Jump to content

Melodist

Member
  • Posts

    559
  • Joined

  • Last visited

Reputation Activity

  1. Informative
    Melodist reacted to B4Quattro in Thermal Grizzly Kryosheet on RTX 4090   
    Hi together,
     
    I want to share with you my experiences with the Thermal Grizzly Kryosheet, a graphen thermal pad in combination with a Watercool Heatkiller V waterblock. I recently upgraded my system to a 7800x3d and an Asus RTX 4090 and heard some stories about the fast pumpout effect of thermal pastes on the GPU because of the very high power consumption and therefore high heat generation. So I tried the graphen pad, and I think the results are acceptable. I also couldn't find any information about the dimensions of the AD102 die (it's around 24x26,6 mm), only the die area of 608mm². I used the 33x33 mm pad and protected the SMD's around the die with kapton tape. These are some pictures of the gpu and my 3DMark Timespy result (don't have a key for Port Royal or Extreme 😞 )



  2. Like
    Melodist reacted to YoungBlade in DDR5 6400 CL32 vs 6800 CL32   
    If they're the same price, you might as well. But if the DDR5-6800 costs more, I wouldn't bother. The real world performance difference will be within a couple of percent if that.
     
    You can see what Intel 13th gen memory scaling looks like in this video from Hardware Unboxed - it's about a 4% difference between low end and high end memory out of the box. Tuning the timings on a cheap kit gives more of an uplift, if the uplift is even worth it. We're still talking about just a few percent when gaming at 1080p with an RTX 4090.
     
     
  3. Like
    Melodist reacted to filpo in DDR5 6400 CL32 vs 6800 CL32   
    Doesn't really matter too much. If you like the higher speed and can spend the money then sure go for that but 6400 is already quite fast and I don't really see a benefit to going faster
  4. Like
    Melodist reacted to RONOTHAN## in Contact Frame's long-term reliability?   
    It's an issue with both. Some boards only have issues with one or the other, but there are some that have issues with both. 
     
    Yeah, I wouldn't really be that concerned. 
     
    It will still have bent a bit, but it's also not that big a deal. The contact frame should still help even if you've installed in a year or so down the line. 
  5. Like
    Melodist reacted to RONOTHAN## in Contact Frame's long-term reliability?   
    A little complicated, some motherboards can have some issues with memory when using the contact frame, though if you have one of those it would become pretty apparent within a few minutes of trying to use it at XMP. It's only a handful of boards, so it's not likely to be an issue, but just be aware of it. 
  6. Like
    Melodist reacted to RONOTHAN## in Contact Frame's long-term reliability?   
    It doesn't touch the board though. The Thermalright (cheap) ones have rubber pads that touch the motherboard in the same space the normal loading mechanism would touch it, and the Thermal Grizzly (expensive) don't even touch the motherboard, floating on the tension of the pins themselves if it's installed correctly.
     
    Given how they work, I wouldn't have any qualms about running them in the long term, hence why every 13th gen system I've build has used one (the Thermalright ones as they're cheaper and are honestly better for any application without liquid nitrogen). One of the systems has been running for two years with no issues (it was a used Z690 board that had a contact frame installed day one from back in December 2021)
  7. Like
    Melodist reacted to RevGAM in 14900K - Define 7 C vs Meshify 2 C vs Torrent C   
    I'm just going to make some observations regarding the use of test benches to provide reviews and evaluations to end users. These are based on inferences, logic and my incomplete understanding of science. I do not pretend to be an expert and I have not yet done any cooler testing as I lack tools and coolers require a bigger outlay of cash. Since my ethics prevent me from striking deals with sponsors, manufacturers and vendors without rules that some of them would (strongly) disagree with, I am the only source of funding, so I can't really afford to go into cooler testing at this time. I would never accept samples directly from manufacturers for testing purposes because some companies send "golden samples" that will distort test results, meaning that the results may not be what most people will experience, and that could cause people to question my test results - because the company manipulated the results. That said...
    A test bench that is horizontal has heat flowing upward (away from the mobo and components). Many computer users choose to buy vertical cases, so the results are not representative for them. The same is true for vertical test benches. A diagonal one produces results that are different from almost all cases. Heat collects in turbulent and "dead" areas of a case, such as between the PSU shroud and GC, whereas this is not possible with a test bench. Cases, except those partial cases with open sides/top, have very different airflow from test benches, especially because of case fans. A test bench without fans blowing across it will achieve different results because it relies entirely/mostly on convection, the natural movement of heat upward. Even if there is an AIO mounted on the test bench, its position relative to the rest of the parts is sometimes diagonal, which will alter results. The argument of comparable is irrelevant because, when comparing different cases, it is impossible to achieve truly comparable results other than in a very general sense, even if the components are the same. If comparing coolers, however, the results are going to be "consistent" with each other, but not with testing in a bench. For these and other reasons, any test bench results are less useful to consumers, as evidenced by the rank of the D15 in that video. For that matter, lab quality testing, or even testing done in a PC where everything is locked down to eliminate confounding factors gives results that are themselves inconsistent with real-world experiences. And then there are the manufacturers that lie about their products, swap components to cheaper ones after lots of positive reviews and don't tell the public, and other shady tactics (like Nvidia deprecating GC features by turning off or decreasing them in a new driver release).
    I have never used any of those cases, so I can only make statements based on what I can see. You'd do better watching reviews. Personally, I would not get the Define 7 - it looks like it will run hot. I think the Torrent is better than it but it looks like the Meshify might have the best cooling power.
     
