"The Fastest Gaming CPU in the World"

[WihGlah]

Any review that doesn't remove the power limits for all chips and use the best RAM available is pointless.

 

The entire purpose of the 3D cache is to make the RAM interface bottleneck less of a problem...

 

Based on those results, my 9900K is faster.

[papajo]

8 minutes ago, WihGlah said:

The entire purpose of the 3D cache is to make the RAM interface bottleneck less of a problem...

 

The key word here is "less" it still is a problem because chache can only be "that big" or other games may not fill it up anyways so giving an advantage to the system with the faster RAM. 


RAM used to be a none issue years ago since the difference between modules was marginal and the workloads more forgiving/smaller/simpler but now its relevant again since the speed of modules can vary hugely even for the same technology (e.g DDR4) and programs use more ram and more frequently. 

E.g a big reason M1 chips had such good benchmarks is the fact that their RAM  is at 4200 Mhz or even higher (depending on the model)  and were compared with laptops that run 3200Mhz or slower so dims with high latency. Not even counting for the latency difference (which is important) we are talking about a 1+Ghz difference in speed here... 

[ryao]

3 hours ago, WihGlah said:

So you used 3200MHz ram and left the 12900K at stock speeds. Utterly useless review.

 

The 12900KS is a factory overclocked 12900K.

[papajo]

For example (to add more to my argument above ) Using faster memory (on both the 3800x3d and the 1200KS) adds +10 FPS in most games and they did not use the fastest memory available just different modules with different speeds where the faster module gave better results, mind that latency here was sadly also ignored. while latency is a big factor it can make the difference in speed latency and hz for ram is like g2g response times and monitor frequency for monitors both need to be low for the best result and if response time is high then it renders a high frequency monitor useless (because of ghosting overshooting etc) same with ram it doesnt get useless though but it would negate a speed difference in mhz if the latency difference is big. 

It also paints a slightly different (and more favorable) picture for the 3800x3d when you use faster ram. 
 

 

[WihGlah]

17 minutes ago, ryao said:

The 12900KS is a factory overclocked 12900K.

Sure it is... because it runs 5.3GHz all core out of he box 24/7...

 

Like they ALL can.

[ryao]

1 hour ago, papajo said:

So a question, you wanted to use the "best DDR5" RAM for intel (10ns 5600cl40) 

But you used a subpar DDR ram for ryzen? Why not the best DDR4 as well e.g the tridentZ 4000Mh/z cl 14 or at least the 3600mhz cl14 ? 


Also I dont want to shame you or anything I have buggers too it is natural and happens to all of us... but could somebody please when you shoot a video just pause it for Anthony (or who ever he just happened to be the one in this video)  to clean his boogers real quick? It creates a gagging effect for me and makes watching the video less enjoyable...

Again nothing to be ashamed I have buggers too etc but ok since you give so much money for good video production and care about the end result to look professional a few seconds for cleaning up one's nose prior to shooting a video shouldnt be that much to ask. 

For what it is worth, the increased cache would lessen the benefits of overclocked memory. AMD also specifies 3200MHz memory:

 

https://www.amd.com/en/products/cpu/amd-ryzen-7-5800x3d

 

while Intel specifies 4800MHz memory:

 

https://www.intel.com/content/www/us/en/products/sku/225916/intel-core-i912900ks-processor-30m-cache-up-to-5-50-ghz/specifications.html

 

They already went overboard with memory clocks for both of them, although in Intel’s case, the memory clock speed was within JEDEC’s specification, while in AMD’s case, the memory was over locked beyond JEDEC’s specification.

[papajo]

10 minutes ago, ryao said:

For what it is worth, the increased cache would lessen the benefits of overclocked memory. AMD also specifies 3200MHz memory:

 

https://www.amd.com/en/products/cpu/amd-ryzen-7-5800x3d

Again as I said before keyword here is less the benefits are still there having bigger and faster cache just decreases a still existing  bottleneck 

 

12 minutes ago, ryao said:

while Intel specifies 4800MHz memory:

 

That's megatransfers (MT) not the same, latency in addition with frequency determine that (and that's the point of choosing lower latency even if the frequency is the same) 

 

and that's the default spec nobody said you cant use faster than this. 

 

13 minutes ago, ryao said:

They already went overboard with memory clocks for both of them

No they did not, and most importantly the did something asymetric  so the advantage from the base frequency/latency was bigger in intel than it was in amd. 

 

Last but not least again it is not only about frequency but about latency as well.  5600Mhz at 40-40-40ms latency makes up for a 10ns response wich is the best there is for DDR5 as far as I am aware of. 

