just bought an x299 cpu+mobo, what kind of performance should i expect from this platform?

[VeganJoy]

saw a rampage vi extreme with monoblock and a 7920x together for 650 so i snagged that, love the board so i'm gonna replace my 10900k lmao. i don't know much about overclocking in general, let alone an hedt system, so i have no idea what to expect going into this. i've been setting up my first custom loop lately with 420mm + 280mm rads so i'm hoping that's a decent amount of cooling. a bit overkill for a strictly gaming system but i've had my eye on this board for a very long time and it just looks fabulous

 

q1: are skylake-x cores basically the same as normal sky/coffee/comet lake cores, just with a mesh topology instead of ring and a bunch of cache? what does this mean for gaming performance? i'm playing at 1440p so any cpu bottleneck shouldnt be the end of the world but i would like to avoid any high latency scenarios that cause stuttering and/or frame lag.

 

q2: aside from a soldered ihs, what kinda differences are there between the three generations of skylake-x parts? (7900x, 9900x, 10900x)

 

q3: aside from beefed up vrms/power delivery and some connectivity differences (m.2s, pcie slots, wifi/bt, etc), are there any other notable differences between the original r6e and the omega/encore? can i plug one of the newer dimm.2 modules with integrated heatsinks into the original r6e dimm.2 slot?

 

q4: does the massive memory bandwidth help in any sorts of gaming scenarios? i plan on getting some b die and overclocking, but i have no clue what frequencies will be supported by the chips, nor how many sticks/ranks to get. would you need 8 ranks of ram to take advantage of memory interleaving? would 4 vs 8 ranks even make a meaningful difference in bandwidth or should i go with 4 for lower latency?

 

q5: what cpu-related things should i keep in mind when overclocking? are voltages/clockspeeds similar to the non-hedt parts? will wattage be higher core-for-core at the same voltages, and does the massive die/ihs help with heat dissipation? 300w through a 10900k seems like itd be harder to cool than 400-500w through an lga2066 chip but idk

 

thanks for any help and tips that you can provide!

 

edit: this is pretty much my absolute dream board, like swapping out a ferrari for a 63 stingray corvette. yeah it's gonna be slower but it's gonna look 10x more badass and get me from a to b anyways

 

[Lurick]

You should expect a TON of heat output and power draw along with pretty gimped gaming performance compared to the 10900K. I mean this is a pretty large step backwards in terms of performance.

[Guest]

I... why did you.... how.... never mind.

I cannot ask enough questions to make me understand what just happened.

[The Blackhat]

Well first off you're going to take a rather substantial hit to gaming performance. 

Second, good thing it's just a 7920X since the original R6E kinda sucked.

Third, overclocking Sky-X could not be any more different from overclocking the mainstream parts, I implore you to do as many hours of reading and research as you possibly can to be educated on the topic before you ruin a somewhat decent CPU like the 7920X.

As for memory, the extra bandwidth is what helps make up for the handicap of having a MeshBus Sky-X part. Without quad channel, gaming would be even worse.

 

I also repeat word for word the post directly above mine.

 

[Guest]

5 minutes ago, VeganJoy said:

saw a rampage vi extreme with monoblock and a 7920x together for 650 so i snagged that, love the board so i'm gonna replace my 10900k lmao. i don't know much about overclocking in general, let alone an hedt system, so i have no idea what to expect going into this. i've been setting up my first custom loop lately with 420mm + 280mm rads so i'm hoping that's a decent amount of cooling. a bit overkill for a strictly gaming system but i've had my eye on this board for a very long time and it just looks fabulous

 

q1: are skylake-x cores basically the same as normal sky/coffee/comet lake cores, just with a mesh topology instead of ring and a bunch of cache? what does this mean for gaming performance? i'm playing at 1440p so any cpu bottleneck shouldnt be the end of the world but i would like to avoid any high latency scenarios that cause stuttering and/or frame lag.

