Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

Demonic Donut

Member
  • Content Count

    317
  • Joined

  • Last visited

Everything posted by Demonic Donut

  1. Don't use afterburner with Radeon. Use the Radeon program. Download MSI kombuster. Open Radeon Software and go to the Performance/Tuning tab. Click the sliders to unlock manual tuning. If you want to OC, bump up GPU frequency 50 Mhz at a time, using kombuster stress test at your resolution for at least 5 minutes. When you get a crash, drop the OC by 10-20Mhz or so and see if it's stable in your games. Frequency setting does not mean that's the frequency you will achieve, frequency fluctuates based on load and power draw. Undervolting is great for lowering temps and achieving higher clocks when power limits and temps are your limiting factor. To undervolt, leave frequency alone and lower voltage 10-20mV at a time and stress test until you get a crash. Then bump it up some and retest to verify your undervolt is viable. To OC ram, do the same thing as with GPU speeds but with the memory speed slider. I adjust power sliders only if I have temperature headroom. My stock sapphire pulse 5700XT can only manage 10% higher power limit before temps get too high for my liking. My card can't undervolt at all, but I recently put an aftermarket cooler on and can OC much farther now. All cards are different, so it's a lot of trial and error. My stock card OC: 2050 @ 1200mV (cant remember middle frequency/voltage setting but the graph has a slightly increasing curve) 1825Mhz RAM 10% Power Limit
  2. If you put it all back together and they don't notice, you will be ok. If they know you took it apart and broke it, they won't cover it. If they won't cover it, you'll have to repair the header yourself. *Edit* If they notice, you can try the ignorance method and say you bought a new fan and it didn't work either.
  3. When you unplug them (any connection really). You need to hold onto the plug that is soldered to the board to support it from flexing/pulling while you pull the wire plug from the fan out. When you plug them in, hold the plug again and support it from being flexed/pushed by the fan plug insertion.
  4. Generally connections on laptops are rather flimsy. They aren't designed to be unplugged/plugged in again. You have to support the fan and board side when you remove them.
  5. I'd inspect to see if there is any visual damage and go from there. Most likely, soldering will be involved. Its not extremely difficult, and generally easier with a good soldering iron.
  6. I'd guess you damaged the fan header.
  7. Both mine and my wife's Sapphire Pulse 5700xts are on the newest non-beta driver. The only issue I've had since owning the card (November 2019) was bad custom fan control. I always update to the newest stable driver. I've never had black screens or driver issues that caused instability. Use Radeon software only for OC and fan control etc. Use PCIe gen 3 like @rijzen recommended. If you need monitoring software, don't use afterburner, use HWInfo. Update your motherboard bios too. Chipset drivers etc.
  8. You might need new thermal pads if they are squished/damaged at all or if they got dirty from dust or your finger oils.
  9. 109 is getting really hot. Do you know if it was die junction or memory or vrm?
  10. Try using FanControl, in this sub forum. I love it. It allows you to control any fan on any temp sensor. Tons of options for fan control. I have my side intake running on GPU die temp, my bottom and rear fans are set to a mix of CPU and GPU (goes to the higher speed based on which temp is higher), and my top and CPU tower fans are just on CPU temp. You can adjust hysteresis, min/max speeds, linear curves, mixes, fixed speeds, synchs (offsets) and all kinds of stuff. If you want to stick with motherboard control, I find it's best to just set your case fans to the highest speed you are ok with noise wise and leave it alone. Or if you have a temp sensor option for the motherboard, you can mount it on or near the GPU and play with that for control.
  11. Run/record HWInfo while you run a test and see what temps are at.
  12. G1/4 is a thread standard. Like NPT for the USA. So all G1/4 are the same. Then just verify tube sizing. Remember some tubing is thicker/thinner walled etc. Mix brands, just don't mix aluminum with copper. Or steel for that matter. Copper can be mixed with brass though. When picking a rad, find the size, thickness and FPI you want and buy the one with good reviews and price.
  13. Definitely get the 3300x. Get a 3600 if the price is close to the same. Last I looked, the 3300X wasn't available under $200 and the 3600 was $175.
  14. I don't plan to do anything crazy with the GPU. Perhaps use More Power Tool to lock in the OC without needing to open Radeon Software, but I'm going to stick with the 1200mV at 2100Mhz target frequency. I found 2150Mhz target (max allowed in Radeon app) gets glitchy in RDR2. I also don't want to kill my card yet, I've only had it since November.
  15. That's the only thing I had seen before as well. From what I've read, these coolers are popular with the Vega cards.
