I planned to buy AMD, but now Intel 13600K is my fav (shouldn't it be *new* 13600 or 13700 instead?)

[Coolio]

Hi guys,

many thanks to those who helped me a lot in my thread in AMD section. After CES'23 announcements I came to the conclusion that Intel will be the best choice for my use scenario: office apps, gaming @ 1440p (4K when possible), occasional multitasking.
I strongly want to improve my 4K experience in the future, but now I'm not so focused on upgrade as I was before - I'm OK to just sell my 2023 build and buy smth. more "4K-ish" in e.g. 2026.

Anyway, now I'm building from scratch (based on Z790 MoBo), so may I ask for your comments on the following please:

1. My initial choice was 13700K ($440, hereinafter MSRP), however I tend more to 13600K now ($340). Firstly, the gain in FPS (~9% @ both 1080p and 1440p) seems to be not significant (I plan to buy a 144Hz monitor, so FPS=144 is my top anyway), secondly, my multi-core tasks (e.g., converting video while browsing) are rather occasional, so the productivity of +20% will be noticeable only in games that use multiple threads - there aren't many of them, right? However, 13700K's power consumption is +74% in multi-core and +24% in single-core tasks. What shall that difference in power consumption mean to me in reality - hotter CPU temperatures, higher electricity bills, noise? If no big difference here, but the productivity really differs - maybe 13700K has more sense? Anyway, looks like AMD's "leadership" of 95C working temp. is unbeated by any Intel CPU...
2. Now that non-K models were announced on CES'23, I think of whether or not they shall be my choice, in particular 13600 ($255 MSRP) or 13700 ($384 MSRP). I wanted to have a rather compact case (e.g., Dan A4 or slightly bigger), so TDP 65W vs. 125W (K-models) looks attractive. There are no tests so far, AFAIK, but what was said in Intel release (e.g., 13600 should be 30-40% faster than 13600K in multi-core operations; 2nd-generation implementation for DDR5 RAM in new CPUs, etc.) sounds persuasive. Don't you think?
3. I'm, however, slightly worried about non-K clocks - 2.7/5.0GHz (13600) and 2.1/5.2GHz (13700) - vs. 13600K, which is 3.5/5.1GHz. Should I want to overclock my CPU and/or play with curve - which makes more sense: K or non-K?

Thank you!

 

[Meeho]

No need to go higher than 13600K. As for non-K, wait for a proper comparison to see what the real world difference is.

 

[Coolio]

Well, here is the comparison of several tech specs that I really wanted to understand - how much do they really matter and for which tasks:

Green - clear winner, red - point of my concern, blue - smth. I don't understand at all.

1. What are the benefits of Turbo Boost Max Technology 3.0 that only 13700 has? [UPDATE] Found the answer myself: it pushes single-threaded tasks to fastest cores. Could be useful for games, I assume.
2. Is 0.2 GHz loss of Efficient-core MAX frequency (13600 vs. 13600K) noticeable at all?
3. In which tasks/scenarios will I feel the loss of 0.6 GHz of Efficient-core BASE frequency and 0.8 GHz of Performance-core BASE frequency (13600 vs. 13600K)?
4. In which tasks/scenarios will I feel the loss of 8.5 Mb of L2 cache (13600 vs. 13600K) or gain of 4 Mb of such cache (13700 vs. 13600K)?
5. Am I right that the senior CPU (13700) has lower BASE clocks (for both Performance & Efficient cores) just to reach low TDP (65W vs. 105W in i5 models)?
6. Am I right that the above-mentioned DDR5 speed limits stay even if my MoBo has them higher (e.g., Z790 with up to 6,800MT/s with 1 DIMM per channel), so in order to reach MoBo's limits I will need to upgrade my CPU in the future?

Thank you guys!

 

[gerardfraser]

Truth is buy whatever you want it is CPU gear and it always changes.(AMD will have the Fastest PC Gaming CPU's next month in Feburary.That would be your best rig to buy)

1440p is fine if you have a very good monitor with higher FPS at the resolution than 4K but really not noticeable if you have a 1440P next to a 4K monitor.I would still pick 4K all day long.

Buy 13900KS for fastest CPU today-Will be best in 2026 compared to 13600K/13700K.I know your not thinking of this CPU.

13600K- Is 100% fine for PC Gaming compared to the 13700K at 1440p and 4K
Z790 go with DDR5 Ram with 4K PC Gaming anything at CL36 5600MT's+ will be the same FPS wise in 99.8% of games at resolution 1440p+


Playing PC Games at 4k the CPU clock has very little effect blah blah blah I could keep spewing stuff .There is no magic in PC gear when PC Gaming.

Buy anything new you will be fine for your use case and will not miss out on anything.

Here is a fine example I did yesterday in the Witcher 3.

