AMD EPYC 7H12 Announced As New 280 Watt Processor For High Performance Computing

[tildearrow]

Wonderful piece for my HTPC. Does it have iGPU? As, currently I'm running on APU Ryzen 2400G.

Really? A 64-core HTPC? Are you going to watch 8K120/16K video?

[tildearrow]

Right now AMD is cheating and fooling the customers with their Ryzen3000 "Agesa 1003abba" microcode update.

Wait what? AGeSa "1003abba"?

How many more letters are they going to use?

P.S. it would be cool if they changed the boring OEM-y " AGESA" name to something cooler, like XBS (Cross-Board Software).

[numacross]

Must be its TDP when running AVX-512 instructions on all 56 cores at full-load.

You know there's a "?" next to TDP that explains what it means, right?

Anyway that is just 2 glued together processors that nobody can even buy since it's only available in the S9200WK series.

It is liquid cooled with the chassis having 3x 2100W PSUs (for up to 4 compute modules).

[numacross]

(test of the Agesa 1003abba) (german)

https://youtu.be/aM8I7NfpzXQ

I watched the English version and he didn't even present performance differences between the BIOS versions, but just muddy frequency graphs with his own convoluted interpretations.

The problem with that is Zen 2's ability to shift frequency so fast that actually detecting it is very hard. This is the reason AMD is going to publish an API for use in monitoring tools because the current methods are simply too slow and error-prone.

Every other test I've seen has notable increases between the ABB and ABBA versions. ABBA is usually faster and more energy efficient. For example this one. This shows a measurable increase in performance including the most important 0.1% low measurement.

[numacross]

no you are wrong he did in fact show "performance" means an overall increase of performance of 0.8% some is up to 4% slower but some other is 5% faster makes it a overall 0,8% increase of performance.

Why are you lying? He specifically says that he doesn't do performance in that video.

but thats not the point AMD is selling CPUs with advertisement on high CLOCK SPEED MHZ but they lie the CPUs can not RUN at the advertised clock speed. and he clearly proof that "Agesa 1003abba" is just a cheat to fool the idiots because the high "CLOCK SPEED MHZ" what they advertise is only without any "load" and if some load hits the cpu the cpu clocks down.

thats clear "Fraud"


but hey without open-source Firmware everywhere "Fraud" is the status quo in the garbage IT industry.

In the same video you used as "arguments" you can clearly see the processor hitting the advertised boost speed on ABBA while it wasn't on ABB.

You're a troll, and not a very good one at that it seems...

[numacross]

he quote 2 other videos of other people who did the performance work. and because of this the performance is known and not like you try to say: it is unknown... no wrong the performance is known.

why should i Lie? i just point to the sources who is the knowledge. at minute 9:57 he points out where you can find the performance numbers.

I linked to one of the videos he talks about in my previous post and... the performance increased across the board. You're lying again.

he proofed multible times in this video that the CPU never hit the boost clock speed if there is load.. it always hit the boost clock speed bevore and AFTER the load. and this is cheating. you act like there is high clock speed but all the time in load it is lower clock and all the time the load is over you clock high to fool the people into your cheat believe of high clock speed.

It doesn't matter when the processor hits the advertised top boost speed. It does now, and it didn't before - the issue is fixed.

There is no "cheating" since AMD never specified at what "levels of load" the processor would do it. The only guaranteed speed is the base clock at all times and "up to" maximum boost speed. Not to mention the usual sustained boost speed is very close to the max anyway while the CPU stays within the TDP which is way better behaviour than that of the competition.