  16. TL:DR and Temps at the bottom. A couple of weeks ago I stumbled across the Morpheus II cooler and grabbed it on a whim. I had never seen an aftermarket GPU air cooler and read some good things about it. Why not buy one and have something new to play with? I figured worst case, I wasted $85 (after shipping from the UK) More likely it'll be quieter, at least. But maybe it'll perform better and let me push my card a bit further, although I was a little worried about vrm and memory temps. Sapphire's cooler works well in that department, especially if you sacrifice some noise. But it really hits a wall with die temps. I found extreme diminishing returns above 50% fan speed, and it screams like a banshee. I decided to use 2 Noctua P12 redux fans. I would have liked to try some Arctic P12s, as I've heard good things about them, but I had the Noctuas already. That 5 pack of Arctic's for $30 on Amazon is tempting though... I also purchased some alphacool copper heatsinks for the VRMs and RAM (10x10 and 14x14), but I wasn't sure if they would fit under the Morpheus. I did everything in steps to establish how much each change affected temperature. My temperature benchmark was Furmark at 1440p with all stock GPU settings, except fan curve. Case fans were locked at my max curve speed, 60%. I let Furmark run until temps settled out, but at least 15 minutes. My ambient temperatures will be noted, but are usually at 70F +/- 2 degrees. I also pulled the cooler when I got the card just to look at the cooler design, and repasted with Noctua H2 paste so it isn't technically stock. I found no temperature change from the repaste though. Evening One: I pulled the backplate and fan shroud, then mounted the two P12s with some zip ties to the stock cooler and started up Furmark. The first thing I noticed was that the new fans are roughly equal in noise at full speed to the stock fans at about 40%, and with a lower pitch sound that I prefer. I ran my stock fan curve at 50% max and was satisfied with temps but not with noise, so this alone was a great upgrade for me. Memory temps saw a nice improvement at the same noise level, but not compared to stock fans at 50%. I noticed that die temps were pretty close. Fan speed maxed on the stock fans gives similar die temps to 50%, but I never paid close attention to memory or vrm temps at high stock fan speeds. Evening Two: I pulled the card all the way apart and started test fitting the included heatsinks to the memory chips and VRMs etc. Unfortunately the Morpheus didn't come with enough of the right size/shape Heatsinks to apply to the RAM, VRM and various other chips that the stock sapphire cooler contacts. The Alphacool heatsinks hadn't come yet, so I applied 1mm Thermal Grizzly minus 8 pads to the chips and reinstalled the stock cooler with the Morpheus included thermal compound. I wanted to see if there was any difference in thermal pad quality. The Morpheus thermal compound gave me about 5C higher temps and a larger Delta between junction and average temps, so I decided to pull the cooler again and repaste. After some isopropyl and gentle scrubbing with a soft toothbrush to clean the extra goop off, I repasted with some older Noctua H1 I had left over. Disaster! It wouldn't post. After clearing CMOS and reseating the GPU, I was worried I killed the card somehow. I would get the Aorus splash screen, then reboot. Couldn't enter bios as well. It was late and I headed to bed. I would do troubleshooting the next evening. Evening Three: I had a stroke of luck, my tower booted up and everything was normal. My first guess was some unevaporated Iso had been causing an issue. Junction temps ended up even worse than with the Morpheus Paste. I thought I had a mounting pressure issue or bad paste job. Pulled the cooler again, paste looked good but I believe I zip tied my fans too tightly and tweaked the cooler and PCB just enough to mess with die contact. Cleaned the board up and started prepping for copper heatsink application. When cleaning again, I found that Iso will wick under the PCB layer around the die and takes some encouragement to dry/get out of there. I got as much out/dry as I could with compressed nitrogen. Evening Four: Alphacool heatsinks arrived. They definitely get a thumbs up from me. Well packaged, well made and not too expensive. The RAM heatsinks won't clear the cooler in all places though. Heatpipes and mounting hardware interfere somewhat. I attempted to use a bandsaw to trim the heatsinks quickly… I don't recommend it. Once you hit the fins, they bend/bind and rip the heatsink out of your holding implement. A very fine saw might work, especially hand powered, but I don't have one. I then switched to tin snips, and while it's not the prettiest, it's functional. Most of the fins I could simply bend out of the way, but some required trimming. I don't love the VRM solution Morpheus provides, but it works. I couldn't find a copper VRM heatsink and the copper ones I have are too big. Once the heatsinks were trimmed and passed test fitment, it was on to attaching them. I used the included double sided "thermal tape" with pretty good luck. There is plenty of extra to use if you mess up or if it moves and touches you etc. Once they were all attached, I applied conformal coating (2 coats) to the die area in preparation