
I think this is a big reason we see less low end ~$200 cards in development because iGPUs will handle that part of the market vs a 1650-4050 now and in the near future
Still for those running older cpu and systems, they still might want that rtx 5050 or rx 7400 or rx 8400
@@UmmerFarooq-wx4yo at what point would these same people begin to be held back by a cpu bottleneck and have to upgrade anyway ? A “5050” would be knocking on a 2080’s door power wise so you are most likely going to bottleneck. Also we’re starting to see these new low tier cards use less and less pcie lanes so these cards would restricted with older cpu/motherboards anyway. At that point the used market is a better solution than a new low end card
@@TopNotchPanch that 3770k gave a 12400k a run for its money.
@@UmmerFarooq-wx4yo I seriously doubt that even if they both used pcie gen 3x16 cards
Excelent video Rma its the most important thing on cpu demanding games! What motherboard did you use?
Now a maxed out oc on air on the gpu with this tweaked memory vs a 1650/super would be pretty fascinating!
It pleases me so much to see how far iGPU's have come. Gets me really excited for the future of handhelds.
indeed
Probably by the end of the year we will see AMD's "Strix Halo" architecture which has a 256 bit (instead of 128 bit) memory interface. It is LP-DDR5x only, so no desktop variant, even a cut down version, is possible. So we should see about double this performance in handhelds soon enough.
Belo video Fábio, quando vai sair comparativo do novo driver 24.3.1 ~ 24.2.xx ?
Em principio amanhã!
10:34 did you say "my fridge was able to run.."? Or what did you say there?
Exactly what I said haha
@@AncientGameplays Wait so are you playing this on a smart fridge with a Ryzen 7 8700G, why would a smart fridge have that? Or was that a joke I didn't get hehe
Of course that was a joke haha@@Meyus_
can you show how to get 8000mhz and how stable it is on your computer?
Basically i just enable xmp and raised ram voltage. Done. 100% stable
100% stable is an extremly huge claim with just enabling xmp when most of the time 8000hz needs tweaks, good voltage, fans on ram, cpu lottery and high end motherboard. What is your testing methodology for stable ram? My bet is that you would get errors in 1 minute in karhu ram testing. Running high speed unstable ram just for benchmarks is not genuine since instabilities might ruin some peoples PCs..
@@sharky9056 my testing is these games tested, TM5 running for 5 hours without a single error and running these games smoothly. Ram voltage 1.4v, soc at 1.3v. Done
Games are not a good test for ram and TM5 ram test is laughable even on the hardest custom presets... only pi and karhu are good test for exposing errors. Reddit post in ram community incoming with your claims.
My friend Karhu sucks (sorry). Used it for several months and had a ram instability in the past due to trusting that same software, the minute I started TM5 with the 1usmus profile I had an issue in the first seconds. Everything I testing with TM5 in the last years, all my systems, are rock stable, so don't come to me telling me that Karhu is better than TM5 and that "games don't stress anything" on the RAM side. I've been there on the same boat as you and I know for a fact that's not true@@sharky9056
good to know, thanks!
Thank you as well for always supporting the channel!
Is there no difference for X3D cpu?
This is iGPU sidd
I see clear difference for iGPU, but I don't know if it's same for X3D @@AncientGameplays
Not fully sure about the 7000 series, but for 5800X3D higher clocked RAM did not make that much difference. Simply because with the huge chunk of local L3 cache, less transfers to RAM were necessary.
And the iGPU in the 7000 series CPUs is more or less a basic display adapter, it's too slow anyways.
@@johnscaramis2515 prob depends on games
@@PokeStoryHarry There are always differences depending on the engine. However if you are playing a game that is depending on RAM speed, it most likely does not benefit from the additional CPU cache (or to a limited extent). CPU frequency could also be a factor.
That's why channels like this one or HUB are important, they check for such differences.
Amazing Benchmarks!
Thanks Fabio! I just bought a laptop with a Ryzen 7840U (which has 780M) on LP-DDR5x 6400Mhz, and was wondering how much I "lost" by not choosing another (much more expensive) vendor for a similar product with 7500MHz. This video gave me the overview I has been looking for.
