Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Oct 11, 2021.
It was a floating heatsink. No mounting kits for 1700...
I want to see the intel cpu being on top of 1080p cpu limited games, don't care about nothing else, multicore ryzen can't be beat and intel isn't even trying with it's real+idle cores
This isn't a fair benchmark due to fact the AMD's L3 cache is broken on the released version of Windows 11 and has mention in my earlier post the R9-5950X is almost a year old.
It's only a year old.
It's not, but just adding more performance without thinking of efficiency isn't great either. 250W for the CPU package is double that of the 5950X using all cores: https://www.anandtech.com/show/1621...e-review-5950x-5900x-5800x-and-5700x-tested/8. Looking forward to seeing actual benchmarks without any OC features turned on!
Maybe wait for "7950x" LOL
It's kinda funny we have soo many synth benchmarks leaks yet gaming leaks are close to zero . Why ?
Probably because those with access to these parts couldn't care less about gaming benchmarks.
The good news is, it seems Intel has finally provided realistic TDP numbers, rather than the nearly useless numbers they have provided in the past. The bad news is, that wattage is a bit alarming, for 8 P-cores. If the 8 E-cores were also under full load, that doesn't speak well of how efficient they are. Though, I suppose that would explain why Intel is charging roughly $50 per E-core, because they otherwise seem to be more focused on performance. Perhaps PCIe 5.0 or DDR5 causes more power draw, but if that's the case then this is bad news for laptops.
~250W in and of itself is a reasonable number for a flagship CPU, but again... it's just 8 performance cores. If it were 16 performance cores then I'd be like "oh ok, sounds pretty good" and move on. If this were still 14nm I'd say "that wattage makes sense".
The 5950X also has dual-channel memory, which would cripple its performance quite a bit.
Sending 900€ of to Ebay land to receive a 4 year old CPU, Mainboard and waterblock and then cross your fingers that it is even stable anymore or that it is not full of stab marks from a delid or something.
You are a lot braver then me.
5800X doesn't do 225 watts on this test.
I only hit 133watts with PBO on when I test it.
Indeed, thanks for the report.
But shorely we're just AMD fanboys here, the fact I called it
doesn't mean a thing.
Wintel continues to disappoint and it's not even out of the gate, but what can we expect? So many transistors (= just current valves) on such a large node cannot run cool. AMD clearly does have the process advantage.
FPU only is selected.
PBO is on plus CO and no all core overclocking if I go stock the power usage isn't much different.
Impressive, but in the same article they quote a Forza Horizon benchmark comparison where the 12900K leads the 5950X by a whopping... 2%.
Fair enough, I wasn't aware of that. I'm getting so tired of these leaks. Intel should really do us a solid and lift the damn embargo already. Or at least post some cherrypicked benchmarks of their own.
That looks like 100% gpu bound to me?
9900k has like 40% ! higher minimum fps than 2700x
Nice try cherry picking
Cache is king and low latency is king in most games.
Unless you are measuring it with a multi-meter , vrm data can be unreliable. using the cpu package power reading is usually more accurate for estimates.
CPU Package Power is based on VID reading along with amps measured from a CPU MSR (via Serial VID interface).
VID can be substantially different from vcore if you are using manual voltages or your AC and DC Loadline values are not synched separately with Loadline Calibration (this gets tricky; DC Loadline on auto voltages should be set to the same value in "Milliohms" as loadline calibration, and if you don't know the mohm values for each LLC level on your motherboard, well...good luck. And AC Loadline is a completely confusing, different ball game as this affects CPU supply voltages on auto or offset vcore modes, and you absolutely DO NOT want AC Loadline to be a HIGH mOhms value when VRM Loadline (Loadline calibration) is a LOW mOhms value, at the same time).
Accurate wattage readings can usually be found from Asus EC on Maximus boards, or the VRM specific section (where VR VOUT) is on Gigabyte boards. If supported, these can be found in HWinfo64. MSI boards have had some shenanigans with their IR 35201 / ISL 69269 VRM readings (some crappy stuff about "die sense (VCC_Sense)" or "socket sense" set in the BIOS completely changing the VRM Readings or loadline levels when they should NOT do that)
in the case of ryzen you have the SMU(system management unit) providing the package power measurement, It isn't perfect, but it will provide better estimates than vrm vcore*amps output alone, since you have to also do the SOC vrm aswell if you want the total power draw for the chip. its also very convenient.
Point was mainly that @tty8k is getting power numbers wayyy outside where they are supposed to be with his method, it would wise to consider other measurements available before concluding that all ryzen cpus are drawing double power or perhaps identify if there is something horribly wrong with his computer.