AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

[Journey]

Stumbled across the user manualk for teh Giugabyte Xtreme X570. Grabbed soime bios screenshots which are interesting

[/IMG]

By stumbled across you mean, you went to their website product page, support section, and downloaded it?

 

[Journey]

I'm not confused, 4.1Ghz @ 1.428V suggests they stoked the fire

Sorry you are confused, since my post was about your incorrect assumption, and was written before the image were posted. Ta.

 

[Jokester]

We can’t be sure CPU-Z is picking up the Vcore right there.

 

[decto]

Sorry you are confused, since my post was about your incorrect assumption, and was written before the image were posted. Ta.

Nope - still not confused

 

[decto]

We can’t be sure CPU-Z is picking up the Vcore right there.

CPU-Z and HWinfo are in the same ball park for CPU vcore

 

[Journey]

but power usage scales exponentially with frequency

Nope - still not confused

I think you'll find you are... might want to read how power scales with frequency.

 

[cliffy]

By stumbled across you mean, you went to their website product page, support section, and downloaded it?

In a round about way that would be correct, wasnt there yesterday though

 

[Jokester]

CPU-Z and HWinfo are in the same ball park for CPU vcore

Yes, they could be reading the same thing from the processor, and misinterpreting it on the basis they think it’s an older generation chip.

 

[decto]

Yes, they could be reading the same thing from the processor, and misinterpreting it on the basis they think it’s an older generation chip.

True, but it's also at >80C so either they are over clocking on the box cooler, or it's pushing some power

 

[pmc25]

This chipset chip is the same as the IO die on the processors themselves, and is effectively a massive overkill for 99% of the users out there.
I don't know why AMD didn't design something else for a chipset.

Because of it, the mobos become so expensive.

I don't understand why two 75 sq.mm dies + one a little bit larger should be sold for $750.
It boggles my mind how small and cheap these chiplets are.

Complete nonsense.

It's a very small component of the increased cost, and makes sense for AMD themselves as they're not redesigning the wheel again.

Main increase in costs are higher spec power delivery and VRMs, far more complex traces for the PCIe 4.0, higher layer count PCBs (even on 'low end' boards) to meet the signal integrity spec for PCIe 4.0, more heatsinks and active cooling (fan) .. and these are the first wave of new boards being shifted back to Taiwan from PRC, in order to avoid Trump's tariffs for the US market, which is is the biggest % cost increase at the low end (I think virtually all the X570s are going to be made in TW).

 

[decto]

I think you'll find you are... might want to read how power scales with frequency.

Well if you want to be a pedant, I could have typed 'power usage scales exponentially with frequency because in 30 years of overclocking I've yet to find a CPU where you don't need to ramp up the voltage with any significant overclock and this has a drastic effect on power usage. I'm also not aware that AMD has invented any new physics to avoid this effect' but hey, I didn't realise this was a dissertation.

I'm quite aware of the power laws and how power usage scales with frequency so you get a mutiplied effect of both frequency increase and square law power increase from voltage uplift, I was keeping it simple.

Still ~12.5W per core @ 4,1Ghz is a good result for AMD (assuming the 200W in the images is correct). Odd that everyone jumps on 5.5W but doubts the vcore, amps and package / core power readings.

12.5W @ 4.1Ghz would give us 100W + SoC power on a mildly overclocked 3800X which is a little above the stated 105W TDP and so seems to be in the ball park.

In my view these high core chips are about efficiency, the 3950x can run 16 cores at 3.5Ghz, giving you the equivalent of a 7Ghz 3800X on a 105W power budget - a fantastic achievement.
If you then have a game or task that needs a few fast cores, it can scale it's power budget to give you a few fast cores (vs intel I9 class) in a way that threadripper didn't so you get a really versitile chip for diverse workloads.

However all core, high ghz will have a high power cost as there is no free lunch, the images show 1.428V @ 4.1Ghz and 1.572V for 4.25Ghz, so even if there is an offset issue somewhere in the vcore reading, you're pushing an extra 0.15V for 0.15Ghz which is getting well into law of diminishing returns.

Also it's notable there are no higher overclocks, they stopped at 4.25Ghz @ indicated 1.6V.

Odd if it was so easy to go faster they didn't post something a little more exotic.

 

[TrixX]

This chipset chip is the same as the IO die on the processors themselves, and is effectively a massive overkill for 99% of the users out there.
I don't know why AMD didn't design something else for a chipset.

Because of it, the mobos become so expensive.

