Why I will probably buy 7700X instead of 5800X3D or even 7800X3D [please tell why shouldn't I]

Coolio

Weaksauce
Joined
Jan 8, 2021
Messages
118
Hi guys,

after spending hours on reviews, feedbacks, etc. I'm very close to thinking 7700X is probably my best choice. I will substantiate that below, and please share your thoughts - will highly appreciate that!

Goal: build a new PC from scratch (i.e., there are no previous configs that tie my hands), so that it is compatible with at least 1 significant upgrade (mainly CPU/GPU) within the next 5 years.
Use scenario:
  • office use
  • gaming @ 1440p/144Hz (it seems to be a sweet spot today), or 4K when possible. The future upgrade of CPU/GPU should improve my 4K possibilities (otherwise upgrade makes no sense), but I will stay within a 144Hz/FPS paradigm (jumping to 244Hz monitor makes no sense in my use case, so I will better upgrade to 360Hz in 8-10 years - but that will be a totally new build)
    • I will play mainly openworlds/platform/RPG (incl. some AAAs). Never tried MMORPG, but not sure there's extra value for me in it over a RPG. Definitely not racing or sports
    • not an FPS-slut, so I will prefer higher resolution/detalization with reasonably (!) lower FPS over extremely high FPS @ 1080p
  • occasional audio/video coding/converting

These are my arguments:
  1. Since I want to be able to significantly upgrade at least once within the next 5 years, platform-wise AM5 wins over AM4, as there won't be any new CPUs compatible with AM4 socket anymore.
  2. With the said future upgrade in mind, Zen3 CPUs have no chance to be sold for more or less decent money in 5 years, while used Zen4 will stay attractive when Zen5 is on the market.
  3. x600 CPUs (or x700 if multitasking matters) are the "best bang for the buck" solution, since their performance can be significantly boosted by an appropriate GPU - specifically on 1440p/4K resolutions which are less CPU constraint, but more GPU constraint. Buying the x9x0 series won't eliminate the "GPU bottleneck" problem, but these CPUs cost more - this budget will work much more efficiently when invested in a GPU instead. They say that at 1440p+ resolutions e.g. 5600X & 4080 will outperform 5800X3D & 3080.
  4. What about 5800X3D? [I'll ignore the AM4 "aging" problem and will talk pure tech here]:
    1. When compared within Zen3 family, it delivers much more FPS vs. 5600X/5700X, but only @ 1080p (1440p/4K are GPU constraint, as mentioned earlier, so 5800X3D will be no different from 5600X/5700X on these resolutions)
    2. When compared with Zen4 family CPUs, 5800X3D without top GPU has parity with 7600X/7700X @ 1440p/4K, so again - even gaming CPU needs a good GPU to outstand at high resolutions and AAA games. Given that both 5800X3D and 7700X cost $360 (hereinafter Jan'23 prices in EU), and that I will need a decent GPU for either of them - no sense to choose aging AM4 for the same money. As for 1080p, 7700X brings 92.4% of 5800X3D's FPS rate, so even if I [hypothetically!] don't buy a top GPU, I will sacrifice only 7.6% of FPS while staying on a newer platform. 7600X brings pretty similar 91.3% of 5800X3D's FPS rate, but it costs $250, so wasn't multitasking part of my use scenario I would have invested those $110 (vs. 5800X3D) in a GPU. However, in my case I'm paying them for CPU to gain +27% (7700X vs. 7600X) in multi-threading performance.
  5. What about 7800X3D (or whatever it will be called)? Firstly, no doubt it will be an expensive thing (some say around $900???). Secondly, following the 5800X3D logic, the GPU will still be a game changer at higher resolutions (GPUs are part of AMD's business too, right, so it won't be smart to cannibalize the friendly portfolio). I assume this future CPU's "GPU bottleneck" will definitely rise from 1080p to 1440p, but 4K bottleneck will stay in place. The clockrate will probably be lowered not only for tech reasons (as it was done in 5800X3D), but also to avoid intra-Zen4 competition. Finally, Zen4 is DDR5-based, which is way faster than DDR4 - and that somehow reduces the need for L3 cache (which is the 5800X3D/7800X3D's key strength). To summarize, 7800X3D will definitely make me happy at 1440p and partially at 4K, but it won't eliminate the necessity of extra expenses on top-rated GPU if I want to enjoy 4K to the full extent. Budget-wise it will be $900/CPU + .... [what? at least $700/GPU, right?]

So, my vision of the potential config for my use case is the following: 7700X ($360) + NVidia GeForce RTX 3080 ($700) = $1060. I've heard the opinion that should I have chosen 7600X, RTX 3070Ti would have been more appropriate, so your comments on the balance between the CPU/GPU of my choice will be also appreciated.

Thank you and have a great 2023! :)
 
You might want to see what CES brings in a few days and if any dates, maybe pricing for X3D models. Your the best person to decide what will work for you, your budget etc. Pretty much agree, unless upgrading a current AM4 system to keep the costs low if budget constrain with a 5800 X3D, it makes better sense if building a new complete system to be AM5.
 
Your the best person to decide what will work for you, your budget etc.
True, but it's not a bad idea anyway to listen to those who possibly know more than me. :) For this post to not look noob and boring I shared my arguments.

CES or not - gaming CPUs won't be cheap, specifically after AMD saw 5800X3D was damn popular and hence - dangerous for x9xx series. Play the L3 cache card and rise the price (at least while DDR5 prices are high), says my long-term marketing background. :)
 
It doesn't hurt to be on the latest platform/socket and there's not much difference monetarily between them, with the older tech phasing out and getting more expensive it doesn't make sense to invest more into a dead socket. DDR5 can be found fairly cheap if you cut out the bling / aesthetics especially.

I've got a 7700X and find it to be quite the performer on a B650E chipset with 6000 Mhz CL36, I believe it to be good investment and should only get better. I am also SO happy that AMD went LGA.
 