for TG Conductonaut. Evening Six: Getting close! Heatsinks attached and seem well adhered. Conformal coating is dry and everything is ready for the cooler to be mounted. Spread TG Conductonaut on the die and cooler and mounted to the card. Attached fans with included clips (which kind of suck and barely hold the fans on) and seated the card. On to final testing! Wow. This cooler is impressive! Memory and VRM temps are a bit higher, but those die temps! Played with overclocking the card and was thoroughly impressed with temperatures. I am running the same OC I did before but with power limit increased to 25%. I don't love the memory and VRM temps, I want better performance everywhere. I found Arctic makes a thermal adhesive, and reports say when cut with their same version of thermal paste (Arctic Alumina), you can make a semi permanent bond. The adhesive is rated at 8 w/mK, which is the same as the Thermal Grizzly thermal pads I have and the material layer will be thinner. If anyone has a source for a good copper GPU VRM heatsink, let me know. The aluminum one is fine, but it's aluminum and not copper. I'm going to purchase more low profile RAM heatsinks as well so I can get rid of a couple of my hacked up ones. TL:DR I wanted to try to improve the temps on my Sapphire Pulse 5700XT with new fans and a custom cooler. Performed tests at each stage. All tests done with case fans locked and stock GPU bios with only GPU fan speed being modified/held. Furmark was ran until temps stabilized, then HWInfo reset and recorded for 5 minutes. Stock Card - 40% Fan Speed Ambient Temp: 70F GPU Temp: 74C GPU Junction Temp: 87C Memory Junction Temp: 88C VRM Temp: 75C Average Clock Speed: 1803Mhz Average Power Draw: 193W Stock Card - 50% Fan Speed Ambient Temp: 70F GPU Temp: 71C GPU Junction Temp: 84C Memory Junction Temp: 82C VRM Temp: 69C Average Clock Speed: 1808Mhz ASIC Power Draw: 193W 2 Noctua P12 Redux on stock cooler - 100% Fan Speed Ambient Temp: 71F GPU Temp: 71C GPU Junction Temp: 85C Memory Junction Temp: 84C VRM Temp: 69C Average Clock Speed: 1811Mhz ASIC Power Draw: 193W Morpheus Core II Stock Settings Ambient Temp: 74C GPU Temp: 53C GPU Junction Temp: 65C Memory Junction Temp: 88C VRM Temp: 70C Average Clock Speed: 1819Mhz Average Power Draw: 193W Morpheus Core II High Power Limit OC Ambient Temp: 74F GPU Temp: 61C GPU Junction Temp: 80C Memory Junction Temp: 96C VRM Temp: 88C Average Clock Speed: 2060Mhz Average Power Draw: 255W
  17. We use a laptop so a USB dongle is fine. Or anything else a laptop can handle...
  18. My wife and I occasionally watch TV in bed on our laptop. Currently we share in ears like a high school couple, and while it works, it's getting old. One of us moves and pops the other person's out, a dog or cat catches the cord while they move and it unplugs and audio plays over the speakers and wakes up the baby etc, etc. I would really like to go wireless, but am not sure if we'd be happy with them. I absolutely despise out of sync audio, and I'm worried I will hate Bluetooth headphones. Once I notice out of sync audio, it's all I can notice and ruins the show/movie for me. I know some wireless over ears are low latency because they aren't using Bluetooth, but I want in ear for comfort when lounging in bed. Is Bluetooth time delay a thing of the past? From what I've read, it is decent but not great. Are there faster wireless standards for in ears? I'd like to keep cost under $100 or so. My cheap option is to use a 3.5mm splitter. It will help with some of our problems but I'm still stuck with a cord that can get caught and yanked on. Any and all suggestions welcome.
  19. Nothing can be done. Anecdotally, people have had luck with a better power supply. Buildzoid mentioned in a video that coil whine happens the most when loads change or travel through a certain point. Makes sense to me considering it's a harmonic vibration. You can try to overclock/underclock and/or overvolt/undervolt to alleviate it. Only other option is to replace components.
  20. 100ms is acceptable for me for online gaming. But 50ms or lower is my personal sweet spot, especially for twitch shooters and other games where reaction and timing matters a lot.
  21. As long as the depth of the cold plate etc is the same, it should give better temps.
  22. Don't be afraid of liquid metal. Use some conformal coating and you'll be fine. Its cheap and easy to use. On cutting heat sinks. Try to find one that matches for height and width. That way you aren't cutting fins. I recently tried cutting some alphacool copper heatsinks for a GPU project with a band saw... Didn't go well for me. Too soft and they bend and bind to the blade.
  23. If you know the spec of the part and are good with soldering you can certainly try. I've heard people have tried to use hot glue to dampen coils, but I don't know if it worked. My old Mobo had coil whine. Anytime it was accessing the SSD I could hear it... New computer isn't as bad but I still hear electrical noise, not from my GPU though. I'm pretty sensitive to high frequency, so it can drive me nuts. When I met my wife, she had an old tube TV and I threw it away and bought a flat screen for her simply because I couldn't stand the noise from the TV.
×