I don't understand why two 75 sq.mm dies + one a little bit larger should be sold for $750.
It boggles my mind how small and cheap these chiplets are.

So you didn't get the prices you wanted so have gone from the most rabid AMD supporter to being incredibly negative post Computex. It's amazing to see, but still disappointing.

How do you define making money back on 7yrs of R&D to get to the point of Zen 2?? Not to mention funding Zen 3, 4, 5 etc...

 

[Scooby-DoobyDoo]

I'm undecided between a 6 core and 8 core. I will probably wait for the reviews as on the 2000 series the boost was a lot worse on the 8 cores when multiple cores used.:

 

[Grim5]

I wouldn’t get a 6 unless you’re happy to replace in 1 or 2 years when it starts to get hammered in games

 

[Dirk Diggler]

Sorry, I'm at work so only managing a fleeting look in... And let's be honest, this thread is a tough gig for lurkers

Have prices been released, and if so could someone post a link or chart etc.?

Thanks

 

[Potatowithearsontheside]

Well if you want to be a pedant, I could have typed 'power usage scales exponentially with frequency because in 30 years of overclocking I've yet to find a CPU where you don't need to ramp up the voltage with any significant overclock and this has a drastic effect on power usage. I'm also not aware that AMD has invented any new physics to avoid this effect' but hey, I didn't realise this was a dissertation.

I'm quite aware of the power laws and how power usage scales with frequency so you get a mutiplied effect of both frequency increase and square law power increase from voltage uplift, I was keeping it simple.

Still ~12.5W per core @ 4,1Ghz is a good result for AMD (assuming the 200W in the images is correct). Odd that everyone jumps on 5.5W but doubts the vcore, amps and package / core power readings.

12.5W @ 4.1Ghz would give us 100W + SoC power on a mildly overclocked 3800X which is a little above the stated 105W TDP and so seems to be in the ball park.

In my view these high core chips are about efficiency, the 3950x can run 16 cores at 3.5Ghz, giving you the equivalent of a 7Ghz 3800X on a 105W power budget - a fantastic achievement.
If you then have a game or task that needs a few fast cores, it can scale it's power budget to give you a few fast cores (vs intel I9 class) in a way that threadripper didn't so you get a really versitile chip for diverse workloads.

However all core, high ghz will have a high power cost as there is no free lunch, the images show 1.428V @ 4.1Ghz and 1.572V for 4.25Ghz, so even if there is an offset issue somewhere in the vcore reading, you're pushing an extra 0.15V for 0.15Ghz which is getting well into law of diminishing returns.

Also it's notable there are no higher overclocks, they stopped at 4.25Ghz @ indicated 1.6V.

Odd if it was so easy to go faster they didn't post something a little more exotic.

https://hothardware.com/photo-gallery/NewsItem/48380?image=big_3950x_cpuz.jpg&tag=popup
This image shows all core 5.0GHz for 1.608V...
A little more exotic since it used LN2.

Edit: that's a 0.18V increase for a 0.9GHz uplift in frequency, or a 12.6% increase in Voltage for a 22% increase in clocks. To me, that suggests the listed voltage for the 4.1GHz sample is not correct, or at least not as low as it could be set.

 

[MaXxBoulton]

but... when reviewing, one has to use the best available hardware. That's AMD now?

Says AMD.. we have yet to see some real benchmarks.
As a fan boy I hope AMD do good. As a realist I thanks a 9900k at 5.3ghz will still be the better cpu for benching and so on.

 

[Potatowithearsontheside]

The other odd thing about your analysis is that you state that the node shrink will have directly benefitted the per core power consumption, yet we know that the 2nd gen weren't consuming anywhere near the 12.5w per core that you're attributing to the 3950X.

 

[theone8181]

IId be amazed if the fans on the choose don't throttle down. I can't imagine it's going to be working that hard for most people. Surely 11W is only when going crazy and doing aan nvme to nvme transfer.

 

[Murphy]

For me this puts x570 in whatever category is opposite of premium and desirable.
Why would I pay more for a board that consumes triple power and doesn't downclock?

I'm not entirely sure we're seeing the whole picture here, yes GN said it doesn't down clock but IIRC some board manufactures have also said something along the lines of the fan being temperature controlled and/or adjustable / capable of running passively, the only way both statements can be true is if the chipset, supposedly based on the I/O die from EPYC, powers down unused parts of the chip.

We know Infinity Fabric includes a control fabric with a fairly sophisticated system management unit so it's not beyond the bounds of reason that the chipset would just power gate the unused parts and thus reduce power requirement significantly.