 

It's a Hot son of a Beep. Probably won't see 5ghz on ambient without a really decent water loop. An 8700K is easily 10% faster by nature.

 

5 minutes ago, VeganJoy said:

q2: aside from a soldered ihs, what kinda differences are there between the three generations of skylake-x parts? (7900x, 9900x, 10900x)

Not a great deal in difference. You hit the nail on the head, cache overclocking is a thing. Ring is cache frequency.

 

5 minutes ago, VeganJoy said:

q3: aside from beefed up vrms/power delivery and some connectivity differences (m.2s, pcie slots, wifi/bt, etc), are there any other notable differences between the original r6e and the omega/encore? can i plug one of the newer dimm.2 modules with integrated heatsinks into the original r6e dimm.2 slot?

Sure.

5 minutes ago, VeganJoy said:

q4: does the massive memory bandwidth help in any sorts of gaming scenarios? i plan on getting some b die and overclocking, but i have no clue what frequencies will be supported by the chips, nor how many sticks/ranks to get. would you need 8 ranks of ram to take advantage of memory interleaving? would 4 vs 8 ranks even make a meaningful difference in bandwidth or should i go with 4 for lower latency?

Well, depends on how brave you are. I use mine with 8700K at 4000mhz+ CL-16 1.58v

5 minutes ago, VeganJoy said:

q5: what cpu-related things should i keep in mind when overclocking? are voltages/clockspeeds similar to the non-hedt parts? will wattage be higher core-for-core at the same voltages, and does the massive die/ihs help with heat dissipation? 300w through a 10900k seems like itd be harder to cool than 400-500w through an lga2066 chip but idk

 

 

Let's see if you can handle 1.35v or more.... then we'll talk   

 

[The Blackhat]

1 minute ago, ShrimpBrime said:

 

It's a Hot son of a Beep. Probably won't see 5ghz on ambient without a really decent water loop. An 8700K is easily 10% faster by nature.

 

Not a great deal in difference. You hit the nail on the head, cache overclocking is a thing. Ring is cache frequency.

 

Sure.

Well, depends on how brave you are. I use mine with 8700K at 4000mhz+ CL-16 1.58v

Let's see if you can handle 1.35v or more.... then we'll talk   

 

Cache on Sky-X isn't Ring, it's Mesh, and labelled as such typically.

[Guest]

Just now, The Blackhat said:

Cache on Sky-X isn't Ring, it's Mesh, and labelled as such typically.

Quote

just with a mesh topology instead of ring and a bunch of cache?

I was responding to this part of the question, not sure if he knew what ring was. 

[-rascal-]

The i7-7920X is clocked MUCH lower (2.9 GHz Base, up to 4.3 GHz Boost).

IPC Skylake is also (slightly) lower than Comet Lake.

If you are not going to do any overclocking, gaming performance will be DOWN from your i9-10900K.

 

Other then soldered IHS, refer to 1st answer, slightly lower IPC...

 

Not much difference.

 

No, the memory bandwidth does not help, or barely if at all; it only really helps with CPU intensive / heavy multi-core workloads.

Which is WHAT the HEDT platform(s) were designed for.

You don't 12-core / 24-thread CPU for gaming...generally.

 

Overclocking?

Similar to non-HEDT parts for the most part.

You'll need even MORE cooling than a i9-10900K....because 12-cores running at 4.5GHz+ will generate more heat than 10-cores running at 4.5GHz+.

You are not going to reach ~5.0 GHz with such a high core-count CPU...

 

 

IMO, I cannot wrap my head around WHY you would replace your i9-10900K with a i7-7920X...

[Guest]

2 minutes ago, -rascal- said:

 

IMO, I cannot wrap my head around WHY you would replace your i9-10900K with a i7-7920X...

lol. right. 

 

I'm an Enthusiast/hobbiest myself. I often ask myself the same question.

