Project H20-Micro a Watercooled 3 Liter system with GTX970

Bieberfever said:
If anyone (OP included) is interested in doing a custom/reproducable ultra SFF gaming case I have one of these (Carbide 3D) coming in the next month or so for the purposes of making such a case. Could use some help/expertise.
Click to expand...


haha, I have one too for my SFF case, currently sitting in the kitchen. But I'm about to jump on my project soon!
 
qwerkus said:
Awesome reply: thanks a lot; the latter build is approximately what I'm try to emulate ! So with a gtx 750 ti, 4x will be totally fine. Will try to post some pics once it's done :) I found a french site dealing with pci-e 3.0, and they found a difference of about 1-2% only with gtx 980 cards. So cards are slowly pushing towards the maximum capacity of x4 lanes, but nowhere to the x8 lanes. So x16 is only useful for professional cards ? Now I understand why QinX isn't worried at all plugging his 970 onto a 4x riser...
Click to expand...
Why, WHY could I not find this forum months ago?
So, my build uses the same ASUS Q87T Motherboard, with a x4 to x16 flexible adapter cable so I can move my graphics card around a bit, and I'm trying to run a EVGA GTX 750 TI SC.
According to the EVGA website, that model takes 60w.
Based on this article on a different forum it sounded like the x4 slot should be able to provide the required 60w for the GPU:
GTX 750Ti on a X4 PCI-E port

So I hooked everything up.... and it worked! Kinda....
At first it was merely my benchmark software that would throw errors and say it failed.... eventually that went away (IDK what I even did to get past that anymore) but when I started doing testing with games, they all randomly crash themselves or the 'graphics driver' crashes, crashing the games.
I believe it is a power issue, especially now.
To make things a bit more complicated, my PSU is only 150w via a brick connected to the external DC connector.
Now, I'm only using a i3-4330.... which is rated at 54w TDP, so with everything else (20w peak for RAM, 5w for mSATA SSD, we'll say 10w for mini pcie wifi/bluetooth card?, +USB, +Motherboard?) even if the PCIe x4 did provide the necessary 60w, my PSU would be cutting it very close....

So my conclusion was to buy a larger PSU, and I was eyeing up a 180w one, since that should push me over the edge enough.... but now I'm wondering if I need something even bigger, and to modify my PCI riser cable to inject more power....? Thoughts?
 
The x4 riser may be you problem. The PCIe slot can provide up to 75W through bus power, but only for the x16 interface, with the limit being lower for x4/x8 and x1. Depending on how your motherboard's BIOS/EFI is configured, the card may request 75W with its 4X link, and the motherboard may simply say "NO". It's entirely down to whether you motherboard sticks to the letter of the specifications, or will go over and above what they require.

Second point of suspicion is the external DC brick. If it;s a laptop brick, those are designed to provide a sustained constant load to charge a battery. A CPU or GPU under load is not a sustained constant load, it is constantly and repeatedly drawing high power spikes then dropping down again. Your average load over time may be under the 150W limit, but those load spikes may jump well over that,
For a normal PC power supply this is not an issue at all, the PSU is designed to handle that as a requirement of being ATX compliant. A laptop brick is not, so may shut down when it encounters a power spike. This is a notorious problem with the R9 Nano in compact cases, so you may be finding the same problem with the 750Ti and your smaller power limit.[/I][/I]
 
EdZ said:
The x4 riser may be you problem. The PCIe slot can provide up to 75W through bus power, but only for the x16 interface, with the limit being lower for x4/x8 and x1. Depending on how your motherboard's BIOS/EFI is configured, the card may request 75W with its 4X link, and the motherboard may simply say "NO". It's entirely down to whether you motherboard sticks to the letter of the specifications, or will go over and above what they require.
Click to expand...

My interpretation was that all the power pins were still in the x4 pins, that only PCIe graphics cards can request more then 25w and since it is indeed a graphics card it can still request 75w, and since it is the only PCIe slot that it might actually be able to supply the 75w to the x4 slot. Again, that's my interpretation, so I could definitely be wrong.
Unfortunately I can't find anywhere in the manual or in the BIOS about power distribution to the PCIeX4, but I did force it to use V3 instead of 'Auto' and oddly enough it seems to be more stable today then it was yesterday.

