Pentium Gold 8505 based Router

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

fta

Active Member
Feb 19, 2017
154
163
43
94
Also, can you check if hardware p-states (HWP, aka speed shift aka SST) is enabled?
 

oneplane

Well-Known Member
Jul 23, 2021
844
484
63
I wouldn't be surprised if the firmware was more simplistic due to the TPM2.0 requirement. They aren't going to get Intel to sign off on their fTPM or PTT unless they go through some form of compliance.
 

Infected1

New Member
Feb 21, 2023
8
0
1
I got the 8505 unit from cwwk today which was the fastest delivery of the 4 units I now have (~10 days from order to door). BIOS is pretty sparse compared to the other 3 - I think it was dated 02/07/2023. Haven't had time to play with it yet other than to slap 64GB of RAM into it, a Crucial NVME drive, and start up memtest86+ on it. After an hour and a half of memtest86+ (all-core), the chassis was 60C (140F) which, to me, was definitely uncomfortable to touch. It comes with a Daijing 96W power brick which I will likely keep using. One of the other units I have came with a Daijing and it was the most efficient of the three. Draw during memtest86+ was 36-38W.
Thats great man! And yeah that really was fast. so the Bermuda Triangle was obviously avoided by them with grace I suppose. Sub 40W is still acceptable. Can you check idle temps and consumption? And maybe also benchmark it a few times in a row to check scores and if it will throttle. Also, maybe check the CPU info for "ES" chip. When you say the chassis was,do you mean you used a thermal camera? What was the CPU core temps?
Also would it be possible for you to check its cooling system, whether or not the dual copper pipes was in need used. I would still put a fan tho.
 

skimikes

Member
Jun 27, 2022
83
79
18
Can you do a test for me? Run prime95 with the smallest fft size and see if it crashes?
No issues with either small or smallest fft size - let it run for about 2 hours. Unfortunately the OCZ-RD400 drive that I had an old copy of Windows 10 on with all of my tools appears to have bit the dust while I was running the test. Given the drive was from 2016, which makes it almost 7 years old, I'm not going to put the blame on the 8505 unit. I'll have to start over with a new drive. I have Crucial, Teamgroup, and Hynix drives handy so it's not really an issue - just bad timing.

Here were some grabs I took before it went poof:

8505-hwinfo64.PNGcpuid-p95.png8505-hwmon-p95.png

cpu-z and hwmon are from during the prime95 run.

The unit drew between 37 and 38W during the prime95 small fft (max heat/power) run. Heat is very effectively transferred from the processor to the chassis. I already got the chassis to reach 60C by running memtest86+ on it for several hours with no fan. For prime95, I just used the same USB A/V cabinet fan that I was using on a dfifferent unit and it keeps it quite chilly so I'm not going to bother disassembling this one. Chassis temp was measured using a Fluke-62 IR thermometer (I don't have an IR camera to play with).

SST is greyed out according to hwinfo64.

Idle CPU under Windows 10 was 14-15W, so worse than N5105/N6005, on par with J6413. It would bounce up to 20W as Windows did whatever Windows things Windows does in the background while it was idle. The power meter did spike to 50.1W at one point while Windows was upgrading since I haven't plugged that drive in for a while, but that's the highest I measured and I can't seem to reproduce that with prime95. Not sure what else I can use to generate max power on the processor.

There are 2 NVME slots. Observed the unusual behavior that a drive plugged into one slot would always first go to a UEFI prompt regardless of what the boot order was set to. From there, ctrl+alt+del would always boot into Windows. Moved the drive to the second slot and it doesn't have that problem. Yanked the power cord a few times to make sure that it would auto-boot on AC restore and it did, but only from the second slot which, in this model, is provided via an adapter. Not sure if that is a drive issue since it was that old OCZ-RD400 or if that's just how this unit behaves with all drives. Will figure that out over the next few days.
 
Last edited:

fta

Active Member
Feb 19, 2017
154
163
43
94
No issues with either small or smallest fft size - let it run for about 2 hours.
Thanks! So my 1235U version crashed when I ran that. I've now diagnosed it to the junk power supply they ship with it. It couldn't handle the initial rush of current when the test started. Replacing it with a GlobTek 90W power supply has solved that issue. Based on the weight comparison of the included PS and the GlobTek, there's no way I would trust running the included PS 24/7. My guess is it has very little if any safety design in it.

SST is greyed out according to hwinfo64.
Your BIOS has the same bug as mine, which means you can't use HWP. I have this fixed in my BIOS now, though. I'll demonstrate it when I post my review of the 1235U box.

The power meter did spike to 50.1W at one point while Windows was upgrading since I haven't plugged that drive in for a while, but that's the highest I measured and I can't seem to reproduce that with prime95.
Interesting. This is probably why you don't get the crash. The 1235U will spike over 55W no problem on initial startup and then is limited to the PL2 at 55W. I wonder if PL2 is set lower in the 8505 version?