You can't tweak memory in most of the laptops available just a few of them support memory tweak.
@@c4nchi I'm perfectly aware. But I was still kinda curious about how much I was leaving on the table by going with 6400 instead of 7500.
@@c4nchi I thought you can do it in all of them through smokeless.
10:35 I'd never thought to play League on my fridge.
😂😂🤌
you have a video showing how to tweak the timing?
I do
Honestly for Integrated every single FPS matters when your performance is already pretty poor. For example 48 vs 58 fps in Spider Man is a huge difference going from laggy to almost smooth gameplay. Now obviously buying hella expensive 8000mhz kits wouldn't make sense but there is a GSKill Neo 6400/CL32 which is Hynix A-die and it can be manually pushed out to 8000. (My Corsair Dominator Titanium 6400/cl32 is A-die though but more expensive than the G.Skill kit and it's also way easier to find A-die Gskill kits as the part number on the back of the ram stick ends with the chip type.)
Great video though.
Fighting the urge to get one of these just for memOC shenanigans
haha, these are indeed great
The 7500f has been around for some time, should be the same die.
@@heinzletzte.6385 Memory OC is significantly different despite both being Zen 4.
@@atirta777 whats the difference? binning?
@@heinzletzte.6385 Even if the architecture is Zen 4 on both, the whole die design is different. Ryzen 7000 is chiplet, while 8000 is monolithic which means the mem controller and infinity fabric also work very differently. 8000 series design is also meant to support faster lpddr5 for low power devices so that could also help stability at higher frequencies. People are pushing these new chips up to 10000MT/s, while Ryzen 7000 is still stuck at 8000MT/s
@AncientGameplays, Fabio, I'm sure you have this already lined up but, just in case you don't yet, can you please do an episode on the new drivers - 24.3.1? In particular I'm sure many will be interested in the supposed unlock of overclocking for the memory on the 7900 GRE. I've found this to be disappointing personally, since I can only reach around 2376, about a 60 Hz increase on the memory clock. Beyond that they don't seem to be stable at all. I'm wondering if AMD will release updated drivers in the future that address this? Anyway, just a suggestion. Keep up the excellent work you're doing and thank you.
Will get it released tomorrow 💪💪
@@AncientGameplays As we say in the UK, you are a scholar and a gentleman Sir. Much appreciated. 😊
As a side note here, I help with a medium sized channel, about 27, 000 subscribers, which is done by my lovely wife and it would be great, when you have some time, if you could come on her show and chat a bit about your enjoyment of computers and technology. I think you'd be a great guest. Valeu por tudo, Fabio.
not even 5 minutes watched and.......
ABSOLUTORIUM !!! 👍👍😎😎
haha, thanks!
When are you going to test new drivers 24.3.1 ?
Today will release the video
Hey brother, I just wanted to let you know, that you can manually get your FCLK and UCLK up to 2400mhz with this chip without decoupling the frequencies. I was able to get mine to 2400 FCLK for 7200mhz, and 2467 for 7400mhz. I haven't seen anyone try this on TH-cam. I was able to do this on a B650m MSI mortar wifi nothing special. I used 1.4v for ram. I use this chip with my 4070, and haven't tried the APU yet but without being decoupled it should give even more performance. If you have time it would be cool to see if there is any benefit to raising your FCLK that high. Love the channel!!!
Both fclk and uclk were at 2200MHz 💪
Yes Hynix chips OCd and tuned at 7200-7400 is the optimal for 8000G APUs, not 6400, not even 8000, cause many games like the lower latency, so generally 7200 tuned is the best for these APUs :)
@@chovekbcan you post some more info on the 7200 setup? What’s your build and settings? Stable?
Please make a DDR5 8000 tweaked + 8700G vs other budget GPUs like 6500XT, 3050 6gb , A380.
It would only make sense to compare it to pcie powered GPUs only imo.
