Build Log: OCEANBLUe Upgrade

I laid hands on OCEANBLUe for the first time since it traveled with my parents to Moscow, Russia. I did get the opportunity to perform some work on it remotely, but performance remained sluggish. Maybe it was time to give the hardware a rest.

After my parents came back to the United States, my father expressed an interest in upgrading OCEANBLUe. I ordered parts.

Order Placed: OCEANBLUe Upgrade

I livestreamed much of the initial build, stopping when it was time to sit down to dinner.

A couple of notes from the build process:

Plastic brackets for HSF needed to be removed, but I didn’t need to remove the motherboard from the tray to get at the metal bracket underneath

Orientation of the AMD Wraith cooler is important – AMD logo protrudes, making the RAM slot closest to the CPU inaccessible

Houston, we have a problem

My first attempt to power on the system was unsuccessful – the system did not POST successfully, and the board indicated that the CPU was at fault. Symptoms were as follows:

The LED strip lights up briefly, then goes dark. With CPU HSF spinning, the CPU status LED remains lit

CPU orientation is foolproof, but I double-checked to be sure.

I decided that the CPU probably was not getting enough power: OCEANBLUe’s existing PSU had a 4-pin ATX 12V power connector, whereas the motherboard takes an 8-pin connector.

I wanted to get the machine up and running ASAP, so I made the decision to purchase a new power supply unit instead of taking a gamble with a four pin to eight pin adapter, even though I would have liked to have gone the latter route. I selected an EVGA 450 BT (Amazon.com) over a slightly cheaper power supply that lacked the 80 Plus efficiency rating.

When I received shipment of the new power supply on July 9, I resumed work on the build. With the new power supply installed, the system still failed to POST. No change in status.

Diving deeper into the GIGABYTE GA-AB350-Gaming 3 motherboard’s documentation showed that support for AMD Ryzen 5 processors was introduced in BIOS revision F6. Is it possible that the motherboard I received shipped with BIOS revision F3? If that’s the case, my options are either to ship the motherboard back to Gigabyte for an upgrade, or to borrow an AMD Ryzen 7 processor for the sole purpose of flashing the BIOS (I did this before when building my DIY FreeNAS server).

Feedback online suggested that the GIGABYTE GA-AB350-Gaming 3 should support Ryzen 5 out of the box, regardless of BIOS version. Some posters wrote about long first boot times, anywhere from five minutes to two and a half hours. These seemed unreasonable to me.

In any event, I submitted a ticket through GIGABYTE’s eSupport ticketing system. Unfortunately, I’d set my profile to indicate that I was in the UK. I corrected my profile and submitted a new ticket.

While waiting for a response, I phoned in to GIGABYTE’s technical support line. A pre-recorded message advised me that all of their technical support representatives were busy helping other customers, but I could leave my number and my call would be returned.

I followed troubleshooting guidelines and powered up the board with the RAM installed in each of the board’s four slots, clearing the CMOS between attempts. Still no change in status.

I received a call back from Robert, a GIGABYTE technical support representative, and asked if they could check to see which BIOS revision the board had gone out with. He placed me on a brief hold, and informed me that his supervisor had stated that the board should POST regardless of BIOS revision. I asked if he could run the serial number of the board to be sure, and was told that I could expect a call back soon.

Robert called back to state once more that it should POST with any AMD Ryzen CPU – perhaps there were bent pins on the motherboard? I told him that I’d checked and found nothing out of place. He suggested returning the board to the retailer.

I phoned back in to GIGABYTE technical support again later. I had managed to find a speaker to wire up to the motherboard. No matter what, the motherboard wouldn’t issue any beep codes. I didn’t think polarity was important, but I verified anyway. The GIGABYTE technical support representative had me try powering on the system without the CMOS battery inserted. I jumpered the CMOS reset pins on the board. Still no change in status. Maybe, just maybe I’d managed to foul up the CPU installation. I put the probability of this occurring at zero, but checked anyway.

Later, I tested the speaker by plugging it into my workstation. It beeped.

And so begin the RMAs

By this point I was beyond frustrated. I started the RMA process on July 11 by going to Newegg. The box that I received from them had taken quite the hit, so it seemed like the logical place to start.

FedEx came to pick up the motherboard on July 12, and Newegg sent out a replacement motherboard on July 18.

