Rather than an uptick, these are far lower clocks than 32c Naples. The base is 1.4GHz. But anyway, we know AMD will offer lower core count, frequency optimized SKUs as well, so I guess we will see how they fare...I wouldn't expect final clocks on the 64 core model to be much higher than the 2.2 GHz shown. That's the highest base frequency of the current Epyc 32 core model and they will need to fit within the same TDP while doubling the number of active cores and adding full throughput AVX256. If anyone expected a large base frequency uptick while doing all that, I don't know what to tell them other than to have more realistic expectations. This also says nothing about what frequencies a high end consumer focused part might be able to reach.
I know, I’d like to start at 32 then move to 64 when it’s bargain pricing.why stop at 32 GB I got 48 GB of 3200 DDR4 in my system. Yeah it was not cheap,
Rather than uptick, these are far lower clocks than 32c Naples. The base is 1.4GHz. But anyway, we know AMD will offer lower core count, frequency optimized SKUs as well, so I guess we will see how they fare...
Rome already has some HPC wins and people in the industry seem to have a very positive reaction to it, which is a good sign.
Eh?
https://www.newegg.com/Product/Product.aspx?Item=N82E16819113471
Now I will agree, clocks will have to come down for AVX2 workloads to stay within a reasonable TDP.
Taking the quote from TSMC literally this would be right - however far from ok if we are taking real physics into consideration. Therefore TSMC quote can only be true at a certain voltage point - and would be right for the whole voltage range if df/dV would be constant (e.g. the frequency-voltage curve would be linear), which is not the case. Taking a real frequency-voltage curve into consideration the power gain would be less as you approach lower voltages.
As conclusion - you under-estimate the power for lower clock frequencies.
Not qualification samples, IIRC. Maybe Rome's final frequency could still be higher, though.Early Naples samples had the same 1.4 GHz base frequency.
https://www.pcper.com/news/Processors/AMD-Announces-EPYC-Massive-32-Core-Datacenter-SoC
Not qualification samples, IIRC. Maybe Rome's final frequency could still be higher, though.
What is different in the product page from what I said? It shows a base frequency of 2.2 GHz. . .
One thing to consider. Thats a 64 core 128 thread Rome. The EPYC series clocks way less than threadripper to be efficient. If a QS is doing 2.2 ghz (turbo more ??) for 128 threads, I would say thats pretty good. Since the factory 7601 EPYC today has 2.2 ghz, and turbo 3.2, then they just doubled the speed on the same socket at the same speed, and probably the same power draw.I wouldn't read too much into clock frequencies on engineering samples. When Zen initially launched all of the engineering samples had significantly lower clock speeds than the final silicon.
The original Opteron's/A64's being the biggest example as the web filled up with doom and gloom about AMD's lackluster clocks. They came out clocking ~25%-40% faster.
Eh?
https://www.newegg.com/Product/Product.aspx?Item=N82E16819113471
Now I will agree, clocks will have to come down for AVX2 workloads to stay within a reasonable TDP.
Eh?
2.2 GHz is the highest base frequency of the current 32 core EPYCs.
Besides AMD stated those numbers for the Radeon 7 clocked at 1.8, for Epyc 2 wich is clocked at around 2GHz, but we also know from their Matisse demo that those numbers hold at 4GHz as well, i mean, all the necessary info is available, dunno how you came to the conclusion that it would be accurate only at a specific frequency/voltage point.
Yeah my mistake, I already responded to that above:
http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=threads/speculation-ryzen-3000-series.2558009/post-39750501
Yeah - I'm sometimes too lazy to read through all the thread rigerously/too stupid to see all pertinent posts. [delete as you feel applicable]
Second, as i pointed out, it is the differential d(V(f)^2*f)/df which determines the power-frequency function - its just physics and math. Your calculation cannot possibly be correct for all voltage range unless dV/df is constant - which is not the case.
By the end I thought we saw a couple 1.2 and 1.4 GHz opterons. But yeah that was sandbagged heavily before release.More than that, actually. All the Opteron ES chips were clocked at 800MHz. The fastest model on first release was 1.8GHz.
I think we might see some 2990wx like 32c monsters if AMD sells EPYC in 4 chiplet variations.
The value of dV/dF can be easily extracted from the formulae i posted, you ll see that its value is about 1, wich mean that the voltage is proportional to the frequency and that the power/frequency is a cubic polynomial, if the curve was closer to the ideal value, wich is a square law, then for sure power at 2GHz would be higher than what i stated, but that s not the case.
As for typical dV/dF you can check the existing ones, for 14nm the value is constant up to 3.3-3.5GHz for GF and up to 4-4.2GHz for Intel wich is closer to the idealized square law.
Not sure what you want to prove here....frequency-voltage curves are hyperbolic and not linear.
Even in you example you can clearly see that at 2GHz the first derivative is much smaller than at 3GHz and thats again much smaller than at 4GHz - and its not looking like a TSMC N7 curve at all - even if we assume high Vth libraries.
I think it would make better sense to limit the threatripper brand to products under $2500, and simply release a TR4 socketed Epyc. Bringing Epyc brand to the TR4 would be a very good thing in my opinion.
2990wx should fill the bulk of TR4's 32c/64t demand.
I think it would make better sense to limit the threatripper brand to products under $2500, and simply release a TR4 socketed Epyc. Bringing Epyc brand to the TR4 would be a very good thing in my opinion.
What I meant by that is having a high clocked 32c CPU for those that are seeking high clock speeds with a pretty high core count. A 4 Chiplet version, even with a 180w TDP should have a lot of room to turn up clocks even more if they open EPYC to 250w.