We tried reaching you on a private message asking for the Service Tag number to ascertain the warranty but did not receive a response. Please feel free to reply to the private message whenever you are available.

I tried replying, but the bot ignores the tag telling it's incorrect. Please, replace the bot with working one.

Please DM us the information we have asked you in the previous message, we will be happy to assist you. 

I have sent you the tag in text in direct PM

Replied in PM.

The latest bios does not solve the issue.

waiting for a fix.

P.S. sorry for different account, logged from different machine/email

wow, sad to see your issue is still ongoing. And on this note, that's really a weird cpu load graph

Interesting. Seems, it is a new issue with Cascade lake Xeons, and it is NOT related to full load.

I tried overloading it with 200% of simulation work and these spikes disappeared.

Now, do you remember what "Voltage droop" is? It is a very stupid way of fighnting the reactive current spikes when core load drops rapidly. The older CPUs were requesting down-voltage at high loads, which was leading to system instability.

Now remember, that there are (roughly speaking) 3 types of throttle: 1)NOP cycles, 2)Clock down 3)protective full core halt

Number 2 and 3 should increase spikes. but carefully spreading NOPs in the conveyor can probably smooth down the spikes. But for that you have to analyze the conveyor load in advance, or guess the load from a voltage drop on a resistor in Vcc circuit of other cores/whole package.

Seems like Cascade Lake is able to issue such throttling NOPs.... But I wonder how it decides when to do that?

Now look at the graphs. All the spikes happen at low load when cores are ALREADY underclocked and underloaded.

So probablym that is a resistive sensor in VCC circuit, and it is either oversensitive, or senses some noise.

The latter means, the VCC circuit is badly designed.

In the best case, the sensitivity can be tweaked by developers in BIOS, and I hope, DELL will do something about that. Because unneeded throttle reduces performance in real-life applications.

And still, it is not good even in full load.

Just checked how it is going under multitask simulations. Still EDP current throttling, even without shift-downs.

Bad, very bad.

I have a t5810 using a e5-1650v3. Playing around with it I run into the same throttles. The bios has a hidden intel rcsetup menu. I have an external programmer and all the bios modification tools. I have tried flashing my machine a dozen times with a modded bios to gain access to these options. Dell did a very good job at setting many defaults to make other settings not take affect. If and when I figure out how to modify the variables in the efi modules I will have full control over the current limits. I know you have a newer machine but the same would apply. If I figure this out I will post back and go through with you on how to pull a bios dump and I'll modify it for you. It's very upsetting to have such an expensive machine and not be able to use all the power. I modify voltage using throttle stop right now but if I don't change anything I get throttling at 25 percent load instantly with that 6 core...... I also have a massive cpu cooler I custom installed and made a vrm heatsink.