You can see this one: th-cam.com/video/m1GBo8K2MY0/w-d-xo.html
@@AncientGameplays true but in this video you probably didn't use tweaked memory, did you ?
@@DragonOfTheMortalKombat absolutely sir. Fabio did not include what the ram configuration was(that I remember?) for that video. You could open both videos side by side and compare results on game to game and see the % difference and assume that vs the rx 580 etc and give you an idea. Report back what ram configuration you think that the "vs rx 580" video was using. It'll be a lil game for you. 😜
@@christophermullins7163 before making a fool of yourself, did you go and watch the video ? In the RX 580 vs 780M ddr5 6000 cl 36 was used. I am talking about ddr5 8000 with manually tuned memory timings. I never said that he didn't mention what memory was used. Lmao, you think you I wouldn't bother verifying kid 🤪
Be careful, the "Tweaked" versions are not the same for everyone, if you buy a cheaper kit at lower frequencies you won't be able to tweak it as much. Still stands that you'll match or be better than more expensive without tweaking but something to take in mind.
any hynix kit can do these timings. Just saying
It's so funny to me how Fabio makes videos on FSR 3.1 and the new iGPUs and AMD is just like "oh yeah let's drop a new driver update" LOL
I know right? Let me sleep haha
@@AncientGameplays Exactly haha
Does higer ram speed increase the APU's temperature?
Not really, not unless the FPS difference is so big that the CPU needs to work way more and temps would raise, but generally no difference
I'm looking at DDR 7200, CL34 for my Ryzen Zen 5 build. I won't use IGP, but 7200 DDR 5 gives you a nice mix between high bandwidth and latency at CL34.
You'll have to decouple the UCLK which isn't really that great in terms of CPU performance only. Just get a 6400MHz kit and tweak it
It's well documented that decoupling your CPU clock and memory clock leads to _decreased_ performance, not improved performance. It may be beneficial in this niche case that's memory starved, but will hurt overall performance everywhere else. Memory speed matters, but it's also important to be in sync with the CPU's memory controller, and we still have no idea if Zen5 will come with an improved memory controller that doesn't slam into a brick wall at 6000mt/s.
I know your content is mostly focused on gaming, but can you test if the IGPU helps improve performance in productivity applications like premiere pro, after effects or davinci resolve? and how it compares against an 7700X?
I know the Intels APUs have an impact on programs that decode video, but I found anything about the AMD APUs doing the same.
Can try
Can you link me the video for tweaking the RAM please? 😅
th-cam.com/video/yDiy20cT1M8/w-d-xo.html
@@AncientGameplays Awesome! Thank you!
This is incredible.
I have the Ryzen 7 5700g CPU at 1.32V 4.5Ghz all cores i think :D paired with 6700XT tho.
And actually its Good and very effectively with low voltage in big loads :)
so the Ryzen 7 8700g must be a Beast.
Can the "tweaking" be done on 7840 laptops?
if the bios lets you change timings yes
I’m looking at the ram kit you said to get and there are a lot of reviews of people saying that are unstable
Nah, that was in the very early days and from people that know absolutely nothing about what they're doing. Happens a lot sadly
Do you have a guide for tweaking timing on RAM by chance?
Can you pls do a Ram OC Tutorial?
watch the video mate :D
@@AncientGameplays oh, sry mate. I just skipped to the results of the different games as that is the most important part for me 😅
Great job
Fabio is back from the dark side and he brought cats😎
😂😂
Thank you. Great video.
I thought 5600 was standard DDR5 speed for AM5. I imagine uplift from that is even better, you get like 20-25% better result.
I have only miniPC with 7840HS that can not be really tweaked (BIOS version shows 0.01 ?!! )
Noone should be getting 5600MHz nowadays I believe
@@AncientGameplays Too late 🙈
@@DS-pk4eh no, 6000 has been the standard since launch, I don't think 5600 was ever the recommended speed.
Is it possible to see such an uplift, or at least close to it, with a Ryzen 7600 or not? I have 32 gb DDR5 6400mhz. Thank you!