A 3600 mhz cl 16-18-18 is a 8ns-10-10 ns response which is mediocre for a DDR4 e.g the best RAM out there in terms of response time is DDR4 4000 14-14-14 which yields a 7 ns response time while still having a high frequency 

Also even if we forget the speed the memory used has a significant price difference if they did not care about the speed (which they obviously did at least for the intel side otherwise why use DDR5 ) they could at least use DDR4 and DDR5 modules of comparable price since a 3600 cl 16-18-18 kit is dirt cheap a 16GB kit with those specs goes at about 100$ or maybe even less (in europe you can find them well under 100 euro) 

While a cl 40-40-40 5600Mhz DDR5 16GB costs about $300+ usd...  

[Ultraforce]

2 hours ago, WihGlah said:

Any review that doesn't remove the power limits for all chips and use the best RAM available is pointless.

I don't know too much about CPUs and RAM but that just can't be a correct statement. I'm fairly certain most users don't artificially remove power limits for all of the chips for a variety of reasons and probably don't necessarily buy the best possible RAM so having reviews which deal with something closer to what someone uses in the real world probably makes sense sometimes. Maybe it doesn't make sense for an LTT video but there's definitely a place for reviews where ignoring power draw is not done.

[papajo]

1 hour ago, Ultraforce said:

I don't know too much about CPUs and RAM but that just can't be a correct statement. I'm fairly certain most users don't artificially remove power limits for all of the chips for a variety of reasons and probably don't necessarily buy the best possible RAM so having reviews which deal with something closer to what someone uses in the real world probably makes sense sometimes. Maybe it doesn't make sense for an LTT video but there's definitely a place for reviews where ignoring power draw is not done.

If they dont buy the best available ram then why use a $300+ DDR5 ram kit for the 1200KS (a 800$ cpu) instead of using a cheap DDR4 kit  (like they did with the ryzen sytem) ? 

You can't have it both ways. 

Want it cheap and more realistic? Agreed.  Then use  about the same budget on both systems

Want to compare each system to its peak performance ? Then use the best parts* for both systems not only for one. (I also wonder if the 3800x3D would have increased performance if besides with faster memory it would be paired with an AMD 6900 xt with SAM enabled or at least with a nvidia card with resizable bar enabled ) 



* This is an other issue I see often in tech reviews not only by ltt, best parts do not necessarily mean "most expensive"  (although in this day and age they would probably be quite expensive albeit not the most expensive) it means to have the best specs (actually better best real performance cause sometimes specs are misleading or nominal) for the particular test scenario. That's especially true for ram kits I mean ok the most expensive ones are going to be faster than an average cheap kit but its not a given that it would be faster than a cheaper kit.

[Ultraforce]

7 minutes ago, papajo said:

If they dont buy the best available ram then why use a $300+ DDR5 ram kit for the 1200KS (a 800$ cpu) instead of using a cheap DDR4 kit  (like they did with the ryzen sytem) ? 

You can't have it both ways. 

Want it cheap and more realistic? Agreed.  Then use  about the same budget on both systems

Want to compare each system to its peak performance ? Then use the best parts for both systems not only for one. (I also wonder if the 3800x3D would have increased performance if besides with faster memory it would be paired with an AMD 6900 xt with SAM enabled or at least with a nvidia card with resizable bar enabled ) 



 

That is why I was responding to With rather than you, I can understand that maybe they didn't make the right choices for this specific review, but feel the need to refute the claim that it is always improper to be testing at anything else other than voiding one's warranty by maximizing speed.

[Dogzilla07]

12 hours ago, starsmine said:

Yes the 12700F, when not "running stock" is a good value part. 

12 hours ago, starsmine said:

Going from a 2600X to a 5800x3D is far bigger bang for the buck then completely swapping platform, and that's who most of the buyers for the 5800x3D

12 hours ago, Jeppes said:

And why on earth would cripple the 12700F by not unlocking power limits and not using a good b660-board?

Not exactly, neither is 12700F unlocked the best value for new builds, neither is the 5800x3D the best bang for buck for previous AMD holders, 12700F power unlocked is close, but is still not the best value, nor the best bang for buck.

 

The current best value, best bang for buck is irrefutably the 12400F combined with the Asus Strix B660-G and fast-in-price-dropping 2x8GB of DDR5. The combo costs less than 5800x3D + cheapest board + cheapest comparable RAM. Why do you think board is so hard to get in the US ? 

 

And as for 12700F compared to 12400F combo, they are very close, but at least in Europe the 12400F combo comes down 17 euros cheaper (not counting coolers). As can been seen on the EU's version of pcpartpicker on the link below. And yes a 12700F overclocked to ~5.4GHz is a very good option as well (but the 12400F has a higher chance of reaching those clocks, due to being a different die) The US pricing should be similar, slightly higher for nearest comparable MOBO.