I received shipment of the replacement motherboard on Thursday, July 20. I remembered the GIGABYTE technical support representative’s advice to test outside of the case. With only the CPU + CPU HSF installed, the board’s CPU status LED lit up to indicate that the CPU was faulty.

SuperBiiz approved my RMA request on July 21 and issued a return label for me on that same day. I cleaned the AMD Ryzen 5 1600, placed it back within its packaging, and dropped it off at a nearby UPS Store on Saturday, July 22.

Meanwhile, my father indicated to me that he would be interested in driving a couple of ultrawide displays from this new workstation. The NVIDIA GeForce 9800 GT has a maximum digital resolution of 2560×1600. The NVIDIA GeForce GT 1030 (specifications) may fit the bill nicely: it’s also available in a passively-cooled configuration. Unfortunately, it doesn’t support NVIDIA G-Sync. I’m also considering the Radeon RX 550 to stay under the AMD umbrella.

I phoned in to SuperBiiz on Monday, July 31, to see if there was any update on my order. The CSR informed me that their system showed the item was received, but there was no additional information. She suggested that I call back on Thursday.

I called back on Thursday, August 3, and was informed that the CPU had been sent to their vendor. The turnaround time was longer than I had the patience for, and I was certain that the CPU was faulty. I asked if I might escalate to a return instead of a replacement order – the CSR got it approved, and I went back to the drawing board.

I ordered another AMD Ryzen 5 1600 and an NVIDIA GeForce 1030 from Newegg on the morning of August 4. Unfortunately the CPU was not available for ShopRunner two-day shipping, but the GPU was.

My order arrived on August 8. I tested by building on top of the motherboard box. I powered the machine on, hoping for the best.

Just the same as before, the CPU light came on as soon as the board powered up.

I called GIGABYTE support and spoke with Robert again. I brought him up to speed on my situation. This time, he offered that a BIOS update may help move things along. Robert offered that he could give me a call back the next day, but I told him that I could drop him an email if there was any improvement.

I called Micro Center. They moved at the tail end of 2014, and I wanted to be sure that there was a space where I could tinker with the build in-store.

Meanwhile, I drafted a plan of attack:

Plan to start by swapping the CPU to a Ryzen 7. If the board POSTs successfully, I will upgrade the BIOS, swap back to the Ryzen 5, and test again. Hopefully that’s as far as I will have to go.

If the board fails POST with the Ryzen 7, I’ll swap RAM.

And if all else fails, I’m throwing in the towel, because I just might not have the patience for building systems 😀

On August 9, I headed to Micro Center with parts in tow. It takes about half an hour for me to reach the store. After parking my car, I went in, and headed for the Knowledge Bar, which struck me as being an appropriate first step.

I signed in there, then asked an associate who was standing by the laptops whether there was any space to tinker with builds, explaining that I’d brought my parts with me and wanted to get to the bottom of my no POST issue. She told me that the old store had a designated space for tinkering, but there was no such space in the new store, and the Knowledge Bar was probably my best bet. I thanked her and moved on.

I headed to the Build Your Own PC counter, where I explained my situation to an associate working that area. He asked me which motherboard I was using, and remarked that it was picky on RAM. He told me that I could test at their counter, so I went back to my car to retrieve the parts.

The associate brought me a module that was known to be compatible. Running that lone stick of RAM resulted in a successful POST, and I was greeted by the GIGABYTE UEFI for the very first time. The board was running BIOS F6.

That stick of RAM had relatively loose timings, so I opted for something tighter. I verified that the board would POST using my newly selected RAM, paid, and went back home.

I installed Windows 10 Professional – the process went smoothly.

I was outside of the return window for the Patriot RAM that I initially ordered, but I decided that I ought to test it further now that I had the system working. Was it DOA?

With only the Patriot module installed, the board predictably failed to POST. The strange thing is that the board POSTs successfully with both modules installed, and all 16GB of RAM are addressable. I don’t understand.

Final Parts Listing

CPU: AMD Ryzen 5 1600 (Amazon.com)
Motherboard: GIGABYTE GA-AB350-Gaming 3 (Amazon.com)
Memory: 8GB Kingston HyperX DDR4-2400 HX424C15FB2/8 15-15-15-35 (Amazon.com, datasheet)
Memory: 8GB (1*8GB) Patriot Viper 4 DDR4-2400 PV48G240C5 15-15-15-35 1.2V (Amazon.com)
SSD: SK hynix SL308 250GB HFS250G32TND-N1A2A (Amazon.com)
PSU: EVGA 450 BT, 80 PLUS Bronze Certified (Amazon.com)
OS: Windows 10 Pro

Final Notes

My eagerness to get this build completed resulted in some unfortunate oversights, resulting in significant expenditure of time and energy to rectify.