DDR5 can be found fairly cheap if you cut out the bling / aesthetics especially.
Thank you for your comment! I totally'd love to avoid aesthetics, in particular RGB blows'n'whistles.
  1. Wanted to go for one of EXPO kits to boost DDR5-5200 to 6400 - does it make sense? Any drawbacks I should be aware of (e.g., instability, etc.)?
  2. Do you have experience with 7000 series ECO mode [105 -> 65 TDP] - how much does it decrease power consumption in office use scenario, is the productivity cut noticeable?

I've got a 7700X and find it to be quite the performer on a B650E chipset with 6000 Mhz CL36, I believe it to be good investment and should only get better.
Better - how? I mean what is your upgrade plan so far and maybe you have comment on my GPU choice as well?
 
I've got a 7700X and find it to be quite the performer on a B650E chipset with 6000 Mhz CL36, I believe it to be good investment and should only get better. I am also SO happy that AMD went LGA.
i just purchased an identical setup this week and have been very happy with my upgrade.
 
If gaming only I would wait for the non x chips to launch. Might put pressure on prices downward. If not the 7600x makes more sense than the 7700
 
Finally, Zen4 is DDR5-based, which is way faster than DDR4
I agree with your choice. That said DDR5 is not WAY faster. It can be marginally faster if you are willing to pay for the upper echelon speeds, and even that depends on your game/application. DDR5 does have higher rated speeds but it also has much higher latencies. If you are considering bottom barrel DDR5 then you are definitely not entering the jet age when compared to top rated and cheaper DDR4.
 
If gaming only I would wait for the non x chips to launch. Might put pressure on prices downward. If not the 7600x makes more sense than the 7700
I am considering non-x CPUs in this thread - 7600X and 7700X and my choice was 7700X, because my scenario is not gaming only (which I mentioned).

DDR5 does have higher rated speeds but it also has much higher latencies. If you are considering bottom barrel DDR5 then you are definitely not entering the jet age when compared to top rated and cheaper DDR4.
Good info, thank you! Am I right that overclocking (e.g., DDR5-5200 to 6400) won't help much, and latency will stay a bottleneck, compared to low-latency DDR4? Do you think makes sense to buy now best possible latency, but smaller volume (e.g., 16Gb), wait for prices to go down and buy another 16Gb in e.g. 1 year?
 
I am considering non-x CPUs in this thread - 7600X and 7700X and my choice was 7700X, because my scenario is not gaming only (which I mentioned).


Good info, thank you! Am I right that overclocking (e.g., DDR5-5200 to 6400) won't help much, and latency will stay a bottleneck, compared to low-latency DDR4? Do you think makes sense to buy now best possible latency, but smaller volume (e.g., 16Gb), wait for prices to go down and buy another 16Gb in e.g. 1 year?
If you have the budget, then just go for 7700x, it won't disappoint in regards to gaming + productivity.
Or if you can wait, perhaps the 7700 non-x can be considered as it's of lower TDP and may require less cooling capacity + probably has improvement (next stepping batch) compared to the X variant.

if you leave nearby MC or can afford to order online, perhaps the MC deal for free ram still exists?
if it's still exists, then just go for the free ram bundle (it's 32gb kit with XMP/EXPO 6000mhz c40 if I'm not wrong).

If the deal already expired, then just go for 32gb kit with 6000mhz c40, it will probably contains Hynix M-Die chip or Samsung B-Die which supported by Ryzen 7000.
The M-Die one can be overclocked higher past 7200mhz in Raptor Lake + Z790.
 
Or if you can wait, perhaps the 7700 non-x can be considered as it's of lower TDP and may require less cooling capacity + probably has improvement (next stepping batch) compared to the X variant.
Dammit - my bad! I've read Vengance_01's message as "non-gaming", i.e. non x3D. :( Good point regarding non-x - noted! Shall lower TDP mean lower productivity too? I mean what exactly will I sacrifice when hitting this road?

Hynix M-Die chip or Samsung B-Die which supported by Ryzen 7000.
The M-Die one can be overclocked higher past 7200mhz in Raptor Lake
  1. Can M-Die be overclocked higher than 7200 in Ryzen 7000 too (with relevant MoBo of course)?
  2. To which extent does overclocking eliminate the DDR5's high latency problem (if it does at all)? If I have a choice between e.g. DDR5-5200 cl32 and DDR5-6000 cl40 (both overclockable) - which is better and why?
Thank you!
 
... Do you think makes sense to buy now best possible latency, but smaller volume (e.g., 16Gb), wait for prices to go down and buy another 16Gb in e.g. 1 year?
It appears that running 4 sticks of DDR5 is very problematic for current memory controllers, so I'd say no. If you were to do anything along those lines, it would be buy 2 new sticks and sell the old, probably for a loss.
 
I am considering non-x CPUs in this thread - 7600X and 7700X and my choice was 7700X, because my scenario is not gaming only (which I mentioned).


Good info, thank you! Am I right that overclocking (e.g., DDR5-5200 to 6400) won't help much, and latency will stay a bottleneck, compared to low-latency DDR4? Do you think makes sense to buy now best possible latency, but smaller volume (e.g., 16Gb), wait for prices to go down and buy another 16Gb in e.g. 1 year?

Just to clarify, relatively cheap B-die DDR4 can get you ~4000mhz with a total latency of around 45ns. DDR5 has to make it to 7400+ CL32 to beat that and AM5 isn't capable of it in the first place so don't worry about it.
 