I had Windows power saver plan running and minimized game and switched to high power plan. Notice the difference on my 13700KF FPS and CPU usage,then notice the FPS stuff.

 

[Coolio]

Playing PC Games at 4k the CPU clock has very little effect blah blah blah I could keep spewing stuff .There is no magic in PC gear when PC Gaming.

Buy anything new you will be fine for your use case and will not miss out on anything.

Well, any of those 3 will be fine, but they are obviously not the same, so I ask for some explanation: which particular spec may influence what, hence - concrete questions. I like to withdraw conclusions too, just need some help with that.

1440p is fine if you have a very good monitor with higher FPS at the resolution than 4K but really not noticeable if you have a 1440P next to a 4K monitor.

AFAIK, 144Hz @ 1440p is good enough these days, and 144Hz @ 4K is possibly the best for the amateur gamer. Not sure, though, what you meant by the part I underlined in your quote...

13600K- Is 100% fine for PC Gaming compared to the 13700K at 1440p and 4K

13600K vs. 13700K comparison is a minor headache for me now. What really interests me is 13600K vs. 13700.
Cores/threads are obviously not a subject for discussion for game purposes. But what about 1GHz+ difference in base clocks (in favor of 13600K)? What about +4Mb of L2 cache (in favor of 13700)?

anything at CL36 5600MT's+ will be the same FPS wise in 99.8% of games at resolution 1440p+

What I've heard was that timings are more important for gaming (while clocks - for huge data flows). Anyway, since Z790 MoBo supports DDR5 up to 6,800MT/s, but 13600K/13700 are limited with 5,600MT/s, I'll be bottlenecked on CPU side, so DDR5-5600 is the top of what makes sense with these CPUs, right? I mean makes no sense even to overclock such module (to e.g., 6000), is that so?

 

[kirbyrj]

The base frequency is almost irrelevant. Pretty much any non-budget basement board is going to allow you to slightly tweak the power limits allowing for unlimited boost, etc. Even under a load, the cores are never at the "base frequency." I would guess if you're board was extremely limited and you were operating exactly within the 65W limit, the base frequency would show itself after the initial 56 seconds of turbo boost (or whatever it is programmed into the PL1/PL2 settings). Another thing is you can limit the 125W CPUs to less than 125W in the bios.

Personally, I'd get a 13700 and a mid range B660 or B760 board if you don't plan on overclocking. There is no difference between Z790 and B760 other than the ability to unlock the multiplier for overclocking on the Z790 with "K" CPUs. If you buy the 13700 non-k you don't have an unlocked multiplier anyway. Might as well get a cheaper board. You can still overclock the memory (XMP settings) with a cheaper motherboard. And MSI even has a "Max" version that lets you overclock the bus speed of locked CPUs to increase performance. Any increase in speed though comes at the expense of heat and power draw which you're trying to avoid.

 

[gerardfraser]

Well, any of those 3 will be fine, but they are obviously not the same, so I ask for some explanation: which particular spec may influence what, hence - concrete questions. I like to withdraw conclusions too, just need some help with that.


AFAIK, 144Hz @ 1440p is good enough these days, and 144Hz @ 4K is possibly the best for the amateur gamer. Not sure, though, what you meant by the part I underlined in your quote...


13600K vs. 13700K comparison is a minor headache for me now. What really interests me is 13600K vs. 13700.
Cores/threads are obviously not a subject for discussion for game purposes. But what about 1GHz+ difference in base clocks (in favor of 13600K)? What about +4Mb of L2 cache (in favor of 13700)?


What I've heard was that timings are more important for gaming (while clocks - for huge data flows). Anyway, since Z790 MoBo supports DDR5 up to 6,800MT/s, but 13600K/13700 are limited with 5,600MT/s, I'll be bottlenecked on CPU side, so DDR5-5600 is the top of what makes sense with these CPUs, right? I mean makes no sense even to overclock such module (to e.g., 6000), is that so?

Well I can tell you only from my experience,here it is. It does not make a difference in what you buy,anything new is the best way to go and then AMD in February will be the fastest PC Gaming CPU out there.

Last couple years my PC Gaming CPU's and I could not tell the difference today with FPS counter turned off on any of them at 1440p/4K

AMD 2600X/3600X/3600XT/3800X/3800XT/5600X/5800X/7700X
Intel 10850K/12600K/12900K/13700KF

Here are 4 PC Gaming rigs I used in the last couple months. So depending on the GPU ,the GPU determined what resolution I would play my PC Games with such as 2K/4K. The CPU was never a factor as they are all around the same.
I also ran 1000's of benchmarks with ram timings.Ram timings affect low resolutions and low settings not so much higher resolution.