Earlier 2020, purchases Ageia Physx Card. PCI-E version because I only have legacy PCI version.

And it's sitting in a box, I've used it once lol.....

[The Blackhat]

1 minute ago, -rascal- said:

The i7-7920X is clocked MUCH lower (2.9 GHz Base, up to 4.3 GHz Boost).

IPC Skylake is also lower than Comet Lake.

If you are not going to do any overclocking, gaming performance will be DOWN from your i9-10900K.

 

Other then soldered IHS, refer to 1st answer, slightly lower IPC...

 

Not much difference.

 

No, the memory bandwidth does not help, or barely if at all; it only really helps with CPU intensive / heavy multi-core workloads.

Which is WHAT the HEDT platform(s) were designed for.

You don't 12-core / 24-thread CPU for gaming...

 

Overclocking?

Similar to non-HEDT parts for the most part.

You'll need even MORE cooling than a i9-10900K....because 12-cores running at 4.5GHz+ will generate more heat than 10-cores running at 4.5GHz+.

 

 

IMO, I cannot wrap my head around WHY you would replace your i9-10900K with a i7-7920X...

The clocks are not the main reason gaming performance takes a huge hit. Even overclocking is not going to bring it even close to a 10900K.

No, the IPC between all the Sky-X revisions are identical.

Yes, the bandwidth helps massively and is the only reason Sky-X isn't even further behind the mainstream parts.

No, the overclocking is completely different and involves totally different processes and methodology due to the necessity of cache OC and uncore voltage modifications, as well as VCCIN.

As for heat, core for core, a 10900K is harder to cool due to core density. But the 7920X just plain chugs much more power. 

 

[-rascal-]

7 minutes ago, ShrimpBrime said:

lol. right. 

 

I'm an Enthusiast/hobbiest myself. I often ask myself the same question.

Earlier 2020, purchases Ageia Physx Card. PCI-E version because I only have legacy PCI version.

And it's sitting in a box, I've used it once lol.....

 

Ahaha fair enough.

 

I was kind of like that too...cuz m0ar cores is better.

I went from a 4-core Phenom II X4 955, to 6-core Pheneom II 1090T, to a FX-8350 ...cuz 4 < 6 < 8.

Was reluctant to go from a "8-core" FX-8350 down to a 6-core i7-6800K ... in my mind, that's going a step backwards (you can only go up lol.

6-core i7-6800K to a "side-grade" 6-core i7-8086K...was tempting to actually go HEDT, cuz gotta be 8+ core to keep the legacy going

 

Can't be weak forever 

[VeganJoy]

so yeah maybe i shouldve put a disclaimer that this is my dream board lmao, it could run as slow as my old r5 3600 and i'd be happy. the only high fps game i play is league of legends and i presume that a 7920x can keep me above 165fps

 

besides all the rgb bling on this board will give me +420% fps over my 10900k so i'll be fine...

 

20 minutes ago, Den-Fi said:

I... why did you.... how.... never mind.

I cannot ask enough questions to make me understand what just happened.

i figured you'd be perplexed lmao, but this is like half your fault lol. if my build looks even 1% as good as your x299 build i'll be ecstatic haha.

 

17 minutes ago, ShrimpBrime said:

Well, depends on how brave you are. I use mine with 8700K at 4000mhz+ CL-16 1.58v

and i was thinking 1.45v for my b die kit might be a bit scary, how long have you been running that voltage?

 

9 minutes ago, The Blackhat said:

No, the IPC between all the Sky-X revisions are identical.

figured, confuses me as to why a 10900x sells for twice the price of a 7900x but oh well lol

 

9 minutes ago, The Blackhat said:

No, the overclocking is completely different and involves totally different processes and methodology due to the necessity of cache OC and uncore voltage modifications, as well as VCCIN.

is there a basic guide available like the memtesthelper guide for ddr4 or like buildzoid's and derbauer's skylake guides? speaking of, is memory overclocking basically the same for skylake-x? capacity is obviously gonna be absurd so should i go with 4 or 8 sticks?