Nah, this is only iGPU sided, watch this one though:
th-cam.com/video/B5K8Rg-oDwU/w-d-xo.html
@@AncientGameplays I have already seen it but thank you very much!
Hello, what settings was used for 8000MHz tweaked? Hardware used was AM4/DDR4 or AM5/DDR5? Thanks
:| i have a 10100f+gt1030+8gb ddr4 2999...
So i ll probably wait for the upcoming apu, which to have strix point igpu.
😅 I am not a laptop guy so, I'm not talking about the 8850hs.
I'm talking about the 9700g (speculation)
:| i know it's going to take around 2 years for the 9700g to hit the market.
Pray for the upcoming glory 😑🙏🐣
(780m is not so good)
Also: rdna 3.5 is actually a refresh of rdna 3. We may be able to see a refreshed processors like 8750g (😑🙏 praying for it to come out asap.)
😑🙏 Praying for the upcoming strix halo as well (which has 40cu igpu)
Please be cheaper (and be a desktop gpu... Not the laptop) 😑🙏 guys pray.
Dragon Dogma 2 in 4 hours, so people are going to be really testing that game out to see what exactly really messes up the CPU usage...
But I think their game engine is the issue, basically can't handle the heavy workload they ambitiously put on it.
Hello , excellent video , the reason I went with 8700G and 6000mhz , 30CL Expo from Kingston.I set up only Expo profile , I am satisfied with results but I would like to optimize the timings as you did.I am a bit of noob for this but I cannot find how to set/optimize DRAM timings and all other configuration? Is there a separate video for this?Thanks in advance :)
Why did you do that ? Why choose unrealistic timings ? Nobody runs 6000@cl38. Why not start with conventional timings, 5600cl36 / 6000cl30 / 6400cl36 ?
I explained that in the beginning, but when you skip parts you come to the comment section clueless...
Upgrading from a rx 6600 should i get a rx 7900 gre or a rtx 4070 super? what do you recommend please!
Same price?
@@AncientGameplays 100€ price gap
@@woIfmann 4070s more expensive? What about models?
@@AncientGameplays yes! 4070 super is more expensive; the model is Msi rtx 4070 super gaming x, as for the Rx 7900 gre its the xfx rx 7900 gre qick 319 bulk edition
I would get nothing more than the GRE Pulse/Pure or Nitro+ if I were you, or that Gaming X then@@woIfmann
Hi, just a thought. What if you ran the same exact test, but add in a dedicated GPU. Would the higher freq tweaked memory scale the same? At 1080p of course to test the CPU limit. Sounds like a good test to do since this CPU is seen as "buy this system now and save for a dGPU later" kind of system. Hope you see this, great video nonetheless 👌
Do you have a video to tweak the ram? I have 6000 MHz CL30 and I wouldn’t mind tweaking a bit, I play at 1440p and not all games would benefit but it is fun to try and get some more fps in some games, I have a 7950x and 7900 xtx, I use no upscaling, all native
It is shown in this video :D
crazy numbers
I bought this APU Ryzen 7 8700G and I'll use just the iGPU Radeon 780M + 32GB of RAM 2x16GB at 7200Mhz because of, the motherboard Asrock A620I Lightning WIFI support (7200mhz maximum frequency) the case is a InWin Chopin MAX :)
Nice
How do you exactly tweak the ram
Watch buildzoid's ddr5 video
0:30 Linus is Proud man 😂😂💯🤝
hahaha
When did Ryzen get past 6000MT?
Months ago, with Bios updates
Is that ps5 xbox level of performance with 8000mhz?
Nah, not even close
Of all the Fabios I know, you look the most Fabio.
Hahahaha
how you get 8000??? can you post settings?
With an 8000mhz kit as shown. Default settings
Can you film a video about micro stuttering? High fps but not smooth etc.? For AMD
I talk about that a bit in my 20 tips video
Where can i see the guide for tweak the RAM?