Sept 2022 AMD 7700X/RTX 3090- Highest FPS with CPU in games I played by a couple FPS comparing same GPU Such as RTX 3070Ti/3080/3090/RTX4090


Oct 2022 12600K/RTX 4090-


Oct 2022 12900K/RTX 3090-


Nov 2022 13700KF/RTX 3070Ti-This Rig I would play at 2K/1440P because of GPU

 

[Coolio]

The base frequency is almost irrelevant.

Thank you - good to know that! I was initially confused with Intel's release, which said smth. like "new i5-13600 will have a +30-40% boost in multi-threaded performance vs. its "K" version", while its Base/Turbo clocks are lower than K's. I just wondered how they were going to achieve that - that's where my questions about importance of clocks arised from. We'll see what real tests will show though.

Another thing is you can limit the 125W CPUs to less than 125W in the bios.

... and compared to non-limited 65W (e.g. stock 13600 vs. slightly limited 13600K) this will be..... what? I mean in practice.

There is no difference between Z790 and B760 other than the ability to unlock the multiplier for overclocking on the Z790 with "K" CPUs. If you buy the 13700 non-k you don't have an unlocked multiplier anyway. Might as well get a cheaper board. You can still overclock the memory (XMP settings)

Very useful - thank you! I was thinking of 13700, but it's ~$380 vs. ~$340 (13600K), so if I'm able to save those ~$50 (or more) on MoBo - rises the chances of i7.
I'm not going to o/c CPU, but I will probably o/c DDR5, which you said is supported by B760.

By the way - does underclocking/undervolting make sense with Gen 13 CPUs to drop operating temps and keep high clocks, or is Turbo Boost 3.0 being ON enough, since it will manage clocking/volting levels itself in real-time?

And MSI even has a "Max" version that lets you overclock the bus speed of locked CPUs to increase performance. Any increase in speed though comes at the expense of heat and power draw which you're trying to avoid.

You mean non-K CPUs here I assume. Not that I'm trying to avoid heat as such, I'm trying to engineer the build which will be properly cooled (hence - pretty silent) and (if possible) locked in ~10L case.
So this "Max" thing replaces Turbo Boost 3.0, or rather complements it?

 

[Coolio]

I also ran 1000's of benchmarks with ram timings.Ram timings affect low resolutions and low settings not so much higher resolution.

Thank you - appreciate visual part too! Well, I've heard that RAM timings are more important for games (frequent, low-data tasks), while heavy-data tasks are more sensitive to clocks. Just quote from someone on Reddit.
As for the build (old/new/this/that) - well it's a hobby to the certain extent, right? Learn, build, know smth. new. And I feel much more secure when I understand what I'm doing. Just the way I am, not that I don't believe in out of the box advices.

 

[chameleoneel]

Well, here is the comparison of several tech specs that I really wanted to understand - how much do they really matter and for which tasks:

View attachment 540250

red - point of my concern,

1. What are the benefits of Turbo Boost Max Technology 3.0 that only 13700 has? [UPDATE] Found the answer myself: it pushes single-threaded tasks to fastest cores. Could be useful for games, I assume.
2. Is 0.2 GHz loss of Efficient-core MAX frequency (13600 vs. 13600K) noticeable at all?
3. In which tasks/scenarios will I feel the loss of 0.6 GHz of Efficient-core BASE frequency and 0.8 GHz of Performance-core BASE frequency (13600 vs. 13600K)?
4. In which tasks/scenarios will I feel the loss of 8.5 Mb of L2 cache (13600 vs. 13600K) or gain of 4 Mb of such cache (13700 vs. 13600K)?
5. Am I right that the senior CPU (13700) has lower BASE clocks (for both Performance & Efficient cores) just to reach low TDP (65W vs. 105W in i5 models)?
6. Am I right that the above-mentioned DDR5 speed limits stay even if my MoBo has them higher (e.g., Z790 with up to 6,800MT/s with 1 DIMM per channel), so in order to reach MoBo's limits I will need to upgrade my CPU in the future?

Thank you guys!

RE: red - point of my concern
Basically, the 13400, 13500, and 13600 are actually specced halfway between Alder Lake and Ratpor Lake. That's why they don't have the increased L2 cache or the same default memory spec. However, they do have increased clock speeds compared to Alder Lake----So they must be using the same refined silicon for Raptor Lake. And they also have increased L3 cache. These changes are to help keep them from being too close in performance to the 13600k. And also probably keeps Intel's costs even lower, for these high volume parts which will go into a lot of Dell PCs, etc.

1. The 13700 and 13700k have Turbo Boost 3.0, because not all of the cores boost to the same speed. Only 1 or 2 cores boost to 5.2ghz. In a multicore workload, the speed of all cores will be lower. 13600 and 13600k do not have Turbo boost 3.0, because all 6 cores run at the same speed, always.

2. The lower e-core speeds and cache will only be noticeable in multicore work. Like a Cinibench score or the amount of time it takes to Run a render in Blender, encode a video, etc.