 

9 minutes ago, The Blackhat said:

As for heat, core for core, a 10900K is harder to cool due to core density. But the 7920X just plain chugs much more power. 

i'm assuming the large amounts of cache and mesh topology are the main contributors here to this? it's coming with an ek monoblock which should help the vrms out, but it has a circular coldplate that doesnt cover the whole ihs? and the cpu's been delidded in the past which i've heard is bad for monoblock compatibility, but i dont think rockitcool still makes their height-adjusted ihs anymore. i'm not opposed to swapping the chip for a 10900x or something like that for compatibility with the block

 

 

[Guest]

2 minutes ago, -rascal- said:

 

Ahaha fair enough.

 

I was kind of like that too...cuz m0ar cores is better.

I went from a 4-core Phenom II X4 955, to 6-core Pheneom II 1090T, to a FX-8350 ...cuz 4 < 6 < 8.

Was reluctant to go from a "8-core" FX-8350 down to a 6-core i7-6800K ... in my mind, that's going a step backwards (you can only go up lol).

6-core i7-6800K to a "side-grade" 6-core i7-8086K (was tempting to actually go HEDT, cuz gotta be 8+ core to keep the legacy going)

And the 6800K completely KILLS the poor FX 8 core which your Phenom II Hexa was just as fast in Multi-threading and faster on single core.

However, I had quite a few AMD Phenom I/II and FX processors. Super fun platform for sub zero OC. Cold and Volts, they clock like mad. I had gamed several times in the 6ghz and 7ghz range respectively. Mostly went in for the max Cpu frequency. 

My 980BE, wish I still had it. Damn thing hit nearly 6.4ghz man  

 

 

[The Blackhat]

4 minutes ago, VeganJoy said:

so yeah maybe i shouldve put a disclaimer that this is my dream board lmao, it could run as slow as my old r5 3600 and i'd be happy. the only high fps game i play is league of legends and i presume that a 7920x can keep me above 165fps

 

besides all the rgb bling on this board will give me +420% fps over my 10900k so like

 

i figured you'd be perplexed lmao, but this is like half your fault lol. if my build looks even 1% as good as your x299 build i'll be ecstatic haha.

 

and i was thinking 1.45v for my b die kit might be a bit scary, how long have you been running that voltage?

 

figured, confuses me as to why a 10900x sells for twice the price of a 7900x but oh well lol

 

is there a basic guide available like the memtesthelper guide for ddr4 or like buildzoid's and derbauer's skylake guides? speaking of, is memory overclocking basically the same for skylake-x? capacity is obviously gonna be absurd so should i go with 4 or 8 sticks?

 

i'm assuming the large amounts of cache and mesh topology are the main contributors here to this? it's coming with an ek monoblock which should help the vrms out, but it has a circular coldplate that doesnt cover the whole ihs? and the cpu's been delidded in the past which i've heard is bad for monoblock compatibility, but i dont think rockitcool still makes their height-adjusted ihs anymore. i'm not opposed to swapping the chip for a 10900x or something like that for compatibility with the block

 

 

More than likely slower than a 3600 without a very, very lucky overclock. It is a massive step back from a 10900K and massive is definitely not an understatement. The delid shouldn't mess with your monoblock too badly. It's an EK monoblock, either way it's not gonna perform great, the minor distance between the IHS and the block can be remedied with thinner thermal pads on the VRMs if you have problems.

As for mem OC on Sky-X, I do not know of any beginner guides on the subject, but I also never bothered to look since I was experienced with memory overclocking on other platforms. 4 sticks will be your best bet.

[Guest]

6 minutes ago, VeganJoy said:

 

and i was thinking 1.45v for my b die kit might be a bit scary, how long have you been running that voltage?

