Topic: Risers 103c fried my GPU's, be aware (Read 482 times)

I would not judge the quality or load handling capacity of a riser board by the number of capacitors mounted on it. There is little if any benefit of having extra 220uF hanging on a 3.3V rail which sees minimal load. Quality of those caps is highly questionable to be honest.

Obviously I'm saying this based on what I see on the picture in OP. While it's definitely a new and likely better design, their sloppy design decisions tell a lot about quality. For example, looks like they wanted to mount a fuse on the input 12V rail, but ended up placing a diode instead (part labeled F1). This will result in measurable sag on the 12V rail. Then, instead of a diode (D5) they have a large ceramic capacitor mounted, which may help with stability of the DC-DC converter. D6, D7 and D8 also look like caps to me and diodes don't belong there anyway. Its hard to say, but I bet it was an amateur job. Still beats the ugly VER006C we all use and pretty happy with and it even comes with an LED lol.

The VER:008C looks a lot better and at least comes with all parts mounted matching the silkscreen designators... Note that I haven't tried them yet. Just going by the looks...

any riser will burn if you connect it improperly/overload the lines, regardless of connector type.

yeah they check them, modified bios voids the warrenty

also, personally my own little tip is to get risers with 4 capacitors (your burned ones have three i see)
those tend to handle loads better
I mean, SATA is a bad idea always imo anyways, go PCI-E for no worries, but get better risers man, they are a big failure point

Lesson Learned!

molex riser is not a good idea either, they tend to burn espacially with AMD Cards.

It is always better using molex directly from PSU to power risers, using Sata conectors is a bad idea, not recomended.

How many SATA connectors per cable did you use?

You can only use the 2 sata slot to powerup your riser, using above that number will fry your cable and it could be harm with your gpu.

I once fried a PPro 200 CPU by dropping a screwdriver onto a motherboard...

Your powering scheme had probably nothing to do with the death of the CPU. As long as your ground levels on both PSUs were close enough, it should not have caused any problems. While in *most* cases it's fine and some people strongly believe it's The Proper Way to power risers (from the same PSU as the motherboard), in *some* cases it may become a major problem.

For example if your GPU has PCIe and AUX 12V rails connected internally (which is unlikely) and the actual voltages produced by PSUs are different (likely), you will have rather high current flowing through motherboard PCIe ground plane, USB3 cabling and risers, and also the GPU 12V and GND traces which may lead to a meltdown of various degree of severity.

The only proper way of powering GPUs with risers is to treat them as a single "unit" and use the same 12V power source for both, the riser and the GPU. Then you just connect this "unit" to the host (motherboard) via USB3 cable which is not supposed to carry any power in this topology.

Think of it as if you would connect your TV or a monitor or whatever via HDMI or DP or USB-C LVDS cable. Actually they are all very similar to the PCIe interconnect on the physical link level, which is in part the reason they used those off-the-shelf USB3 cables for PCIe riser extension. Even clocks and frequencies are pretty much the same for PCIe 2.0 and USB3.0...

Just sharing my experience and hope to get some valuable comments.

Currently all the risers 103C on 6pin and GPUs are powered by HP Server PSU 1200W; a Corsair 750W PSU power the motherboard and HDD. This has been running flawlessly for over a month. I had earlier power all the risers thru the Corsair PSU using both sata and molex connectors and the HP PSU to power the GPUs. This method also works for me.

However, I have once fried the CPU (ya CPU, not GPU) while all GPUs and risers powered by HP PSU except for 2 risers powered thru the Corsair PSU using sata. I'm not sure if this is not proper perhaps someone could share you thots.

After hearing your incident (TS), I'm not worried I mean I'm getting a bit worried because all my other spare risers are 103C  

Good question which got me thinking... In all likelihood they produce 3.3V from 12V input using a step-down regulator. On the more recent risers I have on hands (ver: 008C) which also lacks the 3.3V LDO (it's not a MOSFet), they use a newer FR9888 step-down DC-DC IC which is specified up to 93% efficiency.

I still want to understand why up until now they were using two regulators (12V->5V and 5V->3.3V) on all typical ver:006C riser boards. I was under the impression that 5V was required by PCIe bus, which it s not. Strange...

You probably right. If a GPU manufacturer design is lazy enough and uses 3.3V rail straight from PCIe connector to save a few cents on yet another step-down circuit, it may very well blow up in an OV condition.

For the sake of clarity, there is no 5V power going to the GPU. I was mistakenly assuming PCIe has 5V rail, just like the old PCI bus did. It does only provide 3.3V up to 3A plus the 12V of course.

Honestly it's beyond me why they still use a 3.3V PCIe power rail on a GPU which already has like four or five step-down regulators hanging off of the almighty 12V rail...

Speaking of those OP's deadly risers... If that PCB is indeed misrouted, it has to be a really gross error and not a simple input 12/5V wire swap. They must have routed 12V into some PCIe signal pins or straight through to the 3.3VCC rail in order to fry a GPU with it. The DC-DC step-down converter can take a lot of abuse and will definitely survive a 12V instead of 5V input.

This model of risers, how they work without the 3.3v mosfet? they get the 3.3v from the usb cable? if thats true then you can't use dual psu with then.

If you plug a GPU into broken ass PCBs and blow them up, you have no reason to expect a warranty replacement on the GPU.

If, for example, they did route the +12 into the 5V, yeah, that's gonna blow stuff up, and probably not something any GPU manufacturer designs for.  To do so would cost an extra $15 per card, at least, so that users can plug them into poorly designed $6 risers...

I'm not OP, but wonder what do you mean by mistake in this context?

IMO a properly engineered power delivery circuitry on a decent GPU would've survived a reasonable overvoltage mishap... In the end of the day these are just DC-DC stop-down converters and the good ones do have fairly wide input voltage ranges.

BIOS mods are another story I don't want to touch with a 6-foot pole...

They will however deny your warranty for being stupid.

You dun ****ed up, you have no right to make someone else pay for your mistake.

No problem, I think they will not check the bios.

Yes always the same PSU to power riser and GPU.

Im going to RMA those cards maybe any chance for raplace, too bad they had modified BIOS, do you think Sapphire check what bios was on the fried card?

Were you using same PSU to power your riser and your GPUs?

Can also be a misrouted riser PCB. In that case they will fry GPUs consistently by supplying 12V instead of 5V or even reverse voltage for example. Sorry for your loss and thanks for the heads-up!

Ok but why then risers V006C with cable 6pin >>> Sata works like harm? Made more than 20 rigs with them all powered through SATA and never had any problem, the same with those old red riser only with sata port, they work even with 1080ti.

If there were not enyough power they should not see the card, not burn them i think