3. Basically, same answer at #2
4. The increased cache for the Raptor Lake K CPUs, seems to be part of why they perform so much better in gaming, compared to Alder Lake. The extra clock speed is another reason. The cache also seems to help the e-cores perform better, in general.

5. Partially correct. But also, to help keep heat even lower for basic tasks. The 60w parts get put into a lot of really small PCs, with very small heatsinks. However, you can unlock the power limits in the bios of a decent motherboard and then the 13700 will use as much power as a 13700k. maybe even more power--the non-k parts are often slightly lower grade silicon.

6. No, that is not correct. For the Non-K CPUs, You can use XMP and overclock RAM on B660, B760, Z690, and Z790. The limitation for B660 and B760, is that you cannot increase the voltage of the system Agent, which is basically the voltage for the memory controller. So, you will be limited on top end RAM overclocking and/or getting the timings really low, beyond the XMP.

 

[kirbyrj]

Thank you - good to know that! I was initially confused with Intel's release, which said smth. like "new i5-13600 will have a +30-40% boost in multi-threaded performance vs. its "K" version", while its Base/Turbo clocks are lower than K's. I just wondered how they were going to achieve that - that's where my questions about importance of clocks arised from. We'll see what real tests will show though.


... and compared to non-limited 65W (e.g. stock 13600 vs. slightly limited 13600K) this will be..... what? I mean in practice.


Very useful - thank you! I was thinking of 13700, but it's ~$380 vs. ~$340 (13600K), so if I'm able to save those ~$50 (or more) on MoBo - rises the chances of i7.
I'm not going to o/c CPU, but I will probably o/c DDR5, which you said is supported by B760.

By the way - does underclocking/undervolting make sense with Gen 13 CPUs to drop operating temps and keep high clocks, or is Turbo Boost 3.0 being ON enough, since it will manage clocking/volting levels itself in real-time?


You mean non-K CPUs here I assume. Not that I'm trying to avoid heat as such, I'm trying to engineer the build which will be properly cooled (hence - pretty silent) and (if possible) locked in ~10L case.
So this "Max" thing replaces Turbo Boost 3.0, or rather complements it?

I don't think there's a lot difference between the k and non-K other than the multiplier it runs at and I'm sure the k version is somewhat better binned for higher clocks.

I don't think it makes a lot of sense to undervolt the k series. Most of the value is in the unlocked multiplier. But you always can do that if you wanted to. Occasionally you get better deals on the k parts.

Yes, I was talking about the non-K locked CPUs. The max board complements the turbo boost because a lot of the turbo boost is based on the multiplier not the bus speed. The max board has a external clock generator so it doesn't throw off the board in other aspects. I've done this before with a 10700 and I was pleased with the results. Only back then you had to use a z490 board and now you can just use a b series motherboard.

In your shoes I would probably get a b760 motherboard with good VRMs that allow for an unlimited boost and then get the 13700 non-K.

 

[Coolio]

chameleoneel Thank you mate for such a detailed reply and plenty of useful info - highly appreciate your patience, really!

1. The 13700 and 13700k have Turbo Boost 3.0, because not all of the cores boost to the same speed. Only 1 or 2 cores boost to 5.2ghz. In a multicore workload, the speed of all cores will be lower. 13600 and 13600k do not have Turbo boost 3.0, because all 6 cores run at the same speed, always.

2. The lower e-core speeds and cache will only be noticeable in multicore work.

OK, look: 13600K has E-cores at 2.60/3.90GHz (BASE/TURBO), while 13700 has them at 1.50/4.10GHz - so much lower base and pretty higher turbo (same with P-cores actually). Am I right, that Turbo Boost 3.0 handles both E-cores and P-cores, so if the task is within 1-2 cores workload (e.g. gaming) 13700 will perform better despite of its lower base E-core/P-core clocks, because at least 1-2 cores will be automatically boosted to their Turbo clocks (which are higher for 13700 vs. 13600K)?

The 60w parts get put into a lot of really small PCs, with very small heatsinks. However, you can unlock the power limits in the bios of a decent motherboard and then the 13700 will use as much power as a 13700k. maybe even more power--the non-k parts are often slightly lower grade silicon.

As for MoBo I am focused on a B760 these days, kindly recommended by kirbyrj as the one that makes more sense if I don't hit the K-series road (which seems like I won't ). So given that 13700's silicon is probably vulnerable, does it make sense to unlock it from 65W to e.g. 85W (with appropriate cooling of course), or there are more risks than benefits and I will gain more from o/c my RAM and (maybe) GPU?

The limitation for B660 and B760, is that you cannot increase the voltage of the system Agent, which is basically the voltage for the memory controller. So, you will be limited on top end RAM overclocking and/or getting the timings really low, beyond the XMP.