Ryzen Dram calculator and TH-cam Actually hardware overclocking
still waiting for strix point halo (codename sarlak) apu. dunno if it's mobile only or not but 40cu apu sounds crazy af. i hope it comes to desktop too
I'm interested in this as well as Lunar Lake. It will be interesting to see how they overcome memory bandwidth issues on such powerful iGPUs. I'd assume they almost HAVE to park some GDDR6 right next to it to fully utilize that many CUs. AFAIK these are only intended for laptops so they have a little more leeway in dictating the motherboard setup. Putting these into a desktop socket would absolutely bottleneck the iGPU _hard_ on memory bandwidth.
Is dlss 2 balance better than fsr 2 quality at 4k?
maybe the same quality
But with less ghosting.
More performance off course.
But even perfomance looks good on 4k
I mean dlss perfomance
fsr 2 performance i don't know
Yeah I prefer dlss performance most of the time in most demanding games except tlou. For some reason dlss performance implementation on that game seems terrible,and generally unbearable imo.
@@clem9808 yes, grass looks bad.
So 32GB 7200 C34 for 140€ was pretty good value by the looks of it. Haven't tried to OC it.
Indeed it was
I expected some new info about how to tweak the memory but if it's the same method from previous video then my latency will stay around 60(
Well anyway great video❤
Yeah, tweaking is basically the same, but higher frequency or more agressive timings
hi Fabio, I know it's off-topic, but I have a RX7800XT paired with a R5 5600X, as I am playing at FHD I noticed I am CPU bottlenecked, especcialy in some games like helldivers 2, do you reccomend me to go for the R7 5800X3D, or should I wait until the Ryzen 9000 series? Side note: I found a nice kit of ddr5 2x16gb at 7800mhz cl36, would it be a great pair with the am5 platform? or should I stick with 6000mhz and tight timings? (the difference in price is 10 euros)
You should get a 1440p monitor
@@torradasuprema395 i don't wanna play 2k 100fps, i prefer fhd 150fps
Problem is, I suspect that having a dgpu is still more value for money since high high speed ddr5 is gotta be pretty expensive. And getting amd to work with high-speed ram is not an easy task, closer to rocket science with a bit of lottery depending on which chips are in your ram.
Is it better to have more GBs or more MHzs?
In term of iGPU? MHz
6:53 i love sarfield 👍funny typo sounds like garfield
Curious if it's cheaper to build a 8600g build or 7500f with maybe a 5700 xt?
PSU problem and 5700xt is old
780m is great, but not great enough. It doesn't even have L3 cache like an RX6400 and only has 2MB L2 cache.
I don't think the best iGPU will beat the lowest dGPU of same generation.
true, I do believe though that in some scenarios the 780M equals the 6400
Obviously. The problem with iGPUs always is, and always will be, that they share memory bandwidth with the CPU. On a 128bit bus. The 780M is RDNA3 and roughly on par with an RX 470 at default LP-DDR5x 6400 spec. The lowest RNDA3 dGPU is the RX 7600 which is about 2.6x faster.
AM5 IMC is scaling now?? thats great news!
Yup, on AM5 it was never a problem with the IMC itself. Just that at launch the 2:1 mode equivalent to Intel Gear 2 just doesn't work. AMD fixed that a few months ago. The APUs can do 8000+ stable much easier than even Intel.
Can we just lock CPU frequency to save power for GPU?
You can, but you can simply raise the power limits too
iGPUs are always bottlenecked by RAM speed
Not really!
I won one of these from Linus Tech Tips! To bad I don't need integrated graphics and I'm on AM4 lol. Going to sell it and upgrade my graphics card
You can play CS2 144HZ with a IGPU(6400MHz).This is insane
Indeed
I want to see 6000 cl30 or cl28
Subtimings are what makes the biggest difference
@@AncientGameplaysI know, i use 6000cl28 and other subtimings maximaly tuned on A-Die. I wanted to see how "that" compares with more clocked mem you done.
Show these tweaked results against the best budget build at same cost
I did a thing close to that versus the rx 580
was that one with the 2 tweaked ram configs shown here? @@AncientGameplays