I would wait for CES announcements. There will likely be some sales. And, there are rumors that AMD will announced a 7600X3D. if you are mostlyi concerend with gaming, that will be the CPU to get. It would probably be about $350 and would give top level gaming performance, for hundreds less. The 7600X is already a fantastic gaming value, at about $240. Its not much worse than the 7700x, in gaming.
Just to clarify, relatively cheap B-die DDR4 can get you ~4000mhz with a total latency of around 45ns. DDR5 has to make it to 7400+ CL32 to beat that and AM5 isn't capable of it in the first place so don't worry about it.
Depends on the game. Some games, like COD Warzone, are latency sensitive. And do as good or slightly better, with good DDR4. Other games like the extra bandwidth and raw speed of DDR5. Spider-Man, Hitman 3, etc. Averaged out, DDR5 seems to be overall a little better.
 
Dammit - my bad! I've read Vengance_01's message as "non-gaming", i.e. non x3D. :( Good point regarding non-x - noted! Shall lower TDP mean lower productivity too? I mean what exactly will I sacrifice when hitting this road?


  1. Can M-Die be overclocked higher than 7200 in Ryzen 7000 too (with relevant MoBo of course)?
  2. To which extent does overclocking eliminate the DDR5's high latency problem (if it does at all)? If I have a choice between e.g. DDR5-5200 cl32 and DDR5-6000 cl40 (both overclockable) - which is better and why?
Thank you!
1. The 65w TDP of 7600/7700 is default TDP though, and with B650 (non E) board, you still gain 16x pcie 4.0 for graphics and 4x pcie 5.0 for nvme, and yet still can overclock the CPU and the ram, so might be good cheapish setup (7700 non-x + b650 chipset).
That lower productivity maybe around 10% only yet it probably could reach the 7700x level with mild overclock.

2. Right now, the limitations are AGESA & average imc of ryzen 7000 (only handful can run ram to 6400).

3. Take the 6000 c40 and then try to tune it ot 6400 30-38-38-30 1.5v or 6400 32-39-39-32 1.45v.
 
Whichever you get (7700X, 7900X, 7950X), play around with the PBO curve and PPT settings.

For the 7900X build in my sig under full MT load (Cinebench R23) :
Stock : ~185W package power, 93C
Tweaked (-30 curve, 135W PPT) : ~138W package power, ~74C, ~2% higher ST and MT performance

Even going with just -18 curve instead (for less prime silicon) achieved ~1% lower ST and MT performance
Just make sure to validate your tweaks with CoreCycler and plain old idling (eg. overnight)

Ryzen 7000 is incredibly efficient, except AMD decided to nuke it by chasing the last few performance %
My experience mirrors that of OptimumTech's :
 
Last edited:
Just to clarify, relatively cheap B-die DDR4 can get you ~4000mhz with a total latency of around 45ns. DDR5 has to make it to 7400+ CL32 to beat that and AM5 isn't capable of it in the first place so don't worry about it.
I still don't understand why DDR4-4000 is faster than DDR5-7400. How huge should DDR5's latency be for its clocks to be almost doubled to break even with DDR4 in terms of results. :eek: Or am I misunderstanding something here?


The 65w TDP of 7600/7700 is default TDP though, and with B650 (non E) board, you still gain 16x pcie 4.0 for graphics and 4x pcie 5.0 for nvme, and yet still can overclock the CPU and the ram, so might be good cheapish setup (7700 non-x + b650 chipset).
That lower productivity maybe around 10% only yet it probably could reach the 7700x level with mild overclock.
What do you say about the statement (wide-spread, btw), that "all CPUs these days are factory-overclocked to the max"? So while 7600/7700's 65W TDP is definitely attractive, are there real opportunities to o/c them from 3.8GHz to 5.1-5.3GHz without risks? I've also heard rumors that non-X models are based on lower quality silicon...

They even recommend to slightly underclock/undervolt factory values if build is to be silent. This is the reason why I asked if anyone had experience with Eco mode in 7000 series - I'd probably use it for daily office tasks not to bother with manual underclocking (if only it is much more efficient vs. Eco mode).


Whichever you get (7700X, 7900X, 7950X), play around with the PBO curve and PPT settings.
Saved, thank you!
 
I still don't understand why DDR4-4000 is faster than DDR5-7400. How huge should DDR5's latency be for its clocks to be almost doubled to break even with DDR4 in terms of results. :eek: Or am I misunderstanding something here?

Its not in theory, but in reality the memory controllers aren't fast enough yet for chips that use DDR5 and the board topologies aren't as optimized as stuff thats been out for years. Basically the problem isn't the RAM, just look at the leap from the Intel 12th gen to 13th gen as far as DDR5 speed and stability goes. According to JEDEC standards the single-read latency between the various types of DDR have been around the same since DDR3 was dominant, but the round-trip read/write cycle really depends on the whole platform to keep up.

I've been eyeballing the 7900X myself because I can't resist new things but I would probably be better served by swapping my 12700k for a 13900k to feed my 4090.
 
Its not in theory, but in reality the memory controllers aren't fast enough yet for chips that use DDR5 and the board topologies aren't as optimized as stuff thats been out for years.
Got it, thank you! So what will be your recommendation on top clocks & latency as of today - to not spend money for what won't work anyways? Here is what was already recommended in this thread:

Take the 6000 c40 and then try to tune it ot 6400 30-38-38-30 1.5v or 6400 32-39-39-32 1.45v.
Is it the fastest what I can have with DDR5 as of today? Without compromising its stability after overclocking of course. And also makes no sense to spend e.g. $100 more on the MoBo which probably supports higher RAM clocks (e.g. 7000), but that won't give me significant step forward in gaming experience (because of e.g. board topology limits you've mentioned).

Thank you! Just trying to understand where the reasonable "ceiling" is in terms of DDR5 today.
 
Got it, thank you! So what will be your recommendation on top clocks & latency as of today - to not spend money for what won't work anyways? Here is what was already recommended in this thread:

My suggestion is to get a middle of the road kit at a good price. DDR5 top end kits are $500+ and it makes no sense to buy. I would buy one that is economical and change it out in a few years with the CPU if you're going to do that since the controller will be better and higher speed modules will be cheaper. AM5 maxes out about 6000 CL30 if you're willing to fight it for a long time, so a 5600-6000 CL40 kit is probably where your target should be depending on if you want 32GB or 64GB.
 
and change it out in a few years with the CPU if you're going to do that since the controller will be better and higher speed modules will be cheaper
Thank you so much, noted! Shall the MoBo that I will choose for my 2023 build somehow consider the future RAM upgrade? Don't want it to become a bottleneck with faster DDR5 that I will buy in e.g. 2025. Which specs should I pay attention to in terms of specifically RAM compatibility?
 