Not sure I understood, sorry. Let's assume I have a 13700 on B760 with DDR5-6000 CL40. Why should I ignore the "up to DDR5 5600 MT/s" part of the CPU's spec, how comes I can use a higher-clocked (6000 MHz) RAM and even overclock it? Maybe MT/s is not the same as MHz and that is the source of my confusion... And finally - did I get it right that o/c RAM on B760 with non-K CPU is not much beneficial compared to what factory XMP profiles offer, so I can just rely on them without losing much?

 

[Coolio]

The max board complements the turbo boost because a lot of the turbo boost is based on the multiplier not the bus speed. The max board has a external clock generator so it doesn't throw off the board in other aspects. I've done this before with a 10700 and I was pleased with the results.

Thank you for your comments - very informative! Do you have any saved links maybe on how to manage that MSI "Max" boards' overclocking properly..?

In your shoes I would probably get a b760 motherboard with good VRMs that allow for an unlimited boost and then get the 13700 non-K.

99.99% that exactly will be my choice - too many reasons for that combo. Will appreciate if you kindly share names of a couple of decent B760 MoBos to pay attention to. I will read reviews myself too, of course!

 

[kirbyrj]

Not sure I understood, sorry. Let's assume I have a 13700 on B760 with DDR5-6000 CL40. Why should I ignore the "up to DDR5 5600 MT/s" part of the CPU's spec, how comes I can use a higher-clocked (6000 MHz) RAM and even overclock it? Maybe MT/s is not the same as MHz and that is the source of my confusion... And finally - did I get it right that o/c RAM on B760 with non-K CPU is not much beneficial compared to what factory XMP profiles offer, so I can just rely on them without losing much?

MT is essentially the correct term for what people call MHz with RAM because the MHz are actually half of the MT due to the double data rate design. B660/B760 will generally allow the bottom 90% of memory to run fine with the limit of no SA voltage adjustment. chameleoneel is right in that you might run into situations where you'd have to downclock or loosen timings on the highest end RAM, but if you're budgeting for the highest end RAM, then you might as well budget for Z790, etc. as it will run significantly more expensive than decent mid-range RAM. I don't know that you'd even be able to tell a difference above DDR5 6000 CL36 or something like that running on XMP settings.

Thank you for your comments - very informative! Do you have any saved links maybe on how to manage that MSI "Max" boards' overclocking properly..?


99.99% that exactly will be my choice - too many reasons for that combo. Will appreciate if you kindly share names of a couple of decent B760 MoBos to pay attention to. I will read reviews myself too, of course!

Here's some info in the Max motherboard:
https://wccftech.com/msi-demonstrat...0m-mortar-motherboard-features-new-oc-engine/
https://www.tweaktown.com/news/8962...re-i5-13600k-pushed-to-huge-5-8ghz/index.html

Big limit though is that it appears to be DDR4 only. Not sure if there is a B760 max with DDR5 or not.

I haven't specifically looked at the B760 range, but you could look at some B660 reviews and extrapolate where the B760 would land depending on the board.
https://www.techspot.com/review/2432-intel-b660-motherboards-midrange/
https://www.techspot.com/review/2426-intel-b660-motherboards/

Personally, I had the MSI B660M mortar (non-max). It was a great board and worked fine with my 13700k. I ended up upgrading it to the Z790 Tomahawk only because someone bought the Mortar off me. Day to day, I don't notice any difference between the two.

 

[chameleoneel]

chameleoneel Thank you mate for such a detailed reply and plenty of useful info - highly appreciate your patience, really!



OK, look: 13600K has E-cores at 2.60/3.90GHz (BASE/TURBO), while 13700 has them at 1.50/4.10GHz - so much lower base and pretty higher turbo (same with P-cores actually). Am I right, that Turbo Boost 3.0 handles both E-cores and P-cores, so if the task is within 1-2 cores workload (e.g. gaming) 13700 will perform better despite of its lower base E-core/P-core clocks, because at least 1-2 cores will be automatically boosted to their Turbo clocks (which are higher for 13700 vs. 13600K)?



As for MoBo I am focused on a B760 these days, kindly recommended by kirbyrj as the one that makes more sense if I don't hit the K-series road (which seems like I won't ). So given that 13700's silicon is probably vulnerable, does it make sense to unlock it from 65W to e.g. 85W (with appropriate cooling of course), or there are more risks than benefits and I will gain more from o/c my RAM and (maybe) GPU?



Not sure I understood, sorry. Let's assume I have a 13700 on B760 with DDR5-6000 CL40. Why should I ignore the "up to DDR5 5600 MT/s" part of the CPU's spec, how comes I can use a higher-clocked (6000 MHz) RAM and even overclock it? Maybe MT/s is not the same as MHz and that is the source of my confusion... And finally - did I get it right that o/c RAM on B760 with non-K CPU is not much beneficial compared to what factory XMP profiles offer, so I can just rely on them without losing much?

