Zen 2 for Distributed Computing: Any interest?

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
I don't need PCIe4 (yet) and I despise chipset fans, so I'm happy to wait for BIOS updates to make it back to X470 and X370 boards.

As a bonus, these older boards use less power because no power-hungry PCIe4 chiplet (up to 15W!).

If I had a need for fast PCIe4 NVMe storage devices or the absolute maximum PCIe performance I would consider X570, but as it is nothing appealing to me. Especially when I have purchased top-tier X370 and X470 boards for <$100 effective.
 
  • Like
Reactions: TennesseeTony

ao_ika_red

Golden Member
Aug 11, 2016
1,679
715
136
So far it's only MSI being sold around here and it's at least $50 pricier than previous X470 board.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,226
9,990
126
I wanted to fire up BOINC and test out WCG, on my shiny new R5 3600 3rd-Gen Ryzen CPU.

I already had BOINC 7.14.2 installed, with my projects, all set to "No New tasks", because I hadn't been BOINC'ing lately.

I fired up BOINC manager, clicked on WCG, clicked "allow new tasks", clicked "update", still no new tasks.

I thought perhaps my preferences were wrong, so I set them to allow 90% of CPUs, 100% CPU usage, clicked "Run according to Prefs".

Then I thought, well, maybe, new CPU, maybe have to run CPU Benchmark again. So I did.

Still, can't seem to get WCG tasks. I guess I'm going to try another project too, but this seems curious, somehow. Do I need to re-install BOINC, after changing CPUs?

Edit: Guess the problem is with WCG? I allowed PrimeGrid tasks, and they started flowing in immediately.

Edit: Wow, with PrimeGrid, according to HWMonitor, my Package Power for the CPU is hitting 85.74W, my PPT is 93% of 88W, and my temps are 94.5C out of 95C max, WITH a copper-cored tall 95W TDP R5 1600 heatsink, not the stock R5 3600 65W Aluminum heatsink that came bundled.

My "idle" temps are 60C. Yet, I tweaked it down completely on the CPU, didn't have any "spinners" (heatsink mounting screws that didn't bottom-out, because they weren't attached).

I dunno, I think that these 3rd-gen Ryzen CPUs need AIO WC kits at a minimum for mining or DC loads. Their stock heatsinks are only sized for "gaming loads".
 
Last edited:

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
Sounds like the motherboard is allowing operation out of spec, but that's just speculation on my part.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,226
9,990
126
Yes, I'm pretty certain that this B450 board has a sort of "immature UEFI" for 3rd-Gen Ryzen CPUs. Waiting for a release that includes AGESA 1.0.0.3(AB). This is the 1.0.0.2 version.
 

biodoc

Diamond Member
Dec 29, 2005
6,257
2,238
136
I've got my 3700X up and running. I think the bios in the x470 AsRock master/SLI board has minimal support for Zen 2 because on optimized default settings, there's no overclock at all. I'm running some ODLK tasks because they are short and power draw at the wall is 91 watts, lol. :cool:
 
  • Like
Reactions: IEC

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
I've got my 3700X up and running. I think the bios in the x470 AsRock master/SLI board has minimal support for Zen 2 because on optimized default settings, there's no overclock at all. I'm running some ODLK tasks because they are short and power draw at the wall is 91 watts, lol. :cool:

I've got that same board. They're either AGESA 1.0.0.1 (initial support) or 1.0.0.2 (older June version) for their latest 3.30 version. So you'll have to wait for 1.0.0.3 AB (early July AGESA) or better to get OC options. Personally I don't think it will really be solid until at least 1.0.0.4
 
  • Like
Reactions: biodoc

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,478
14,434
136
I've got that same board. They're either AGESA 1.0.0.1 (initial support) or 1.0.0.2 (older June version) for their latest 3.30 version. So you'll have to wait for 1.0.0.3 AB (early July AGESA) or better to get OC options. Personally I don't think it will really be solid until at least 1.0.0.4
I think the latests is 3.43 bios. It was just released 7/9/2019. But thats for the X470 Taichi
 

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
Do you guys think the Hyper 212 Evo is better than the cooler that ships with the 3700X? I have the Evo on the 1700 Pro right now.
 
  • Like
Reactions: Ken g6

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
My new 3700X runs on an old B350 board:


cpuz3700x.jpg
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
Yeah, boost is broken with AGESA 1.0.0.2 BIOSes from ASUS. The highest ST boost I've seen on my 3900X is 4333... and I see the same max boost all the way down to -150mV vcore offset (I did not test further). It's apparent to me that whatever bug is in their algorithm it causes it to think it's hitting the limits even when it's not.

So for now I'm running a -100mV offset until an updated BIOS comes out. Even at 4300ish ST clocks I'm seeing over 500 Cinebench R20 ST, so there is that. I've basically given up nothing over my OC'd 8700K in terms of ST performance while almost doubling the MT performance at <145W package power versus 151W for the 8700K OC...
 
  • Like
Reactions: phoenicis

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
Thanks, looks like I may have gotten about 100MHz or so at full load with your tip. Honestly I was not expecting much more , it is a 65w chip and I do want it to be efficient. At the moment it's sitting at about 40.5X with a full non-AVX load, 80° with a Hyper 212 Evo.
 

phoenicis

Member
Nov 26, 2017
56
63
91
Thanks guys, interesting data on the new chips. The heat being generated by such a low power draw has me scratching my head a bit.
 

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
I've not checked the actual power draw, but the TDP obviously has to be taken with a grain of salt. If I had to guess, knowing the characteristics of the Hyper 212 HSF, my 3700X is drawing closer to 100W than 65W under continuous load.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
Thanks guys, interesting data on the new chips. The heat being generated by such a low power draw has me scratching my head a bit.

It's the same problem as the Radeon VII. You may have a more efficient architecture and process, but the chips are physically very small surface area so the heat density is higher. This leads to difficulty transferring the heat away quickly enough as well as hotspots, particularly when AMD tends to boost very aggressively at stock.

Undervolt or target boost clocks to the sweet spot for efficiency and suddenly these chips have literally no competition.

I'm at -125mV on my 3900X for now with better-than-stock performance at significantly less power draw. It's a real head-scratcher, but I have to presume an overly aggressive or broken boost algorithm is to blame.
 
  • Like
Reactions: StefanR5R

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,323
4,904
136
I lose performance beyond around -132mV so I'd assume I can go lower but at the expense of performance. I tried -175mV and it dropped almost 5% in benchmarks. Though to be fair you could probably really crank up the efficiency by dropping another 100mV...

Edit: I was using primarily Cinebench and CPU-Z to benchmark. And stepping down 12.5mV at a time.
 

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
Things sure have changed. In the past, going too low on voltage would get you a crash, now it's just a down clock? Pretty amazing!
 

bill1024

Member
Jun 14, 2017
88
73
91
Do you think AMD fixed their implementation of AVX and AVX2 to equal Intel's?
It was so bad primegrid used their SSE4 rather then their AVX.