Thank you so much, noted! Shall the MoBo that I will choose for my 2023 build somehow consider the future RAM upgrade? Don't want it to become a bottleneck with faster DDR5 that I will buy in e.g. 2025. Which specs should I pay attention to in terms of specifically RAM compatibility?

I wouldn't worry about it, that stuff improves with bios revisions and the controller is on the CPU. Any non-bottom barrel board should be perfect.
 
Thank you so much, noted! Shall the MoBo that I will choose for my 2023 build somehow consider the future RAM upgrade? Don't want it to become a bottleneck with faster DDR5 that I will buy in e.g. 2025. Which specs should I pay attention to in terms of specifically RAM compatibility?
The biggest problem you will have with AM5 is board quality and feature sets for the future. You have to pay more attention to the Series of Motherboard and it's listed features than anything else. Gone are the days of simple purchases for AMD. The B650 and B650E and the X670 and X670E and there are a number of feature set nuances between each model. With some E models being quite a bit more spartan than others.

My experience with AM5 has not been a stellar one however, you're looking for a couple things for good stability.
If a PCI-E 5.0 Graphics card is not in your future (as there are none now) you probably don't need the E series of either the B650 or X670. You will want to focus on good power phase distribution for the CPU for long term compatibility and the RAM should be fine on all boards.... Except anything fast.... SO, if you want RAM faster than DDR5-6000 you might consider paying more for a higher end board. RAM compatibility will NOT be the same across all boards regardless of what the series of chipset claims to support if the RAM reaches into an extreme overclocking area. If you ever intend to run DDR5-6400+ I would consider buying a more robust Mainboard.

What do I mean by robust? In AM5? Not Sub 300 Dollar, more like 500+ Dollar Boards. Which will firmly place you in the shit show that is the E Series Feature set. Where you need to pick and choose what you want based on the board partner offerings.

M.2 Slots are all over the place in this generation. With the vast majority of them having at least one PCI-E 5.0 4x slot (and only one in some cases) and a mixture of PCI-E 4.0 and 3.0 ones on most boards.

There is no such thing as future proof. There is only buying what you think you will need to support what you hope the industry tech will provide. There are no guarantees how long AMD will support their new platform. But it's likely going to have support until the end of 2024.
 
SO, if you want RAM faster than DDR5-6000 you might consider paying more for a higher end board. RAM compatibility will NOT be the same across all boards regardless of what the series of chipset claims to support if the RAM reaches into an extreme overclocking area. If you ever intend to run DDR5-6400+ I would consider buying a more robust Mainboard.

What do I mean by robust? In AM5? Not Sub 300 Dollar, more like 500+ Dollar Boards.
Good input - thank you! Do I understand right, that by "RAM faster than DDR5-6000" you as well mean 6000 overclocked to 6400? Or only factory clocks count in this riddle? 'Cause if I have to buy $500 MoBo if I overclock my 6000 vs. $300 MoBo with factory clocks - I won't overclock maybe. :)
 
Good input - thank you! Do I understand right, that by "RAM faster than DDR5-6000" you as well mean 6000 overclocked to 6400? Or only factory clocks count in this riddle? 'Cause if I have to buy $500 MoBo if I overclock my 6000 vs. $300 MoBo with factory clocks - I won't overclock maybe. :)
What I mean is that the initial spec for the AM5 platform was around DDR5-4800, that's the core default that the two AM5 Boards I worked with started the RAM at before Memory Training.

AMD Recommends that the sweet spot for the AM5 Platform is DDR5-6000 and that most boards and CPUs can get to that level.

Anything past that is gonna require more juice and tinkering and fabric clock tweaking and possibly a higher quality motherboard to get the faster ram to run stable.

I think, if you're going in basic you can pickup a Gigabyte Aorus AX X670 for around 275 bucks (US) and it should be a fairly robust board for an early adopter. It lacks the PCI-E 5.0 x16 slot, but it's a nice board for the price with a number of other decent features. And it puts you on the higher end chipset.

If you're not accustomed to the AMD Memory Training thing, this will be a rude awakening for you. Because the entire process may be painless (to some) but it's clunkier and slow as all hell to accomplish on these new boards (that's getting your RAM to Run at it's Rated 6,000 Speed).

You're gonna be spending around 300 bucks just to get on the platform and have some longevity. I do not recommend buying into the platform with a sub 200 dollar motherboard, it will be a cheap piece of shit and may not last.
 
FYI you can buy into the AM4 Platform in the sub 200 dollar price point and be just fine. I have 3 Motherboards I paid less than 150 bucks for that have been stable, allow overclocking and have been running like champs for years new. That's the difference between a platform with most of the kinks worked out vs a new one.

Best value, hands down for a Gamer build with some longevity is a 5800X3D. It can't be beat.

Runner up is an Intel Platform with a 13600-700K.

Or wait for CES and see what AMD unveils. The Consumer 7000 Series Chips might be pretty damn nice. However, I suspect they may intro at the level of discounts that have been applied to the X Series chips for a while now (and the X series may go up in price).

FYI - Depending on what you are doing... I just stepped up to a 13900K and it's amazing. It is up to 20% faster than the top of AMD's product Stack and over 40% faster than my 5900X Overclocked! I'm just using DDR4! And it obliterates what I saw with the 7000 Series. You can probably get a kickass deal on a 13600 or 13700 and a Z690 Board for less than 200 bucks that is more than good enough (with either DDR4 or DDR5).
 