13700 will be faster than 13600 in single core, because it can boost 1 or 2 cores to a higher speed, with turbo boost 3.0.

You can unlock the power limits on a 13700 non-K as high as you want. Fully unlocked, it will use ~250 watts. There is no risk of damaging it. It might actually use a little more power than an unlocked 13700K----because the K CPUs usually have slightly more efficient Silicon (which allows them to have higher clock speeds).

DDR5 5600 is the speed which Intel has officially specced the 13th gen memory controller. Anything higher than that, is technically overclocking the memory controller. If you buy RAM which is DDR5 5600 and overclock the RAM to 6000 --- you have overclocked the RAM and are technically overclocking the memory controller, along with it. If you buy DDR5 6000 RAM and use the XMP profile to run it at 6000---you are not overclocking that RAM. But, you are still overclocking the memory controller.

If your DDR5 is fast enough----you may have to add more voltage to the memory controller, to make it stable at such high speeds. Eventually, it won't ever be stable. 12th Gen memory controllers maxed out somewhere around DDR5 6400. 13th gen is a lot more capable. Over 7000 seems common. Some people are getting 8000.

 

[Coolio]

DDR5 5600 is the speed which Intel has officially specced the 13th gen memory controller. Anything higher than that, is technically overclocking the memory controller.

12th Gen memory controllers maxed out somewhere around DDR5 6400. 13th gen is a lot more capable. Over 7000 seems common.

OK, so (how I understand the situation): 5600 is Gen 13 recommended, but in reality I can go for higher speeds, i.e. overclock 6000 module up to 7000

I don't know that you'd even be able to tell a difference above DDR5 6000 CL36 or something like that running on XMP settings.

If you buy DDR5 6000 RAM and use the XMP profile to run it at 6000---you are not overclocking that RAM. But, you are still overclocking the memory controller.

However, if I'm OK with up to 6000 speed, I can buy 6000 module and use XMP profiles - they will handle it at the said speed. Anything above 6000 (e.g. overclocking 6000 DIMMs to 7000, as mentioned above) will require manual o/c. Right?

The limitation for B660 and B760, is that you cannot increase the voltage of the system Agent, which is basically the voltage for the memory controller. So, you will be limited on top end RAM overclocking and/or getting the timings really low, beyond the XMP.

@chameleoneel is right in that you might run into situations where you'd have to downclock or loosen timings on the highest end RAM, but if you're budgeting for the highest end RAM, then you might as well budget for Z790, etc. as it will run significantly more expensive than decent mid-range RAM.

Finally, the choice of factory DIMM speeds depends on the motherboard: should I choose Z790, I can buy cheaper/slower RAM (e.g. 5600) and boost it to e.g. 7000, while B760 MoBo limits my choice to higher speed DIMMs (e.g., 6000, 6400), because I can only play with DIMM timings on this MoBo (not boost RAM clocks). Even more so - in my attempts to get benefit from new timings, I will have to drop the factory clocks.
TLDR: either cheaper MoBo + losses on expensive RAM, or more expensive MoBo + cheaper RAM.

Did I get everything right now? Sorry guys, RAM has always been kinda maze for me. And thank you for your patience!

 

[kirbyrj]

OK, so (how I understand the situation): 5600 is Gen 13 recommended, but in reality I can go for higher speeds, i.e. overclock 6000 module up to 7000




However, if I'm OK with up to 6000 speed, I can buy 6000 module and use XMP profiles - they will handle it at the said speed. Anything above 6000 (e.g. overclocking 6000 DIMMs to 7000, as mentioned above) will require manual o/c. Right?




Finally, the choice of factory DIMM speeds depends on the motherboard: should I choose Z790, I can buy cheaper/slower RAM (e.g. 5600) and boost it to e.g. 7000, while B760 MoBo limits my choice to higher speed DIMMs (e.g., 6000, 6400), because I can only play with DIMM timings on this MoBo (not boost RAM clocks). Even more so - in my attempts to get benefit from new timings, I will have to drop the factory clocks.
TLDR: either cheaper MoBo + losses on expensive RAM, or more expensive MoBo + cheaper RAM.

Did I get everything right now? Sorry guys, RAM has always been kinda maze for me. And thank you for your patience!

No, B660/760 allows you to adjust RAM speed also. That's the big difference between previous B series boards where you were more limited.

The XMP settings set the timings and speeds to whatever it says on the box. You can go from there and try to further increase clock speeds or lower timings, but it's not guaranteed and will require manual adjustment. Both B and Z series boards allow you to manually adjust these.

