Jump to content


Superherø MPC
  • Content Count

  • Joined

  • Last visited

  • Days Won


Everything posted by Dasa

  1. Things may be looking up for AMD if they can deliver http://www.tomshardware.com/news/amd-zen-cpu-radeon-hbm,29053.html http://www.anandtech.com/show/9231/amds-20162017-x86-roadmap-zen-is-in http://www.overclock3d.net/articles/cpu_mainboard/amd_2016_roadmap_confirmed_8_cores_are_back/1
  2. For the price and my use I think it was the best choice second best choice would have been to pony up the extra $500 for a 2080ti
  3. https://www.kotaku.com.au/2019/02/radeon-vii-amd-australian-price-release-date-specs/ AMD have revealed this afternoon that their new card, which will ship in multiple configurations with a baseline of 16GB in HBM2 memory, 3840 stream proessors, Freesync 2 HDR support and 1TB/s of memory bandwidth, will be available in Australia from $1069. It's also been publicly revealed that the Radeon 7 will run at a boost clock speed of 1800Mhz Availability on all cards in Australia and New Zealand is unknown, although the Radeon 7 launches internationally February 8 Australian time. We'll update this post once a local release date is confirmed. AMD Radeon VII Unboxing & Preview The main die on the "Vega 20" MCM is the centrally located GPU die built on TSMC's 7 nanometer silicon fabrication process. Measuring 331 mm², this silicon packs 13.26 billion transistors, which is a similar transistor count to NVIDIA's TU104 graphics processor. The "Vega 20" GPU die features a 4096-bit wide HBM2 memory interface, which has double the bus width of "Vega 10". This means the GPU has to be surrounded by four memory stacks instead of two. On the Radeon VII, there are four 32 Gbit (4 GB) stacks built on a 10 nm-class node supplied by either SK Hynix or Samsung. We don't know the process node of the interposer, but given it was 65 nm for 4096-bit "Fiji", we can't imagine why it would change for "Vega 20". The interposer has no logic of its own and draws no power for itself. Its only purpose is to provide miniaturized wiring between GPU and memory stacks, and as such, there's nothing to be gained by changing its fabrication process if the desired wiring density has been achieved. A metal-reinforcement brace runs the periphery of the fiberglass package substrate to distribute mounting pressure from the cooler. AMD has used conventional-looking thermal pads over the VRM MOSFETs, which are somewhat reusable. The GPU interface is a whole different story. Defying conventional logic, AMD went with what clearly feels like a high-conductivity thermal pad between the vapor-chamber base and GPU instead of thermal grease. We believe this to be a strip of Hitachi TC-HM03, a high-density thermal pad with graphite strands and a rated conductivity of 25-45 W/m·K, which is higher than the typically 12.5 W/m· K offered by certain diamond-based high-viscosity pastes. Going by the lengths they have gone with the cooling I am guessing they are once again pushing the clocks as hard as they can to try compete. That or 7nm just runs hot so it will be interesting to see if there is any overclocking head room left and just how much power it draws. Considering the RTX 2080 which will be priced similar only needs 6pin+8pin power vs 8+8 I think it is safe to assume that despite the 7nm process it will still be less efficient but then you do get twice as much memory.
  4. By ~100MHz I guess there is something about the balance of shaders ect. that give the Radeon VII a advantage in FireStrike and something in TimeSpy that gives the 2080 a advantage over the Radeon VII and 1080Ti