What I mean is that the initial spec for the AM5 platform was around DDR5-4800, that's the core default that the two AM5 Boards I worked with started the RAM at before Memory Training.

AMD Recommends that the sweet spot for the AM5 Platform is DDR5-6000 and that most boards and CPUs can get to that level.

Anything past that is gonna require more juice and tinkering and fabric clock tweaking and possibly a higher quality motherboard to get the faster ram to run stable.

I think, if you're going in basic you can pickup a Gigabyte Aorus AX X670 for around 275 bucks (US) and it should be a fairly robust board for an early adopter. It lacks the PCI-E 5.0 x16 slot, but it's a nice board for the price with a number of other decent features. And it puts you on the higher end chipset.

If you're not accustomed to the AMD Memory Training thing, this will be a rude awakening for you. Because the entire process may be painless (to some) but it's clunkier and slow as all hell to accomplish on these new boards (that's getting your RAM to Run at it's Rated 6,000 Speed).

You're gonna be spending around 300 bucks just to get on the platform and have some longevity. I do not recommend buying into the platform with a sub 200 dollar motherboard, it will be a cheap piece of shit and may not last.

Can you point me to information on this memory training aspect? 2x B650 (1 E and 1 Non E) platforms here and I've just set XMP or EXPO and gone on with my day.
 
FYI you can buy into the AM4 Platform in the sub 200 dollar price point and be just fine.
Dude, thank you so much for sharing so much valuable info - highly appreciate that! As for AM4 vs. AM5 - I really decided to jump into a new platform. Not that I have lots of money to spend, but I can afford myself a bit of investment in the future - just want this investment to be smart. :) Like I mentioned in the 1st message of the thread - I want this build to be more or less seriously "upgradeable" (CPU + maybe GPU) in approx. 2026. So this is not a build for years, but smth to sell for reasonable money in year 2030 (3-4 years after 2nd upgrade) and build smth totally new again.

I think, if you're going in basic you can pickup a Gigabyte Aorus AX X670
Noted, thank you! ;)

If you're not accustomed to the AMD Memory Training thing, this will be a rude awakening for you. Because the entire process may be painless (to some) but it's clunkier and slow as all hell to accomplish on these new boards (that's getting your RAM to Run at it's Rated 6,000 Speed).
Sorry for maybe a stupid quesion, but does it mean, that even factory-clocked 6000 won't reach that speed without this AMD Memory Training tribal dance?

Or wait for CES and see what AMD unveils. The Consumer 7000 Series Chips might be pretty damn nice.
Yep, waiting for them and YouTube tests too. TDP of 65W sound really attractive, given that I was focused on kinda compact build (~2.5l case), but not sure how much of productivity will be sacrificed vs. 105Whatted X series. Any idea? Btw., rumours are that non-X versions may have lower quality silicon - can be true?

You can probably get a kickass deal on a 13600 or 13700 and a Z690 Board for less than 200 bucks that is more than good enough (with either DDR4 or DDR5).
That will push me to Intel club for years... Totally not a hater, but so much time has already been spent on studying AMD lineup, reviews and stuff... :( In a nutshell: is what you propose a moneysaver or rather a technology-wise smarter decision?
 
1. What do you say about the statement (wide-spread, btw), that "all CPUs these days are factory-overclocked to the max"? So while 7600/7700's 65W TDP is definitely attractive, are there real opportunities to o/c them from 3.8GHz to 5.1-5.3GHz without risks? I've also heard rumors that non-X models are based on lower quality silicon...

2. They even recommend to slightly underclock/undervolt factory values if build is to be silent. This is the reason why I asked if anyone had experience with Eco mode in 7000 series - I'd probably use it for daily office tasks not to bother with manual underclocking (if only it is much more efficient vs. Eco mode).

3. Is it the fastest what I can have with DDR5 as of today? Without compromising its stability after overclocking of course. And also makes no sense to spend e.g. $100 more on the MoBo which probably supports higher RAM clocks (e.g. 7000), but that won't give me significant step forward in gaming experience (because of e.g. board topology limits you've mentioned).

1. It's indeed true but not all are factory overclock to the max. Take a look at the Raptor.

2. For daily office task, then the non-X variant will be a good match. Just wait for a couple of days as those line up will be revealed soon. Of course if you can fine 7700x way below it's MSRP launch, then just grab it.
Also, if you get 7700x, you can always do PBO + Curve Optimizer (CO) to optimize your best core, just do trial on each CO settings.

3. No, the fastest DDR5 setup already can reach 8400~8500 with TM5 testing.
For AM5, currently its limited by IMC of the CPU and AGESA.
Usually, 2 dimms mobo has potential to reach 7000+, for example this Tachyon B650E:

c_ohc=1CFu3GXj1o0AX_Xdni0&_nc_ht=scontent.fcgk22-1.jpg

It's overclock to 7400mhz using BCLK, so it is still limited by IMC but can be bypassed by oc the BCLK of cpu.
 
For daily office task, then the non-X variant will be a good match.
Waiting for them, right, but can't understand yet - how noticeable shall their decrease in productivity be? I mean 65W vs. 105W means some strings attached, right? I assume (maybe I'm mistaken though) that in office tasks/browsing I won't notice anything at all, in multitasking some drop in productivity will be felt, but gaming is probably the scenario that is mostly at risk, uh?

Also, if you get 7700x, you can always do PBO + Curve Optimizer (CO) to optimize your best core
Am I right, that this is smth that is done machine-wide, not task-wide? I mean I can't reduce curve for office tasks only, and have them back (high) again when I'm gaming...?
 
Can you point me to information on this memory training aspect? 2x B650 (1 E and 1 Non E) platforms here and I've just set XMP or EXPO and gone on with my day.
Perhaps it's just me... but nope... there are a shitload of other people out there experiencing the same issue. From what I have seen you're either lucky or you get stuck in a CMOS Clearing festival and a series of restarts praying for the BIOS to re-initialize so you can try all over again.