The only difference between cheaper RAM and higher speed RAM is that it is factory binned for those speeds. I doubt you'd get a set of 5600 that boosts to 7000 because if that particular set could do that, it would have been binned higher and sold for more money. So you can get any B or Z series motherboard and you'll still have a range of memory overclocking available to you. chameleoneel related that you can't adjust the system agent voltage on B series motherboards (I can't confirm or deny that), which might give you a little more latitude with lower end memory overclocking to higher speeds, running higher end speeds, or running extremely tight timings. But if you buy a reasonably priced set of memory (5600 to 6400 range), it should work fine, especially with a 13th gen CPU which has a better memory controller anyway. Once you start getting beyond 6400, the price increases dramatically where you certainly get diminishing returns in terms of bang for the buck.

 

[chameleoneel]

OK, so (how I understand the situation): 5600 is Gen 13 recommended, but in reality I can go for higher speeds, i.e. overclock 6000 module up to 7000




However, if I'm OK with up to 6000 speed, I can buy 6000 module and use XMP profiles - they will handle it at the said speed. Anything above 6000 (e.g. overclocking 6000 DIMMs to 7000, as mentioned above) will require manual o/c. Right?




Finally, the choice of factory DIMM speeds depends on the motherboard: should I choose Z790, I can buy cheaper/slower RAM (e.g. 5600) and boost it to e.g. 7000, while B760 MoBo limits my choice to higher speed DIMMs (e.g., 6000, 6400), because I can only play with DIMM timings on this MoBo (not boost RAM clocks). Even more so - in my attempts to get benefit from new timings, I will have to drop the factory clocks.
TLDR: either cheaper MoBo + losses on expensive RAM, or more expensive MoBo + cheaper RAM.

Did I get everything right now? Sorry guys, RAM has always been kinda maze for me. And thank you for your patience!

Really, its about 2 things:
1. The memory controller on your own CPU. Once you reach the max speed or timings it can handle----you need to add voltage to the system agent, to get better speeds and timings. It doesn't matter if you buy RAM which has an XMP which is already too fast-----or if you buy some slower RAM and are able to overclock it faster. If its too fast for the memory controller, the memory controller (system agent) needs more voltage. And you can't do that on B660 or B760 boards. For 13th gen, that is less of an issue. Because the Memory controller for DDR5 is a lot better.

2. Its also about the RAM itself. If you are trying at add more speed beyond the XMP or timings lower than the XMP, eventually you will need to add voltage to the RAM. you are able to do that on B660 and B760 boards. DDR5 has 3 different voltages. If you want to overclock the RAM beyond XMP, I suggest you read a couple of articles or watch some videos, for overclocking DDR5.

No, B660/760 allows you to adjust RAM speed also. That's the big difference between previous B series boards where you were more limited.

B560/H570 was the first time non-Z boards could use faster RAM. And those boards allowed you to change the system agent voltage! Intel took that away, for the 6 and 7 series non-Z boards : \

 

[Coolio]

But if you buy a reasonably priced set of memory (5600 to 6400 range), it should work fine, especially with a 13th gen CPU which has a better memory controller anyway. Once you start getting beyond 6400, the price increases dramatically where you certainly get diminishing returns in terms of bang for the buck.

If you want to overclock the RAM beyond XMP, I suggest you read a couple of articles or watch some videos, for overclocking DDR5.

Guys, thank you a lot - it's all clear now! I will buy something decent within a 5600-6400 range (as recommended): possibly 6000 СL40 will be my target, and if special offer for DIMMs with higher clocks/lower timings appears - I will buy those instead.
And I guess I will start my o/c experience from just using XMP profiles - I guess they can bring good results, right? As a novice in o/c I will have a chance to see how it works with XMPs, and I can dive deeper in manual overclocking anytime with "DDR5 + 13700 + B760" set, which, while is somehow limited from MoBo side, still allows some manual o/c.

Just a few last questions on RAM, if you guys don't mind:

1. Speaking of decent DIMMs: which RAM chip is better (in general and for XMP/overclocking) - Hynix M-Die or Samsung B-Die?
2. I've heard 2 opinions: "higher speed RAM with good timings should outperform slower RAM with tighter timings" and "on Intel-based builds - improve RAM timings, if you mostly game (i.e., frequent tasks, but light in data), and improve clocks if your tasks are data-heavy". Do you agree that for gaming purposes, when keeping my eye on DIMM special offers and playing with XMP settings - I should pay more attention to RAM timings, than frequency?
3. I've also heard that with Intel (as opposed to AMD) no Memory Training is needed for DDR4/DDR5 to start working on spec-defined (!) clocks. To be honest, I was surprised that smth. written on the box doesn't work out-of-the-box, but do I really need to do smth. extra with my DDR5 DIMMs so that they start working at the speed I paid for (e.g., 6000), or become ready for further tuning via XMP profiles?
4. Last rumor I've heard is that "Gear 1 doesn't work with DDR5 as the default configuration is Gear 2". Do I need to change some settings in Bios for Gear 1 to become active? Are XMP profiles (or manual o/c) somehow dependent on whether Gear 1 or 2 is active?