 

1.45v is an XMP standard of 4267-4400mhz B-Die sticks and higher frequency at 1.50v.

I've had mine as high as 1.9v on attempt to run CL12 at 4000mhz or higher. The breaking point on ambient is 1.9v-2.0v

My 3600mhz kit has been running months at 1.60v, 4000mhz 16-16-16-36 for maybe 9-10 months with some frequencies and timings exceeding this.

 

The 4267mhz kit I have is A2 PCB 19-19-19-39 at 1.40v XMP rating. This kit does 4200mhz CL16 1.610v.

Have benched 4300mhz, but the IMC looses it's stability without pumping mad Vccio and System Agent voltage. 

 

On 8700K (I should mention)

Edited January 22, 2021 by ShrimpBrime
Additional Comment

[VeganJoy]

1 minute ago, The Blackhat said:

More than likely slower than a 3600 without a very, very lucky overclock. It is a massive step back from a 10900K and massive is definitely not an understatement. The delid shouldn't mess with your monoblock too badly. It's an EK monoblock, either way it's not gonna perform great, the minor distance between the IHS and the block can be remedied with thinner thermal pads on the VRMs if you have problems.

i took a quick look at some gaming benchmarks between a 10900x and a 10900k a while back and there was like a 1% difference in framerates, both average and lows? pretty sure it was a 2080ti at 1080p which is probably the same amount of bottleneck i'd have with a 3080@1440p assuming i can get my hands on one...

 

2 minutes ago, ShrimpBrime said:

1.45v is an XMP standard of 4267-4400mhz B-Die sticks and higher frequency at 1.50v.

I've had mine as high as 1.9v on attempt to run CL12 at 4000mhz or higher. The breaking point on ambient is 1.9v-2.0v

My 3600mhz kit has been running months at 1.60v, 4000mhz 16-16-16-36 for maybe 9-10 months with some frequencies and timings exceeding this.

 

The 4267mhz kit I have is A2 PCB 19-19-19-39 at 1.40v XMP rating. This kit does 4200mhz CL16 1.610v.

Have benched 4300mhz, but the IMC looses it's stability without pumping mad Vccio and System Agent voltage. 

 

On 8700K (I should mention)

jesus i know bench voltages arent too damaging if you only bench with them but 1.6v for daily use? and that overclock hasnt degraded or anything?

[The Blackhat]

1 minute ago, VeganJoy said:

i took a quick look at some gaming benchmarks between a 10900x and a 10900k a while back and there was like a 1% difference in framerates, both average and lows? pretty sure it was a 2080ti at 1080p which is probably the same amount of bottleneck i'd have with a 3080@1440p assuming i can get my hands on one...

 

jesus i know bench voltages arent too damaging if you only bench with them but 1.6v for daily use? and that overclock hasnt degraded or anything?

What benchmarks were these? Typically people watch videos from random YouTube videos like "testing games" or "benchmark" and take them as fact, they aren't.

[VeganJoy]

Just now, The Blackhat said:

What benchmarks were these? Typically people watch videos from random YouTube videos like "testing games" or "benchmark" and take them as fact, they aren't.

i thought "testing games" was one of the few good youtube benchmark sites? i know many just regurgitate other publications' numbers but "testing games" always has frametime graphs and all that jazz so i'd figured they were at least a little accurate?

[Guest]

Just now, VeganJoy said:

i took a quick look at some gaming benchmarks between a 10900x and a 10900k a while back and there was like a 1% difference in framerates, both average and lows? pretty sure it was a 2080ti at 1080p which is probably the same amount of bottleneck i'd have with a 3080@1440p assuming i can get my hands on one...

 

jesus i know bench voltages arent too damaging if you only bench with them but 1.6v for daily use? and that overclock hasnt degraded or anything?