EdZ said:
Second point of suspicion is the external DC brick. If it;s a laptop brick, those are designed to provide a sustained constant load to charge a battery.
For a normal PC power supply this is not an issue at all, the PSU is designed to handle that as a requirement of being ATX compliant. A laptop brick is not, so may shut down when it encounters a power spike.
Click to expand...

This is the PSU:
https://smile.amazon.com/dp/B00F4O120E
I *says* it is a Mini ITX Power Adapter, not a laptop brick, but like you mentioned that doesn't mean it's actually built to that spec and could be relabeled laptop brick.
 
J1024 said:
My interpretation was that all the power pins were still in the x4 pins, that only PCIe graphics cards can request more then 25w and since it is indeed a graphics card it can still request 75w, and since it is the only PCIe slot that it might actually be able to supply the 75w to the x4 slot. Again, that's my interpretation, so I could definitely be wrong.
Click to expand...
All slots can draw up to 25W by default (10W for 1x). If a card is a 16x full-height (half-height cards are limited to 25W) card, then it can request that be updated to 75W. The Motherboard is responsible for responding to that request. Technically, to comply with the PCIe standard if the interface is 4x then the motherboard should not agree to switch the slot to high-power mode. This is regardless of the number of physical power traces connected, which will be the same for all bus widths.
This is the PSU:
https://smile.amazon.com/dp/B00F4O120E
I *says* it is a Mini ITX Power Adapter, not a laptop brick, but like you mentioned that doesn't mean it's actually built to that spec and could be relabeled laptop brick.
Click to expand...
FSP's spec page is very clear on a 7.9A maximum current limit for this PSU, which at 19V is 150.1W and with no mention of a higher peak current capacity.

The Asus Q87T spec page and manual fail to list the input voltage (Asus FAIL) so I don't know whether you could test with a full ATX PSU, but if you have a higher wattage power brick you could try that.
 
EdZ said:
Technically, to comply with the PCIe standard if the interface is 4x then the motherboard should not agree to switch the slot to high-power mode. This is regardless of the number of physical power traces connected, which will be the same for all bus widths.
Click to expand...

Also keep in mind that the mainboards traces will could only be wide enough to handle 25W, not 75W, so even if the pins could handle such an amount of power to be drawn, the board could fry itself if it allowed that to happen.
 
iFreilicht said:
Also keep in mind that the mainboards traces will could only be wide enough to handle 25W, not 75W, so even if the pins could handle such an amount of power to be drawn, the board could fry itself if it allowed that to happen.
Click to expand...

Or even more annoying cause a significant voltage drop that only happens when you pull a lot of power.
 
IDK why I thought calling ASUS might actually help. Their 'Technical' Support insisted that the motherboard would be able to power my (theoretical) 500w GPU directly from the PCIe slot, as long as I had a big enough PSU. Fail. I'm going to try getting a 250-300w PSU for it.... if anyone has any recommendations, let me know. I might end up having to modify my PCIe x4-to-x16 cable though, as OP was talking about. :-|
 
Another idea for a cause might be the mb bios trying to run pcie gen 3, but the cable you are using can only reliably run at gen 2. Maybe change to gen 2 in bios.

Ive read about faults and such happening due to this.
 
I have no first-hand experience with ASUS support but I keep reading that their support is hit-or-miss at best.

I'm thinking of doing something similar for powering a graphics card with a thin mini board, just as a stock air-cooled setup. There are various ways you can go about it- using a m.2 to PCIe adapter card, or a mini PCIe to PCIe. But for the mini PCIe, I've only been able to find PCIe 2.0 1x slot adapters. Current cards could bottleneck with that speed.

I also have a leftover powered riser cable from my bitcoin mining days, the kind that have the USB connector for data and molex for power. I'm not planning on using a card that consumes more than 75W so I do not require an additional 6-pin cable.
 
gotbass said:
Are you still working on this project or a similar SFF build?
Click to expand...

I'm not him, but I don't know how active he is on [H] so I'll answer. This project is on hold for now and might not be revived for quite some time to come, as QinX is working on building a business around a little simpler products and this one is too expensive and time intensive to make right now.

You can take a look at the stuff he's doing now here.
 
You must log in or register to reply here.
Back
Top