  5. Finally got my Radeon VII on water although it is not as neat as I had hoped. @2200c 1.293v +60% 1275m (EvenMorePowerVII_1293+) Firestrike Graphics Score 33,246 http://www.3dmark.com/spy/7252701 Timespy Graphics Score 10,459 http://www.3dmark.com/fs/19401306 Max temp 25C core 55C junction 16c Idle These temps are just from 3dmark not from looping something like metro. Storm G5 is now mounted on the Radeon VII with a thick application of grizzly kryo. Ended up making another plate to go between the GPU and the Block so it can use 4 small bolts with 4 large bolts, it's not pretty but it does the job. The CPU block was designed for direct core cooling and is made out of 99.9975% pure silver, it is fairly restrictive with it's jet's. The holes in the base for the water jet's are about the size of the core and the HBM sits off to each side a little. Used some old stained Tygon tubing as it was kinking the least out of my tube selection when trying to fit it in between the GPU and the PSU. Looped metro at 2000c 1138mv +20% 1130m Max temps Core 28c Junction 53c Idle 16c
  6. The discussion popped up on another forum about why AIO don't perform as well as some might expect and the common thought was low water flow but that's just a small part of it. A mate I play games with was good enough to send me his old H80 to play around with and here are the results. 6700K@4.6GHz 1.375vcore Prime95 FTT 16k Corsair H80 Lapped with 220CFM Delta and old fan with guts removed used as shroud. TIM is Thermal Grizzly Kryonaut. Ambient 17-18c Water Idle 18c Rad inlet 23.2c Rad outlet 19.8c Idle CPUTIN 20c Core\Package 15-22c Load CPUTIN 42c Core\Package 88-92c <Hits 88c within about one second of prime95 load So the low flow rate is only adding ~3c to temps with the water being 3.4c hotter after it has been through the water block but this is only a 6700K using ~150w not at 9900K. With the 220CFM delta the rad is ~3c within ambient which is about as good as it gets even with a custom loop so a bigger rad or more airflow wont help. So the problem is in the mount and bad IHS contact guess I will see if I can mod the waterblock for increased mount pressure and maybe delid my CPU. Remount with washers between backplate and MB for increased mount pressure, also used a little less TIM. Ambient 20-21c Water Idle 20.5c Rad inlet 27.2c Rad outlet 23.6c Idle CPUTIN 22c Core\Package 17-23c Load CPUTIN 42c Core\Package 86-88c <Hits 85c within about one second of prime95 load The CPU core delta shows that the CPU temp has dropped by about 8c after the remount with increased pressure. Now that we have water flow rate out of the way let's see how the stock fans perform Fan's off Phone is reporting ~15dBA Idle fan noise With h80 set to low Phone is reporting ~17dBA <Clearly audible but quiet. Idle fan noise With h80 set to medium Phone is reporting ~24dBA <Annoyingly loud. Stock fan's@ medium speed using corsair AIO speed control Phone is reporting ~38dBA (Fairly loud similar annoyance to the Radeon VII) Corsar 2x120x25mm 92CFM@2500 RPM 7.7mm/H20,39dBA – Phone is reporting ~47dBA Detla 1x120x38mm, 220.01CFM@4600 RPM, 65.0dBA – Phone is reporting 63-65dBA Stock fans 2X120~2500RPM Ambient 21-22c Water Idle 23.2c Rad inlet 30.2c Rad outlet 26.5c Idle CPUTIN 29c Core\Package 20-27c Load CPUTIN 46c Core\Package 89-92c <Hits 85c within about one second of prime95 load Stock fan's@ medium speed using corsair AIO speed control Ambient 21-22c Water Idle 24.5c Rad inlet 33.3c Rad outlet 29.5c Idle CPUTIN 29c Core\Package 20-27c Load CPUTIN 51c Core\Package 94-96c <Hits 85-90c within a few seconds second of prime95 load Obviously this is where the cheap little rad struggles with lower fan speeds. The rad is only 15FPI by my count and has 12 water channels, apparently it is supposed to be closer to 20FPI so there was obviously some margin for error in there production. The rad is 42mm thick but the core is only 38mm leaving just 2mm between the fan and the fins.
  7. It's alive! dam thing loves to make my heat stop blank screen first boot. Storm G5 is now mounted on the Radeon VII with a thick application of grizzly kryo. Ended up making another plate to go between the GPU and the Block so it can use 4 small bolts with 4 large bolts, it's not pretty but it does the job. GPU at 2060MHz 1216mv Max temps during first run of TimeSpy 36c core 60c junction Idle 29c core 30c junction Attached Thumbnails
  8. Any idea how liquid metal reacts with 99.9975 % pure silver?