I don't have any troubles no. But I have higher end equipment that can handle it. ROG boards, like my B450-I, really has absolutely zero issues with most any requests I ask of it. The 2700X with a locked IF/Mem ration 1:1 has reached and benched at 3933mhz. Passes stability testing at 3733mhz. Acommpanied Cpu is a 2700X. It loves 3600mhz mem frequency, that much I can say. 

 

Frame rates and benches will always vary. But without some tweaking on the timings, sure results will be minimal. 

 

Memory bandwidth will also accompany a Cpu overclock. Always keep that in mind also. You should be able to measure this with AIDA64 memory and cache benchmark. Watch the Cpu freqeuncy increase push the memory bandwidth higher even if you keep the memory frequency the same between benchmarks.

[The Blackhat]

Just now, VeganJoy said:

i thought "testing games" was one of the few good youtube benchmark sites? i know many just regurgitate other publications' numbers but "testing games" always has frametime graphs and all that jazz so i'd figured they were at least a little accurate?

Not in the slightest. On numerous occasions they have been shown to just either falsify numbers or take numbers from other publications. But I also raise you this question, why would you trust any publication that doesn't even specify their full methodology? For any sort of testing, I mean. For coolers, GPUs, CPUs, anything, you should always know the methodology behind the testing.

[VeganJoy]

3 minutes ago, ShrimpBrime said:

The 2700X with a locked IF/Mem ration 1:1 has reached and benched at 3933mhz. Passes stability testing at 3733mhz.

thats just plain insane, whats your vddg/soc voltages??

 

4 minutes ago, The Blackhat said:

Not in the slightest. On numerous occasions they have been shown to just either falsify numbers or take numbers from other publications. But I also raise you this question, why would you trust any publication that doesn't even specify their full methodology? For any sort of testing, I mean. For coolers, GPUs, CPUs, anything, you should always know the methodology behind the testing.

that is 100% true, i guess i should look to gamers nexus and hwunboxed for some old benchmark roundups...

[The Blackhat]

1 minute ago, VeganJoy said:

thats just plain insane, whats your vddg/soc voltages??

 

that is 100% true, i guess i should look to gamers nexus and hwunboxed for some old benchmark roundups...

An example of the gaming hit of Sky-X is HWUnboxed's video on the 7800X vs the Ryzen 5 1600. While they did the testing at stock on the 7800X, and I am aware there is a large difference between a 7920X and a 7800X, the video is a good demonstration of Sky-X gaming performance core for core. It even loses in some situations to the 1600.

[Guest]

1 minute ago, VeganJoy said:

thats just plain insane, whats your vddg/soc voltages??

 

Vddg on auto actually and SOC 1.200v. More SOC doesn't help. Just creates heat.

 

B-Die likes a good kick in the nuts. Stability is found when keeping the memory at 40c or less. Can't let em' run warm. 50c+ is too much.

[-rascal-]

35 minutes ago, The Blackhat said:

The clocks are not the main reason gaming performance takes a huge hit. Even overclocking is not going to bring it even close to a 10900K.

No, the IPC between all the Sky-X revisions are identical.

Yes, the bandwidth helps massively and is the only reason Sky-X isn't even further behind the mainstream parts.

No, the overclocking is completely different and involves totally different processes and methodology due to the necessity of cache OC and uncore voltage modifications, as well as VCCIN.

As for heat, core for core, a 10900K is harder to cool due to core density. But the 7920X just plain chugs much more power. 

 

 

I thought there was a ~3% IPC gain from going from Skylake to Comet Lake.

My bad, that gain was actually from Haswell/Broadwell up to Skylake.

 

But that's what I'm saying, even if overclocking the i7-7920X to as high as it will go, it won't match the performance of the i9-10900K.

Even more so in gaming.

 

Well I don't know how other people overclock their CPUs, but I still tweak and adjust the Cache voltage / CPU Core PLL / Ring PLL, etc,  even with my i7-8086K.

Somewhat similar like how I did it with my i7-6800K on X99 / LGA 2011-v3.