There are 2 NVME slots.
Seriously?! Mine only has 1 NVME and 1 Wifi NVME. I wish mine had 2 NVME instead.
 

Vince

New Member
Feb 17, 2023
14
12
3
Seriously?! Mine only has 1 NVME and 1 Wifi NVME. I wish mine had 2 NVME instead.
This listing shows a PCB adapter board to use the WiFi NVMe port as another storage slot as an included accessory. My 1215U didn't come with it. Try contacting your seller to see if the'll send one out to you.
 

fta

Active Member
Feb 19, 2017
154
163
43
94
This listing shows a PCB adapter board to use the WiFi NVMe port as another storage slot as an included accessory. My 1215U didn't come with it. Try contacting your seller to see if the'll send one out to you.
Thanks! I'm not very familiar with these NVME form factors. From this listing, I can see the second NVME is just a 1 lane slot with a different form factor. The 1 lane is not going to be very fast, unfortunately.
 

skimikes

Member
Jun 27, 2022
83
79
18
Thanks! I'm not very familiar with these NVME form factors. From this listing, I can see the second NVME is just a 1 lane slot with a different form factor. The 1 lane is not going to be very fast, unfortunately.
Just FYI, via the adapter, I'm getting a bit over 800MB/sec sequential to the PCIe 3.0 x1 slot so while it might not be the 3500MB/sec the drive itself is capable of, it far exceeds the needs of the device. A backup of a pfsense guest on Proxmox ran at about 310MB/sec with lz4 compression on ZFS and that's really about the most IO intensive thing this device will ever do (for me). I haven't migrated the SK Hynix P31 drive to the other NVME slot yet - it's on my todo list. I believe that one is PCIe 3.0 x4 so it should get the full bandwidth of the drive.

So now I've got pfsense running as a guest on Proxmox on it. C-states are currently enabled in BIOS, powerd is disabled in the pfsense guest. Proxmox is running the 6.1.2-1-pve kernel. I've installed the intel-microcode package, removed the /etc/modules.d/intel-microcode-blacklist.conf file that was preventing loading the intel microcode, replaced the relatively ancient /usr/lib/firmware/intel-ucode/06-9a-04 microcode with the latest from Intel-Linux-Processor-Microcode-Data-Files/intel-ucode at main · intel/Intel-Linux-Processor-Microcode-Data-Files, regenerated the initrd and verified it's running updated microcode (for my processor, 0x420 the default; it's currently on 0x429).

The system, at idle, with 6x i226 NICs plugged in, is running in the 11-12W range which is pretty good. Not as nice as the N6005-based CW-N6000 4-port I have that runs at 7.5-8W at idle, but on that I only used 2 of the 4 NICs and VLAN'd the heck out of that LAN interface. It is virtually identical to the N5105 CW-N11 unit with 5 of 6 i225 NICs plugged in, and handily beats the J6413 with 5 NICs connected, which was the most power hungry of my devices at 15W idle.

Idle temps (no fan, no heat sink on NVME yet):
Code:
root@proxmox:~# sensors
coretemp-isa-0000
Adapter: ISA adapter
Package id 0:  +30.0°C  (high = +100.0°C, crit = +100.0°C)
Core 0:        +30.0°C  (high = +100.0°C, crit = +100.0°C)
Core 8:        +27.0°C  (high = +100.0°C, crit = +100.0°C)
Core 9:        +27.0°C  (high = +100.0°C, crit = +100.0°C)
Core 10:       +27.0°C  (high = +100.0°C, crit = +100.0°C)
Core 11:       +27.0°C  (high = +100.0°C, crit = +100.0°C)

acpitz-acpi-0
Adapter: ACPI interface
temp1:        +27.8°C  (crit = +105.0°C)

nvme-pci-0200
Adapter: PCI adapter
Composite:    +38.9°C  (low  =  -0.1°C, high = +82.8°C)
                       (crit = +83.8°C)
Sensor 1:     +31.9°C  (low  = -273.1°C, high = +65261.8°C)
Sensor 2:     +34.9°C  (low  = -273.1°C, high = +65261.8°C)
 
Last edited:
  • Like
Reactions: fta

reutech

New Member
Jan 31, 2024
1
0
1
I just got my unit today! I was upgraded to the 6 port version since the 4 was out of stock. Up to now I've been running an SFF PC with a couple of PCIe nics using pfsnese. Last night I did my first install of OPNsense, complete with encrypted DNS. The PSU in my current router has become unstable after 7 years of use so it's time for an upgrade. I'm well into the "I have no idea what I'm doing" phase so I'm extra grateful for places like the STH Forum.

I've added a 32GB DDR5 kit and a 1TB WD SSD. Planning to run a couple of VMs, which is the new part for me.
 

Attachments