FabioMy 8700k run quite fresh at 4930 Mhz, 1,28 vcore, that run up to 1.31 when under avx workload. I have an aio nzxt. But i think that a good air cooler will work well too.
Single td performance still quite the same as a 10700k... 580 in cpuz and 218 chinebench 20 single

Vayra861.28 is very good for that clock.

I wasnt that lucky...

FabioThis is the very limit. After that 4920 i need to pump 1.34-1,36 to run stable at 5 ghz. Temp in full load under stress went from 76-78 max to 85-86. Really does not wort it...

BubsterIntel Headquarters Parking Lot is gonna be full of Ambulances and First aid Kits and defibrillators Next week during AMD upcoming processor launch...

DemonicRyzen666I'm with this as long it's upgradable to DDR5 and Pcie 5.0.

BwazeZen2 had uplift in single and in multicore compared to Zen1 or Zen+. 2700X to 3700X had 12% increase in single core CPU-Z score (457 to 509), but 13% increase in multicore (4839 to 5465).

It's strange that 3900X to 5900X wouldn't follow the same pattern, especially with the TDP and frequency uplift. I could imagine this result if processor was thermally or power limited.

Parkab0ymulticore comparison

3700x 5433pt @ 88W
5900x 9481.8pt @ (150W) TDP value
+62.5%pt , more 33% cores, almost double the energy +16% performance per-core (same compared to 5900x)

5433/8c = 679.125
9481.8/12c = 790.15

679.125/88W=7.71
790.15/(150W)=5.26

7.71/5.26=1.4657

3700x=46,6% more efficient in the best case if the 9500x doesn't surpass 150w value at full load

5900x 9481.8pt @ (150W)
3900x 8189 @ 142.09W
+15,7%pt

3900x still more efficient by 9,7% in the best case if the 9500x doesn't surpass 150w value at full load

single core comparison

5900x 653pt (impressive but 653/522=1.25, so +25% not 27%)
3900x 522pt
3700x 511pt

if 5900x will be over 150w in full load how they think to cool it?
with a stock cooler 1kg double 140mm fans?

mtcn77The Ryzen 3000 ihs is dished, therefore its coldplate is uneven. The dish makes contact weak. I wouldn't suggest making a direct comparison out of 3700X'es heating characteristics for 5900X for that reason.
forum.level1techs.com/t/dotting-for-better-ryzen-3000-thermals-level-one-techs/147184/10

Parkab0ywww.anandtech.com/show/15043/the-amd-ryzen-9-3950x-review-16-cores-on-7nm-with-pcie-40/2
www.anandtech.com/show/14605/the-and-ryzen-3700x-3900x-review-raising-the-bar/19

OK but power comsumption is a fact, forget temps, for a while (I don't think the 5900x will be an easy beast to tame by air anyway)

Chrispy_I really hope Zen3 isn't a power-hungry monster in an attempt to reach 5GHz.

One of the best things about the 3600 I used to have was that it was pretty much silent on a modest air cooler.

Parkab0yexactly what we are seeing

a 3950x asks 144W at full load to reach 11000+ score on CPUZ multicore and can reach max 4.7GHZ to score around 560 in single-core

that 653 single core score from the supposed 5900x IMHO stands for 5.0 GHZ or even more, who knows how many watts was needed to achieve it!

Parkab0ymulticore comparison

3700x 5433pt @ 88W
5900x 9481.8pt @ (150W) TDP value
+62.5%pt , more 33% cores, almost double the energy +16% performance per-core (same compared to 5900x)

5433/8c = 679.125
9481.8/12c = 790.15

679.125/88W=7.71
790.15/(150W)=5.26

7.71/5.26=1.4657

3700x=46,6% more efficient in the best case if the 5900x doesn't surpass 150w value at full load

5900x 9481.8pt @ (150W)
3900x 8189 @ 142.09W
+15,7%pt

3900x still more efficient by 9,7% in the best case if the 5900x doesn't surpass 150w value at full load

single core comparison

5900x 653pt (impressive but 653/522=1.25, so +25% not 27%)
3900x 522pt
3700x 511pt

if 5900x will be over 150w in full load (where a 9900k can reach 180W being rated 105W TDP) how they think to cool it?
with a stock cooler 1kg double 140mm fans?

a 3950x 16c at full load reaches max 144W.

at 200w my bike push me at 20kmph.

FinneousPJAbout 85% over stock R7 1700. Yeah, I think I can justify the upgrade :)

Chrispy_I really hope Zen3 isn't a power-hungry monster in an attempt to reach 5GHz.

One of the best things about the 3600 I used to have was that it was pretty much silent on a modest air cooler.

Chrispy_This is a new process node from TSMC though, there's literally zero solid data about the efficiency of 7nm EUV at the moment. Your assumptions and guesses are as good as anyone else's at this point. Rumours and leaks point to it being 200-300MHz faster than TSMC's N7 at similar voltages, so in theory 5GHz on Zen3 isn't going to be much hotter than a stock 3900X.

Not long to wait now for actual, official, independently-verified answers at least.....

ratirtYou guys are not serious right? Considering the iterations of of each Ryzen, it was always going with a clock boost and not more power draw so I assume it is not the matter of "how power hungry this chip is" but a matter of how many watts are needed to get it running. Which I assume isn't much though. If you look for power hungry chips you should be talking about Intel.

Chrispy_I'm not expecting XMP to work with Intel timings on an AMD motherboard, but AMD write AGESA firmware, and that includes speculative memory training.

At the moment, that memory training gives up on XMP timings far too easily because it doesn't loosen them enough. Even very loose 3600 timings and 1800 FCLK are waaaay better than JEDEC 2133 defaults.

Say you have a 3600 kit with 18-18-18-40 XMP timings that won't run on a typical Zen2 CPU; regardless of the board, the AGESA firmware will use the XMP primary timings and take a rough stab at the secondary timings before giving up. The thing is, XMP primary timings are almost always compatible with Zen2, it's the auto-generated secondary/tertiary timings that fail to boot on AMD and those aren't even part of the XMP spec. You can likely get that 3600-18-18-18-40 kit to run at 3600-16-16-16-36 on Zen2, so the problem is not the XMP data stored on the SPD.

IMO, AMD need to stick to the XMP frequency and voltage, and then use the primary timings as a reference point to calculate some safe values to attempt on the memory training runs. The best thing they could do at this point is hire 1usmus since his DRAM calculator works really well and is only a megabyte even as a compiled windows application with a GUI. If he can make a bootable timings calculator based off a handful of input variables, AMD can integrate the same kind of thing into AGESA. It doesn't even matter if AMD assumes low-quality RAM modules and runs a super-loose set of timings. Take that 3600 kit I mentioned above; Even if it was run at 20-20-20-55 timings with tRC of ~70 and tRFC of ~600 it would still be so much better than giving up and running the FCLK at 1066 instead of 1800 just because the memory training failed to find bootable values.