Before placing the initial order for parts, I referenced the QVL. The memory that I ordered, PV48G240C5, does not explicitly appear on the QVL. However, PV48G240C5K does.

While the GeForce 1030 specifications (NVIDIA) indicate that it is capable of driving 7680×[email protected], the GIGABYTE GeForce GT 1030 Silent Low Profile 2G (GIGABYTE) that I ordered outputs a maximum 4096×[email protected] through its HDMI-2.0b port. The DVI-D port can drive at most a 2560×1600 display.

Furthermore, Newegg’s product listing for the GIGABYTE GeForce GT 1030 Silent Low Profile 2G erroneously stated that it had three ports: one each of Dual-Link DVI-D, HDMI 2.0b, and DisplayPort. Referencing the product images on Newegg would have uncovered this discrepancy, leading me to seek the ground truth at GIGABYTE’s product page, and avoiding the hassle of returning a product. Fortunately, I was able to return the GPU to Newegg without incurring a restocking fee.

Mini-ITX AMD AM1 Motherboards

I started researching available Mini-ITX AMD AM1 motherboards for my upcoming SFF NAS build. Admittedly, this work was all done in the hopes that the AMD AM1 “Kabini” platform would make for an ideal low-power, 24/7 home NAS. A visit to the FreeNAS forums and some investigation into the hardware recommended for use with FreeNAS has changed my mind, but I leave this table of Mini-ITX AMD AM1 motherboards here for anyone looking for it.

They all use DDR3 and all have two slots for RAM, as is expected with motherboards of this form factor.

Manufacturer ASRock ASRock ASUS MSI
Model AM1B-ITX AM1H-ITX AM1I-A MSI AM1I
Max RAM 32GB 32GB 32GB 32GB
HDMI? 1 1 1 1
DisplayPort? 0 1 0 0
DVI? 1 1 1 1
D-Sub / VGA? 1 1 1 1
Serial / COM? 0 0 1 0
Parallel / LPT? 1 0 0 0
SATA 6Gb/s 4 4 2 2
Onboard Audio Realtek ALC662 Realtek ALC892 Realtek ALC877-VD Realtek ALC887
Audio Channels 5.1 7.1 7.1 7.1
Onboard LAN Realtek RTL8111GR Gigabit Ethernet Realtek RTL8111GR Gigabit Ethernet Realtek RTL8111GR Gigabit Ethernet Realtek RTL8111GR Gigabit Ethernet
PS/2? 1 1 2 2
USB 3.0 2 2 2 2
USB 2.0 4 2 4 2
S/PDIF Out 0 1x Optical 0 0
Expansion 1x PCI Express 2.0 x16 slot @ 4x mode 1x PCI Express 2.0 x16 slot @ 4x mode 1x PCI Express 2.0 x4 slot 1x PCI Express 3.0 x16 slot @ 4x mode
Mini Card Slots 0 1x Mini PCIe 0 1x Mini PCIe
Notes Motherboard has header for additional USB 3.0 ports, additional USB 2.0 ports Supports either ATX or DC-In power supplies, motherboard has header for additional USB 3.0 ports, additional USB 2.0 ports Motherboard has headers for additional USB 2.0 ports

I was thinking it would be very nice to have the ASRock AM1H-ITX equipped with an AMD 5350. This, I reasoned, could get me a livable number of SATA ports, with some room for a PCIe SATA controller. Budget for the build, including case, would have come in at around $500.

While such a machine would serve well in a HTPC environment, it would not be ideal for FreeNAS. ZFS calls for a minimum of 8GB RAM, and prefers ECC RAM to non ECC. It is reported that the AMD AM1 platform should support ECC memory, though none of the Mini-ITX boards mentioned above have explicitly stated that they support this technology. Couple that with the fact that AMD AM1 “Kabini” APUs only allow for single channel memory, and I’m not immediately sold on this platform. However, that hasn’t stopped at least one person from building it..

http://www.techpowerup.com/forums/threads/am1-athlon-5350-freenas-file-server.201676/

He even uses the Silverstone DS380 Mini-ITX NAS case that I had been following, reporting that its stock fans are noisy. Maybe not for me!