I don't JUST mess with DRAM voltage, CPU Core Voltage, VCCIO, VCCSA, and call it a day...

 

26 minutes ago, ShrimpBrime said:

And the 6800K completely KILLS the poor FX 8 core which your Phenom II Hexa was just as fast in Multi-threading and faster on single core.

However, I had quite a few AMD Phenom I/II and FX processors. Super fun platform for sub zero OC. Cold and Volts, they clock like mad. I had gamed several times in the 6ghz and 7ghz range respectively. Mostly went in for the max Cpu frequency. 

My 980BE, wish I still had it. Damn thing hit nearly 6.4ghz man  

 

 

 

I quickly realized the FX-8350 was pretty junky.

But I was able to do 5.1 ~ 5.2 GHz on that chip! And on air!

5.1 GHz was my daily use OC.

Before 5.0 +GHz was mainstream 

 

I don't know...still haven't gotten rid of my Phenom II 1090T, FX-8350, and ASUS Crosshair V Formula board yet...

Still sitting across the room in the corner, collecting dust + webs.

[porina]

40 minutes ago, VeganJoy said:

q1: are skylake-x cores basically the same as normal sky/coffee/comet lake cores, just with a mesh topology instead of ring and a bunch of cache? what does this mean for gaming performance? i'm playing at 1440p so any cpu bottleneck shouldnt be the end of the world but i would like to avoid any high latency scenarios that cause stuttering and/or frame lag.

In execution cores, essentially the same apart from the addition of AVx-512 support. Looking wider, the cache difference is a major factor, and unfortunately not one that is positive for consumer use cases.

 

Before I continue, I currently own and use a 7920X, as well as 8086k/10600k. All three are in gaming systems but they have very different focuses.

 

40 minutes ago, VeganJoy said:

q2: aside from a soldered ihs, what kinda differences are there between the three generations of skylake-x parts? (7900x, 9900x, 10900x)

Cascade Lake-X I think had even more instructions added as part of AVX-512, but irrelevant for gaming.

 

40 minutes ago, VeganJoy said:

q4: does the massive memory bandwidth help in any sorts of gaming scenarios? i plan on getting some b die and overclocking, but i have no clue what frequencies will be supported by the chips, nor how many sticks/ranks to get. would you need 8 ranks of ram to take advantage of memory interleaving? would 4 vs 8 ranks even make a meaningful difference in bandwidth or should i go with 4 for lower latency?

Ram bandwidth certainly affects performance, even in gaming, but how much depends on title and other factors. Dual rank per channel is always good to aim for. I'm running 8x8GB in mine for example.

 

40 minutes ago, VeganJoy said:

q5: what cpu-related things should i keep in mind when overclocking? are voltages/clockspeeds similar to the non-hedt parts? will wattage be higher core-for-core at the same voltages, and does the massive die/ihs help with heat dissipation? 300w through a 10900k seems like itd be harder to cool than 400-500w through an lga2066 chip but idk

Without delidding+LM I found Skylake-X did relatively well compared to mainstream parts. Power running non-AVX-512 code isn't that much different from comparable generation. Note Coffee Lake and newer did have process optimisations to ease power efficiency at higher clocks.

 

AVX-512 code is on another level when it comes to both performance and power consumption, but from a gaming perspective, irrelevant.

 

 

 

I've only OC'd my 7800X which I don't have any more. It was within 100 MHz of Kaby Lake, which is the nearest comparable process generation mainstream CPU. Don't necessarily expect 5 GHz, especially with all cores going. Even with a ton of rads, it will run hot. I haven't attempted to OC the 7920X cores yet since it is horrifically ram bandwidth limited for my use cases. Quad channels were barely enough to feed the 6 core 7800X running AVX-512.

 

What I have found can improve gaming performance is to disable Hyper-Threading. If you can overclock the cache that can help too but I found my 7920X responds much worse than my 7800X did.