 

177 + 202 + 126 = 505 vs 332 + 123 + 67 = 522

 

https://geizhals.eu/?cmp=2659495&cmp=2660688&cmp=2684937&cmp=2659491&cmp=2661685&cmp=1814725&active=0

 

However even though, this is the best absolute option, and what I would do, and I think a much larger number of people are doing that you guys think (because of the easy to follow Der8auer video guide, and dozens of topic pages on hardwareluxx, and a huge presence on reddit) I do understand the appeal for unlocked 12700F on a $140 MSI PRO B660-A and similar boards with 2x8GB DDR4 (it's a completely valid option). But I don't agree with it, as overclocking the 12400F/12700F is extremely simple and straightforward due to the video guide, and games still don't benefit from the extra cores.

[dragn]

how did you test the flight sim benchmark? did you land the plane or just let it fly until the mission ends on its own?

curious why you only get 80avg fps while i with my oc'ed 12900k and ddr5 get 134 avg fps when landing and 160 when letting the plane fly on its own for 8 minutes. thats with a 3080

[CoolJosh3k]

Is it reasonable to expect this new X3D CPU to work with DRR-3800 with the infinity fabric set for a 1:1 ratio?

would it have to be DDR-3600? Or is even that a risk?

[papajo]

7 hours ago, dragn said:

how did you test the flight sim benchmark? did you land the plane or just let it fly until the mission ends on its own?

curious why you only get 80avg fps while i with my oc'ed 12900k and ddr5 get 134 avg fps when landing and 160 when letting the plane fly on its own for 8 minutes. thats with a 3080

I presume they land the plane since it mentions "Sydney landing challenge" although I dont play that game so I cant vouch for that but from the title it seems reasonable to assume that. 

As for the different performance at what resolution and settings do you get those framerates? 

Because they used very high and 1080p which would probably be CPU bound at this resolution which makes sense if you want to try the CPU performance. 

 

Also what are the specs of your DDR5 ram ? 

 

 

[C0d3dSk1llS]

once again, AMD got owned.

[dragn]

1 hour ago, papajo said:

I presume they land the plane since it mentions "Sydney landing challenge" although I dont play that game so I cant vouch for that but from the title it seems reasonable to assume that. 

As for the different performance at what resolution and settings do you get those framerates? 

Because they used very high and 1080p which would probably be CPU bound at this resolution which makes sense if you want to try the CPU performance. 

 

Also what are the specs of your DDR5 ram ? 

 

 

yes its called landing challenge, but thats just a mission instead of playing online. so they use it to have the same scenario every time without variables like weather or location. just want to know what they actually do in the benchmark so i can replicate it.

 

its in 1080p on "high-end" preset, which is the one between medium and ultra

ram is just 6000cl32 with lowered subs and all that. with a 2dimm ram oc mainboard i could up that to 7000mhz

[papajo]

47 minutes ago, dragn said:

yes its called landing challenge, but thats just a mission instead of playing online. so they use it to have the same scenario every time without variables like weather or location. just want to know what they actually do in the benchmark so i can replicate it.

 

its in 1080p on "high-end" preset, which is the one between medium and ultra

ram is just 6000cl32 with lowered subs and all that. with a 2dimm ram oc mainboard i could up that to 7000mhz

So you have

A) lower settings

B) play it online (I presume assets or asset rendering at least for online gaming are less demanding to maintain a good connection and latency) 

C) Have faster ram than what ltt used

D) have overclocked your CPU

 

No wonder you are seeing more FPS. 

[dragn]

2 hours ago, papajo said:

So you have

A) lower settings

B) play it online (I presume assets or asset rendering at least for online gaming are less demanding to maintain a good connection and latency) 

C) Have faster ram than what ltt used

D) have overclocked your CPU

 

No wonder you are seeing more FPS. 

why lower settings? i guess they just wrote it wrong and ment highend, dont think they mean ultra when saying very high.

im not online and used the same landing challenge which is offline only

 

also the 12700k was faster in some games so maybe the 12900k even was throttling? 

 

 

[papajo]

1 hour ago, dragn said:

why lower settings? i guess they just wrote it wrong and ment highend, dont think they mean ultra when saying very high.

im not online and used the same landing challenge which is offline only

 

also the 12700k was faster in some games so maybe the 12900k even was throttling? 

 

 

I think they mean ultra by "very high" or at least a mixture between "high end" and "ultra" 

Because when you go to the global settings and set it to high end all settings are labeled as "high"  and not "very high" 

So I suppose they mean a step above "high" 

So either they chose ultra as a global preset or they have a custom mixture of settings with "high" being the lowest setting in each option and some at ultra. 

[Erdwin]

Yall have a scammer on your youtube page.

[Lightwreather]

6 hours ago, Erdwin said:

-snip-

That happens a lot.

Nothing to do except report and move on

[Erdwin]

13 hours ago, J-from-Nucleon said:

That happens a lot.

Nothing to do except report and move on

They never do anything about it when you report it so im having fun with this! hahahaha They send you to telegram because when you msg them if you dont read what pops up you wont see that it is asking you to also send your phone number to them. I denied it so  there wanting more information! There not gonna like it when they call the number i give them! lol

[chrome_wheels]

This is one great processor from AMD