It either works flawlessly or it doesn't. Here are just a couple examples of the issues facing the AM5 Platform:

https://www.reddit.com/r/ASRock/comments/xugrkz/stuck_on_memory_training/
https://www.techpowerup.com/forums/threads/long-am5-post-times.300869/
https://linustechtips.com/topic/1461447-ddr5-ram-not-working-with-am5-socket/
https://forums.passmark.com/memtest...-stability-issue-inconsistent-memtest-results
https://www.guru3d.com/news-story/a...quires-hundreds-of-seconds-at-first-boot.html
https://forums.tomshardware.com/threads/new-build-boot-failure-diagnosis-x670-am5-ddr5.3782214/
https://www.techgamingreport.com/memory-training-ryzen-7000-takes-up-to-400-seconds-with-4-dimms/
 
Good input - thank you! Do I understand right, that by "RAM faster than DDR5-6000" you as well mean 6000 overclocked to 6400? Or only factory clocks count in this riddle? 'Cause if I have to buy $500 MoBo if I overclock my 6000 vs. $300 MoBo with factory clocks - I won't overclock maybe. :)
What I'm getting a sense of for the AM5 stuff is that the good vs the bad motherboards are something of a crapshoot. Either it's gonna work out of the box as some people have shown here on a B650 board with the DDR5 Running past spec at 7400+ or it's going to give you nightmares until you return the hardware.

You're kind of hell bent on going with the new AMD stuff, so I suspect you will know more from first hand experience in short order.

All I can do is wish you well, hope your experience is a smooth one and good luck!
 
Waiting for them, right, but can't understand yet - how noticeable shall their decrease in productivity be? I mean 65W vs. 105W means some strings attached, right? I assume (maybe I'm mistaken though) that in office tasks/browsing I won't notice anything at all, in multitasking some drop in productivity will be felt, but gaming is probably the scenario that is mostly at risk, uh?


Am I right, that this is smth that is done machine-wide, not task-wide? I mean I can't reduce curve for office tasks only, and have them back (high) again when I'm gaming...?
FYI - Riev90 's Motherboard is between 350 and 400 bucks for the "Lower End" B650 Chipset. That's not a cheap motherboard. That's because the E series boards have a PCI-E 5.0 16X slot on them and M.2 at 5.0 Speeds (instead of only one). It's not a cheap motherboard. Nothing decent on AM5 that will function properly will be cheap. Not yet, at least.

(on a side note I paid 300 Bucks for my Tomahawk Z790 and it has the PCI-E 5.0 X16 Slot, just not the M.2 5.0. Intel boards are offering the same or better features for a lower cost and there is allegedly one more generation on this board in the form of a Raptor Lake refresh at the end of 2023 I believe).

It has some crazy power phases and a shitload of amps

I believe you will want to invest in a high end Power Supply in any case: I usually recommend Seasonic

Due to the new power architecture change of DDR5, the power is directly supplied from the PSU to the PMIC with 5V. If the 5V power is unstable, it may cause an abnormal power supply of the memory and lead to potential system crash. Therefore, maintaining a stable 5V power in any condition has become crucial for those PSUs that are prepared to power the DDR5 memory. To maintain that stable supply of power, we suggest users to utilize power supply units with DC to DC circuit design. Compared to AC to DC circuit (Traditional Magnetic Amplifier Design), DC to DC converter design enjoys higher efficiency ratings and has more stable voltage.

His MB only has 2 Memory Slots and the RAM speeds he's getting are not 1:1 with the AMD fabric clocks. I think he even says as much in his post. If longevity is your thing, you can't just buy more RAM on that board. So be mindful of what board you choose if that is a consideration.

Curve Optimizer would effect everything the CPU does. Effectively you are undervolting the CPU to get it to drop the operating temperatures so that it will maintain a higher boost clock. That may not be exactly correct but its close enough. You have to screw with settings because not all CPUs will behave the same. I used this on my 5900X and it did work, but the speeds my 5900X topped out at were essentially the stock speeds for the Ryzen 7000 Series.

Allegedly you can control the undervolt to each, individual, CPU core. I am guessing that heavily depends on how nice your BIOS is on your board.
 
Last edited:
Waiting for them, right, but can't understand yet - how noticeable shall their decrease in productivity be? I mean 65W vs. 105W means some strings attached, right? I assume (maybe I'm mistaken though) that in office tasks/browsing I won't notice anything at all, in multitasking some drop in productivity will be felt, but gaming is probably the scenario that is mostly at risk, uh?


Am I right, that this is smth that is done machine-wide, not task-wide? I mean I can't reduce curve for office tasks only, and have them back (high) again when I'm gaming...?
1. Regarding the TDP, there certainly be drop in performance but not really noticeable (~10%), but for gaming? nah you will not notice it as IPC of ryzen 7000 or ADL / Raptor are top notch for now. So at 65w the non-X will probably boost to 4.9~5.2ghz on two cores using PBO, that's not too shabby.

2. Done on core utilization and setting on your windows power mode. If you're running heavy tasks, then all core will load to it's max, but if it's light load, then only some core will load. Yo don't have to worry about the Curve Optimizer settings as tis feature is actually one that can dictate the best core to run higher than the other worst core(s).
__________________________________

Legendary Gamer that's not my mobo, though, but regarding cheapish AM5 board, Asrock B650M PG Velocita is a quite good mobo, with 16x PCIE 4.0 graphic lane + 4x PCIE 5.0 for nvme. in my region, the price is ~185US$ and those features I mentioned along with improved VRM already exceeded the spec of last gen several X570's mobos.
 
You're kind of hell bent on going with the new AMD stuff, so I suspect you will know more from first hand experience in short order.

