Kaleid:

170w is OK, unlike modern GPUs this is avoidable because most have absolutely no need for 16 cores at all.

Ryu5uzaku:

If they manage to increase the size of the heatspreader and what not it should be better. Yes the 5xxx series runs hot even with quite low wattage. But also the surface area is small.

Mineria:

Offsetting the cooling block to align it with the die helps (10-15C), besides that choice of TIM and heatspreder material matters as well (5-10C with copper).

tunejunky:

the deal with node shrinkage is a smaller node equals higher performance or lower TDP. as in either/or when AMD introduced ryzen they could have both higher performance and lower TDP because they were three nodes smaller which continued until AL introduced e cores which dropped their (still high) TDP Ryzen 4 will be two nodes smaller than AL and the performance differential should be similar ( 7n vs 5n ) but the power envelope has increased because this gen AMD's uArch has stiff competition unlike last.

TheDeeGee:

You don't need a RTX 3090 Ti either, so even the GPU side can be avoided. Stick with 1440p and all you need is a 3070.

JamesSneed:

Its very likely going to be a thin bin part like intel's KS parts like the 12900ks. I expect AMD to have one chart topper thin binned/special edition whatever you want to call the low volume and high priced halo part.

GamerNerves:

Either that value is for short period clock boost or then Intel is offering tough competition, so AMD has decided to crank up clocks. This indeed might indicate that they are offering more performance to customers (without OC) than they originally planned to. Is Intel going to care about power consumption then? I'm sure they would like to, but it is hard to believe they could improve enough from Alder Lake to really be more energy efficient than Zen4 on high core counts. Alder Lake's IPC is strong though, so they just might win the "gaming processor" battle this time, it's going to be a close call.

tunejunky:

you have good reasoning and i disagree on some points. Zen 4 will crush AL seven ways from Sunday. AL is competitive with Zen 3, Meteor Lake is competitive with Zen 4 ML is Intel's internal pride and joy, the world will shortly see why and it's being fabbed at TSMC 7n AL was just a stop gap proof-of-concept that Mr. Gelsinger knew would "calm the waters" and restore Intel's uArch credibility. if Intel win's "top gaming processor" it won't be AL holding on, it will be Meteor Lake

tunejunky:

you have good reasoning and i disagree on some points. Zen 4 will crush AL seven ways from Sunday. AL is competitive with Zen 3, Meteor Lake is competitive with Zen 4 ML is Intel's internal pride and joy, the world will shortly see why and it's being fabbed at TSMC 7n AL was just a stop gap proof-of-concept that Mr. Gelsinger knew would "calm the waters" and restore Intel's uArch credibility. if Intel win's "top gaming processor" it won't be AL holding on, it will be Meteor Lake

XenthorX:

5900X with slight overclock are already at 190W+ for multi thread workload

JamesSneed:

You can also buy a RTX 3090 Ti and undervlot/underclock it with impressive results. I set some lower voltage limits on PBO for my 5800x. I get very similar performance but much less heat. I like very quiet PC's so I end up tuning things like this vs for performance even If I have to overpay a bit.

JamesSneed:

Its only rumored the graphics tile aka GPU would be using TSMC and the timing it would be TSMC's 3nm. The Meteor Lake CPU tile is being fabed on Intel 4 this will not ever go to TSMC.

PrMinisterGR:

Did something similar with my 3090. I undervolted various steps on the curve itself (I have the MSI Afterburner config saved directly). Even with watercooling, I went from ~1700-1800MHz without memory OC, to ~1900-2000MHz + 800MHz memory (21GBps). This was only from the undervolting on the curve points. I did the same, slowly with my CPU, and it was completely worth it.

tunejunky:

au contraire mon frere Intel will not have access to TSMC 3n until mid 2024. 3n is exclusively Apple until then Intel will be fabbing @ TSMC 7n & 5n according to contract

JamesSneed:

Not on water but I did that as well. Have my two good cores(the ones that boost high) running stock and all the other cores are rolling with -25.

PrMinisterGR:

Not all my cores could do -25, Civilization 6 is a good test for this, as it will max and hang cores that are way too undervolted, something that Prime 95 also does. I tried to leave the fastest cute of each CCX at default, but I just lost performance in the end. Now every core is undervolted, but with increased watt limits, and I regularly turbo at 4,8 multithreaded. Single still hits 5 GHz without clock stretching, and I don't have any weird hangs out WHAE errors or anything. I wish that AGESA 1.2.0.7 was here sooner.

JamesSneed:

Not on water but I did that as well. Have my two good cores(the ones that boost high) running stock and all the other cores are rolling with -25. This is t Are you sure? Intel has reportedly secured the majority of TSMC's 3nm production capacity | TechSpot Maybe its 5nm for the GPU for meteor lake due to timing? It sounds like TSMC will build out a 3nm fab for Intel but that is likely in 2024. Anyhow my main point was TSMC will only be used for the GPU tile not the CPU. The CPU will be made on the Intel 4 node.

JamesSneed:

I suspect based off what we know already that 5nm custom node created for AMD can hit some nice frequencies over 5Ghz. This is a departure from the current 7nm process. I expect this 170w part to be a high clocked version of a 16 core part. If you look at some of the leaks from those who know the 8-core parts are all 65 watt parts.

Pryme:

170W was a typical FX with a slight OC.