  9. 3770K will increase CPU performance in ACO by ~10% going by this https://www.dsogaming.com/pc-performance-analyses/assassins-creed-odyssey-pc-performance-analysis/ Liquid metal under the IHS is fairly safe but I wouldn't be in a hurry to use it elsewhere. Although I am tempted to try it on my Radeon VII. (I also spread it over one of the resistors on my GTX1070 to increase it's powertune %) Temps on my 3770K dropped by ~20c after the delid.
  10. It's only $139 for a Bykski full cover from PLE but people are still hitting around 70c hot spot with full cover blocks and I was hoping to see if i could get a little less. I am going to try get the storm on first and see how it's old silver ass performs.
  11. If you can delid the 3770K and stick some Liquid metal under there you will get a better OC out of it before it overheats. I had a bitch of a time getting mine off though and almost killed the CPU.
  12. Last I looked 3770K were still going for ~$150 online. It will get you 4 extra threads thanks to HT which can boost performance up to 25% in some things and reduce the chance of stutters in games that need more than 4 cores.
  13. Inlet\outlet are to close to screw 90 fittings in but maybe I could fit one and my snake. Drilled out the mount plate today maybe 1mm out with the holes and the other worry is that the screws for the G5 will be to fat to go through the GPU. Got the price back for the laser cut and it is more than I was hoping. Material $25 Programing $50 Setup Fee $100 Total Inc. GST $192.50 Edit: So the screws are too fat and it doesn't look like I can buy any long enough ~70mm that are thin enough ~2-3mm. Removed the barbs to see if I could use some shorter ones but they are larger than your standard water cooling parts. Wonder if I can weld the fat screws to the end of thin ones...
  14. Sent a email of to the laser cutter today but decided I think I am going to have a go at drilling holes in my old Storm G5 pure silver CPU waterblock and mount it only trouble is the hoses will be close to the PSU in the mATX case.
  15. Had a similar problem with my 2600K it was stable overnight for prime95 at 4.9GHz but after days of use it would become less stable needing clocks dropped further and further but leave it at 4.4GHz for a while and it was stable at 4.9GHz again for a few days.
  16. If the CPU can handle 70FPs + at 720p then it can handle 70FPS+ at 4k (not including emulators) and you just need the GPU to go with it now . What speed is your RAM running?
  17. Could probably use zip ties on the current blocks but with soft HBM I don't want to risk crushing it with uneven mount pressure. I could probably cut out a plate that would sit over the top of the current one that would work ok if need be.
  18. Good result I had to delid my 3770K and watercool it to get 4.5GHz stable but then it was a real dud chip that only did 4.2GHz before delid.
  19. Some basics here https://overclocking.guide/ivy-bridge-socket-1155-oc-guide/
  20. It is handling the new monitor nicely although some games still drop below 60FPS I really need to get it on water but have not yet contacted a laser cutter to get a mount made.
  21. The CPU should be capable of 60FPS in most games with a little OC and 1600c9 RAM. You probably wont get 4.5GHz without a K model due to locked multi. 2080 will be ok for 4k in some games but others really need the ti.
  22. Dasa

    What are you playing at the moment?

    Mostly survival city building\management but there is different skill trees to take with moral choices and there is a few other random decisions to make along the way, a nice balance to keep it interesting. The choices don't matter much on normal difficulty but some hard choices become essential for (the greater good) at harder difficulty settings.
  23. Dasa

    What are you playing at the moment?

    Put 56 hours into this game in the last few weeks
  24. Dasa

    Home Server advice

    It will also come down to the software you use as Blue Iris for example only relay works properly with Intel Quicksync at the moment which significantly reduces CPU load and although they have recently added support for CUDA it doesn't work very well.