All I can do is wish you well, hope your experience is a smooth one and good luck!
Well, things are not the same as they were 3-5 days ago: I rather started to tend to Intel, and yesterday I was choosing between 13700K and 13600K with possibly latter to finally become my choice.
Intel clearly wins over 7000 series in both gaming and multitasking, while being cheaper and (!) more stable with other components (RAM, MoBo). The key point of my concern was that 13k is EOL, while my initial plan was to buy a platform with which 1 upgrade in the future will be possible (hence - AM5). However, such insignts as [italic is mine]:

Intel boards are offering the same or better features for a lower cost and there is allegedly one more generation on this board in the form of a Raptor Lake refresh at the end of 2023 I believe).
... make me think that Intel can be my choice, and even if something really extraordinary appears in 2-4 years (be that Intel or AMD) I can alsways skip this planned future upgrade, sell my build and work on the totally new one.
I have some questions on Intel too, but I won't ask them here to stay on-topic, and rather will create a new thread in Intel sub.


Due to the new power architecture change of DDR5, the power is directly supplied from the PSU to the PMIC with 5V. If the 5V power is unstable, it may cause an abnormal power supply of the memory and lead to potential system crash. Therefore, maintaining a stable 5V power in any condition has become crucial for those PSUs that are prepared to power the DDR5 memory.
As far as I understand, this recommendation is valid for all platforms - be it AM5 or Intel 13k - right?

Curve Optimizer would effect everything the CPU does. Effectively you are undervolting the CPU to get it to drop the operating temperatures so that it will maintain a higher boost clock. That may not be exactly correct but its close enough. You have to screw with settings because not all CPUs will behave the same.
[Italic is mine] What about "Intel 13k vs. Ryzen 7000" in exactly curve optimizing? Looks like you're experienced in both, so your comment will be really appreciated.


Done on core utilization and setting on your windows power mode. If you're running heavy tasks, then all core will load to it's max, but if it's light load, then only some core will load. Yo don't have to worry about the Curve Optimizer settings as tis feature is actually one that can dictate the best core to run higher than the other worst core(s).
Ah, OK, so curve optimizing is a dynamic thing, right? We set "rules" and the curve optimizer app applies them depending on the current situation.
However (as opposed to curve optimizing) XMP (Intel) or EXPO (AMD) profiles are "forever", right? I mean AFAIK they are set in BIOS, so the RAM clocks stay the same regardless of whether the current task is single- or multi-threaded and how intensive the load is, correct?
 
Well, things are not the same as they were 3-5 days ago: I rather started to tend to Intel, and yesterday I was choosing between 13700K and 13600K with possibly latter to finally become my choice.
Intel clearly wins over 7000 series in both gaming and multitasking, while being cheaper and (!) more stable with other components (RAM, MoBo). The key point of my concern was that 13k is EOL, while my initial plan was to buy a platform with which 1 upgrade in the future will be possible (hence - AM5). However, such insignts as [italic is mine]:
The 13700 and 13600 are solid choices for gaming and productivity. Intel has run into some manufacturing issues and it looks like we may see 14th Gen compatible with the Z690/Z790 Platform. They are supposed to have a Raptor Lake Refresh AFAIK from recent Intel Slides.
... make me think that Intel can be my choice, and even if something really extraordinary appears in 2-4 years (be that Intel or AMD) I can alsways skip this planned future upgrade, sell my build and work on the totally new one.
I have some questions on Intel too, but I won't ask them here to stay on-topic, and rather will create a new thread in Intel sub.
Technically, any platform you buy is obsolete either before you get it or shortly after. Technology change is continuous. Support on the AM4 platform was tenuous for a long time. AMD clearly didn't want to make everything backwards compatible and features over the course their introduction were lost. IIRC PCI-E 4.0 Functionality was enabled to a degree for pre 400 Series motherboards and then AMD rolled that Microcode update over. There were a couple motherboard manufacturers that supported the new tech in their bios based on AMD's initial microcode and then dropped support as AMD swept it under the rug.

There are no guarantees on any platform. However, what I am thinking is that we will be getting some serious mileage out of the Intel (and even AMD) platforms. With the Intel ones you're guaranteed to get a PCI-E 5.0 Slot and if Video Cards move to it in a generation or two, you can run it at it's full capability. The biggest upgrade we can generally ever do is upgrading our Video Cards. With 13th Gen Intel or Even AMD 5000 Series up, the CPUs will feed the graphics cards just fine. Though 1080P gaming will thrive on Intel and AMD 7000 series platforms more than 5000 series due to frequency.
As far as I understand, this recommendation is valid for all platforms - be it AM5 or Intel 13k - right?
Yup (but you can still get DDR4 Motherboards with Intel, not AMD. I'm Rocking DDR4 with my 13900K and it's stupid fast)
[Italic is mine] What about "Intel 13k vs. Ryzen 7000" in exactly curve optimizing? Looks like you're experienced in both, so your comment will be really appreciated.
I haven't played much with undervolting my 13900K. I'm not even going for efficiency yet. All my cores are locked at 5.6-7 Ghz (E Cores at 4.6-7), they periodically boost to 5.9... I need to spend some time with the configuration to get a handle on the system and get the cores to unload and dial down while at idle. My 4600 Mhz RAM showed up yesterday but I don't have the time to even install it. I am losing a family member and I don't have the time or energy for the tinkering now.
Ah, OK, so curve optimizing is a dynamic thing, right? We set "rules" and the curve optimizer app applies them depending on the current situation.
However (as opposed to curve optimizing) XMP (Intel) or EXPO (AMD) profiles are "forever", right? I mean AFAIK they are set in BIOS, so the RAM clocks stay the same regardless of whether the current task is single- or multi-threaded and how intensive the load is, correct?
XMP is a performance predefined set of parameters that the BIOS identifies from the RAM you use and it runs the memory accordingly.

EXPO similar to XMP. It either works flawlessly, or is a shit show.

