Built another Ryzen rig:

Spec:
3700X on Wraith Prism
Gigabyte B550I Aorus Pro AX
2x8GB DDR4-3600
HP EX950
FSP 450W PSU
Radeon 6670 (ancient card used only for video output)

Set up the rig, plugged a USB drive with Windows installer to it and turned it on, but no display output is visible. The mobo RGB LEDs light up in orange, and the Wraith Prism appears to use the RGB profile from BIOS, though. No error beep codes when I plug in the speaker piece to the 4-pin header on the motherboard.

Tried the following steps, the PC still doesn't boot:
- removed SSD while leaving the Windows installer USB drive plugged in
- reseated SSD
- reseated RAM
- use only 1 stick of RAM in either slot
- removed all RAM sticks (in this case, this was expected to fail - the motherboard also failed to activate RGB LEDs too)

The CPU, RAM and GPU were tested on other machines and they worked fine. I presume I should RMA the mobo next? What else may I have missed out on testing?

Never thought of swapping RAM sticks! I just plugged the Bolt XR in another rig, it POSTed just fine, assumed it was not a RAM issue and never considered RAM compatibility with the BIOS.

Let me see if the BIOS flashback option works. I presume it functions even without working RAM?

Found out the motherboard is on BIOS version F1...

Plugged a 2400 stick inside, capped speeds to 2400 instead of auto, ran Q-Flash to get it to F13h, now it seems to boot fine and managed to install Windows.

Now the problem is trying to enter BIOS again. It moves too fast to booting Windows, and spamming the delete key causes BIOS to fail to POST.

https://www.amazon.com/dp/B08166SLDF

5600X is on sale again, US$339.46 after tax and shipping. Amazon might also potentially refund you part of the costs if they overcharged you for taxes too.

What is the PPT/TDC/EDC readings in HWInfo?

https://www.tomshardware.com/news/amd-shows...ing-improvement

Possible Zen 3+ (with huge 3D stacked L3 cache) to fight Alder Lake?

Yeah Prime95 is going to toast this CPU for sure with small FFTs.

Both brands are moving towards stacking chips anyway. Intel technically did it first with their Lakefield CPU that appeared on a handful of ultra low power laptops last year:

I won't be surprised that they will announce a price drop for the 5600X/5800X before August. 5800X went not too long ago for $418 on Amazon, 5600X is regularly available at $299.

I have the 5800X, and to be frank having this chip be the hottest of the lot does make sense. 142W PPT stock only on a single CCD, unlike the 5900X or 5950X spread across two. Not a surprise that the dies on it were probably cherry picked only for high clocks, not high efficiency unlike the 2 CCD ones.

And frankly you don't want to see (insert brand here) succeed unless you're a shareholder, you want both to be neck and neck with each other. Prices will then be very competitive across both brands then.

I mean PPT/EDC/TDC values read by HWInfo when running Cinebench - the processor may not necessarily hit the max values you set.

Yup that's what I want to see. Interested in seeing how much these values can go up to assuming ideal cooling, and after they start getting throttled due to high temps.

Yup CPU cache heavily benefits games. Problem with Broadwell is it clocks horribly, hence why it barely appeared on desktops.

In non gaming benchmarks, the large cache benefit isn't well utilized though, potentially making the CPU more expensive than its worth. I suspect we won't see this in Epyc or TR.

Intel has their Foveros project, with a shipping implementation using 1 Ice Lake core, 4 Atom Tremont cores, SOC components and DRAM stitched on top of each other. Granted it's a low power component, but the tiny die sizes of each part is nullified by the high number of layers, it looks a lot more complicated to implement than AMD's.

Mostly due to the case modeled after high air flow mid towers, unlike the crappy Dell designs that are still stuck in the 90s.

It doesn't look like you're pushing the SoC that much with your RAM speeds. I see higher SoC power usage on my 5800X using RAM at 3733 speeds and 1:1 IF clock, even at idle. Even Rocket Lake capped at 65W can show a disparity when running Gear 1 speeds at 3200+ due to power pulled away from the cores.

This post has been edited by yimingwuzere: Jun 9 2021, 09:50 PM

Ryzen likes faster RAM though

3200-3466 is the sweet spot for Zen+ IIRC.

Different chips entirely, so can't even call them separate bins.

5700G is the top bin for APUs with onboard graphics and SoC all in one die.

5800X is a pure 8-core chiplet with more L3 cache, but no graphics and SoC on the same die. These are probably less than ideal bins too - 8 working cores and can hit high clockspeeds but can't meet 5950X power targets reliably seems most likely. So it ends up being a power hog for a single CCD.

Not gonna lie this looks like an awful case. I can't see any airflow going inside it from anywhere. Also interestingly enough, HP Malaysia's website says this desktop should be specced with a 5800X not 5700G?

I wonder if that GPU uses the same board layout as the 3070 FE though...

This post has been edited by yimingwuzere: Jun 15 2021, 01:31 PM

Not really an excuse for OEMs, especially with Dell stuck on 90s style heatbox designs.

The Omen seems to have another variant using a thicker front panel and NZXT style cutouts for fan intake, doesn't seem to be the case with yours.

Lenovo has a desktop model with a dual fan intake at the front behind a mesh grille, that looks like a gold standard for OEM designs.

Video guy was expecting better results than der8auer's results (he still used the IHS), ended up just matching it after all that effort. Ouch.

Then again, Ryzen CPUs that use solder since the first gen was always proven to be not worth the delid effort, no surprise it's not any different now.

Yup, but this Chinese guy already did even more than der8auer did, even sanded down the dies so that it's level. So to only see 4C difference is very disappointing.