Thank you and have a nice weekend!

 

[Keljian]

To answer what I know:
1. Hynix M-Die seems to be the pick for ram at the moment
2. If ram timings/speed are the thing that is going to make or break the build, there's a problem
3. I run a 13600k w/ DDR4 , the first boot after changing a memory clock takes time, but so long as you have XMP settings or EXPO settings (ideally just XMP) it should just work. You can tweak higher for a few extra percentage points of performance.
4. XMP profiles sometimes specify what "gear" the processor's controller is in, but the default is gear 2, gear 1 is faster, but by a small margin.

All of that said... I'm going to rock the boat...

Based on your first post, you say a few percentage points of performance is not going to make a big difference

firstly, the gain in FPS (~9% @ both 1080p and 1440p) seems to be not significant (I plan to buy a 144Hz monitor, so FPS=144 is my top anyway), secondly, my multi-core tasks (e.g., converting video while browsing) are rather occasional, so the productivity of +20% will be noticeable only in games that use multiple threads - there aren't many of them, right? However, 13700K's power consumption is +74% in multi-core and +24% in single-core tasks.

Let me give you a couple of insights that may help

• 200mhz (0.2ghz) at 5ghz is about 4%. Keep this in mind with all things you're considering.
• In pure productivity, the difference between ddr4/3200 and ddr5/6000 is not much - in cinebench results, which everyone seems to quote, it's the difference between about 23.5k and 24.0k (2.1%)
• 2 extra p cores (8 cores vs 6) will make very little difference in games as with the technology baked into these cpus, threads that need quick performance will be allocated to the P cores, and background threads (like audio) will be moved to the e cores, Further, I had a 9900k (8c/16t) previous to the 13600k, and most of the time when playing games 4-6 cores were used, the other cores were either parked with my power saving settings (wake a new core on 60% load) or barely ticking over. This includes modern titles like cp2077, shadow of the tomb raider, doom eternal etc..
• The only time a lot of games use more is when loading resources, where some games will multithread the resource loads
• Non-K clocks for e cores shouldn't worry you (as you said your multi core tasks are occasional) - does it matter if they're running at 60-80% vs 50-60% if they're just chugging away in the background?
• The 13600k is actually reasonably efficient. Despite using (stock) all-core clocks, myself and many others have managed a -100mv offset voltage without it effecting stability at all, which results in around 110W at peak. Default peaks are about 185W on most motherboards.
• I'm going to disagree with kirbyrj - undervolting makes sense if it gives you thermal headroom to turbo higher. It's a balance between stability, heat, and power. Less heat, means more turbo room, but less voltage means potential instability. On my 9900k, I ran a -55mv offset, which allowed 5.1ghz with 0 AVX offset, had I increased or left the default voltage, I would not have managed to clock it that high.
• You can limit the K (and other) parts to lower TDPs in the bios.. your call whether this is necessary
• You can also set the voltage and speed of both the P and the E cores, so on a non K part (according to this - which suggests all are unlocked), it is very likely you could speed up those E cores if you wanted to (though intel don't guarantee you'll hit those clocks), but your power usage and heat output would likely be in line with the K parts anyhow.
• It is trivial to set most 13600ks to 5.2ghz/4.2ghz even with an undervolt
• While the 13600k has "fixed" boost speeds, in most bioses you can set per core boost speeds, so you can simulate the 2 core boost behavior by changing a few settings
• As for motherboards... you're not planning on getting a 13900k as far as I can tell and you are trying to keep power down, so going with a somewhat basic board should be fine. I have the Asrock z790 Pro RS/D4, the standard Pro RS (for ddr5) is only a few dollars more than the 760 and 660/690 boards...
• Frankly with any of the processors you're talking about 4k is more than achievable, it's your graphics card that is going to be the limit, and also going to chew a lot of power/heat. When gaming on the overclocked 9900k, it only consumed about 40-80W most of the time, but under full load in productivity apps it was more like 185W. More modern chips will use even less power.

Hope this helps a bit.. it's a lot of words.

 

[Coolio]

Keljian Thank you so much mate for such a detailed and substantiated reply - highly appreciate your help! Need to think this over now...

 

[Keljian]

You’re welcome, for reference a lot of graphics cards use about 280-350w under load. Substantially more than a processor when gaming.

One other point.. all of the CPUs you listed come with integrated graphics and video encoding, even AV1 encoding. There is nothing stopping you using this while you're browsing stuff and leaving the rest of the cpu essentially idle..

 

[Keljian]

Keljian Thank you so much mate for such a detailed and substantiated reply - highly appreciate your help! Need to think this over now...

I found this - (skip to about 10m30s) it shows ram performance figures ddr4 vs ddr5, slow vs fast ddr 5 - result is the same - very little difference between all of them