That's just pure memory speeds. Then there are the considerations about how they communicate with their CPUs. AMD: Fabric Speeds ; Intel: Gear Speeds

1:1 is what you want on both; AMD fabric should equal the RAM Clocks around 3600-4000 on 5K Series (most likely 3600/3800 = 1800 / 1900 Mhz) ; Intel should be Gear 1 if you can pull it off. The performance boost for Fabric 1:1 and Gear 1 is +1-2% in Single Threading. I didn't really examine it past that in Multithreading.
 
Last edited:
Well, things are not the same as they were 3-5 days ago: I rather started to tend to Intel, and yesterday I was choosing between 13700K and 13600K with possibly latter to finally become my choice.
Intel clearly wins over 7000 series in both gaming and multitasking, while being cheaper and (!) more stable with other components (RAM, MoBo). The key point of my concern was that 13k is EOL, while my initial plan was to buy a platform with which 1 upgrade in the future will be possible (hence - AM5). However, such insignts as [italic is mine]:

[Italic is mine] What about "Intel 13k vs. Ryzen 7000" in exactly curve optimizing? Looks like you're experienced in both, so your comment will be really appreciated.

Ah, OK, so curve optimizing is a dynamic thing, right? We set "rules" and the curve optimizer app applies them depending on the current situation.
However (as opposed to curve optimizing) XMP (Intel) or EXPO (AMD) profiles are "forever", right? I mean AFAIK they are set in BIOS, so the RAM clocks stay the same regardless of whether the current task is single- or multi-threaded and how intensive the load is, correct?
1. Whatever you choose between Intel Raptor vs AMD Raphael, both will satisfy you, be it for gaming or for productivity / daily task. They are on the same level performance, but as of now, Raptor wins the Ram tuning possibility due to superior IMC compared to Raphael.

AFAIK: Intel ADL was IMC & Bios limited too at their 1st release, with Z690 / B660 mobo being the limit too for ram tuning.

IMHO: if you're looking for PNP platform but also want to experience the platform tuning, go for 13600K + Z790 mobo. It's worth a try. Take the DDR5 option.

2. Both platform have different terms regarding curve optimizer / performance enhance.
AMD with CO / PBO, Intel with it's TVB. Both optimized by cooling selection, and both can be undervolt stable for daily use. So just as point 1, you can't go wrong with either choice.

3. It's dynamic thing, even if you set all core oc with static vcore, if you change the power settings in the windows, the core will go down as low as under 2000mhz per core with vcore hovering under 1v. Just ensure if you're trying to change the vcore value in the bios, you have to test both load and idle condition. You may want to leave your desktop idle >8 hours to test whether the vcore you set is stable or not. Check WHEA error too (can check via HWInfo).

Regarding the XMP / EXPO, once you set, you don't have to worry about the power, even ddr5 push to it's limit (7800~8200mhz daily), it may use measly 6w max per dimm.
 
Though 1080P gaming will thrive on Intel and AMD 7000 series platforms more than 5000 series due to frequency.
Well, I'm strongly focused on 1440p (not sure smooth 4K experience is within my budget, though), but I assume Gen 13 can perfectly secure 1440p too. Even more so, given that I don't rush for crazy FPS - my monitor will probably be focused on 1440p @ 144Hz.

I'm Rocking DDR4 with my 13900K and it's stupid fast
What does make more sense in 2023 with Gen 13 and e.g., Z790 MoBo in your opinion: take DDR4 with top factory clocks and overclock it to its max, or rather take "best within budget" DDR5 (e.g., 6000 CL40) and overclock it (e.g., to 6400)?

I am losing a family member and I don't have the time or energy for the tinkering now.
PMed you. :(
 
as of now, Raptor wins the Ram tuning possibility due to superior IMC compared to Raphael.
AFAIK: Intel ADL was IMC & Bios limited too at their 1st release, with Z690 / B660 mobo being the limit too for ram tuning.

IMHO: if you're looking for PNP platform but also want to experience the platform tuning, go for 13600K + Z790 mobo. It's worth a try. Take the DDR5 option.
13600K + Z790 is exactly what I tend to since yesterday - you're just reading my thoughts. :) Am I right Z790 has no limitations of Z690/B660?
As for DDR4/DDR5: with Z790 - won't it be better to start with DDR4 in 2023 and overclock it to max, then move to DDR5 in 2024 and o/c it too (but hopefully DDR5 prices will drop down next year)?

And one more question on the topic: I know Ryzen 5000's rule of a thumb was "sweet spot of a RAM speed is x2 of that of Infinity Fabric". What about Intel Gen 14 rule and shall it be the same for Gen 15 too?

You may want to leave your desktop idle >8 hours to test whether the vcore you set is stable or not. Check WHEA error too (can check via HWInfo)
Wow, seems a bit of a rocket science to me now. Do you have some good article in mind to read on Intel curve optimization? Kinda step-by-step guide maybe..

Thank you - appreciate your comments!
 
Well, I'm strongly focused on 1440p (not sure smooth 4K experience is within my budget, though), but I assume Gen 13 can perfectly secure 1440p too. Even more so, given that I don't rush for crazy FPS - my monitor will probably be focused on 1440p @ 144Hz.
You will be fine on either platform in that that regard
What does make more sense in 2023 with Gen 13 and e.g., Z790 MoBo in your opinion: take DDR4 with top factory clocks and overclock it to its max, or rather take "best within budget" DDR5 (e.g., 6000 CL40) and overclock it (e.g., to 6400)?
I just went DDR4 based off my nightmare fueled experiences with AMD. However, if you're stepping into the platform anew you should probably look at going the DDR5 route. DDR4 is gonna require a lot of tweaking and "space magic" to fully match DDR5 though it is possible. I also went that route (DDR4) to eliminate a problem with testing, because I have a stupid amount of DDR4 on hand from other builds & I got a helluva deal on some DDR4 4600 memory (32 gigs for 139 bucks).
PMed you. :(
TY, responded. Very Kind.
 
Back
Top