    If you really want to keep it cool, discard the idea of solid sides and get a full mesh case, or at least one with a side panel that you can attach a fan to in order to either pull cool air in or out. That is if your goal truly is to keep your CPU from thermal throttling. For example:
    ASUS Prime AP201
    darkFlash DLC29 Full Mesh, DLH21 Mini-ITX
    SAMA IM01PRO Black Iron Mesh Panel Micro ATX Tower
    There are others but my time is running out. I hope that helps!
  8. Like
    Melodist reacted to OhioYJ in Noctua NH-U12A Chromax vs Standard Performance difference?   
    So I don't know about that NH-U12A, but I'll say I decided to "waste" money and replaced all of my Noctua stuff with Chromax versions (cooler and fans) several builds back (I say "waste" because I did it purely for cosmetic sake). I had the NH-D15, and replaced it with the Chromax version. I didn't notice any difference at all. If there was one,  it would have been so small it would have been in the margin of error. I've run the Chromax version now on a 9900k, 12900k, and a 13900k.
  9. Like
    Melodist reacted to RONOTHAN## in Noctua NH-U12A Chromax vs Standard Performance difference?   
    It's not paint, it's anodization which doesn't affect cooling capacity. 
     
    The two different versions should perform identically to each other. 
  10. Funny
    Melodist got a reaction from Cavalry Canuck in PC case with sturdy / good frame?   
    For testing a build I used to do that yes but now I have to be afraid that it'll flex and I'd have to put the case fully together before finishing the wiring and testing.
     
    Is it really such a high demand to have a case's frame hold together by itself nowadays?
     
    Would you buy a cooler that falls apart when you take the fan off?
     
    Or the radiator of an AIO that is split into 3 pieces, tubes in between and holds together once you attach it to the case?
     
    The level of complacency when it comes to the quality of cases is frighteningly high.
     
    Anything is just as strong as the weakest link in the chain.
  11. Informative
    Melodist reacted to mariushm in 1 vs 2 Phase AUX Power   
    No, I think aux has something to do with powering the pci-e controller and/or other system-on-chip controllers but I'm not sure... I didn't keep up with these minor rails.
     
    The 8 phases for CPU are not in parallel, they can run all at the same time, but usually they're meant to alternate, two - three phases work for some time for example, then others start to work and let the others cool down a bit, and so on... each phase can do let's say up to 80A of current, but they may be most efficient at 30-40A, so instead of using just 4 phases at 50A to produce 200A of current, they can run 6-8 phases at 30-40A to get better efficiency...  if you run all four at same time for 10ms then stop for 1-2ms and rely on the energy stored in inductors and capacitors to provide smooth output voltage, then you may get fluctuations in the voltage, like for example 1.5v when the phases all run and the capacitors are topped out and then during those 1-2ms of off time, the voltage may drop to 1.48v until the phases restart and top up the inductors and capacitors.  If you alternate between phases, there's never a "dead" time, and a few phases are always ready and working to top up and maintain the voltage to the level the cpu asks.
  12. Funny
    Melodist got a reaction from thrasher_565 in PC case with sturdy / good frame?   
    Apparently people aren't thinking at all otherwise they wouldn't buy a connector from a third party for their 4090 that pushes over 450 watts over a single cable.
     
    And then even people like Steve have sloppy, I don't care reviews where the same problem he mentions occurs in another product and he fails to point it out.
  13. Funny
    Melodist got a reaction from Middcore in PC case with sturdy / good frame?   
    Any other case that you can recommend than the 802?
     
    I'm building a case for audio production but it doesn't have to be nose isolated, just need something solid with good airflow.
     
    PS: Didn't like the flex on the North.
  14. Like
    Melodist reacted to RONOTHAN## in Z790 motherboard and ram for 13600K?   
    Fixed the advantages. The better VRM doesn't really mean much since they're both overkill for a 13900K, the PCIe 5.0 NVMe drive is because it steals 8 lanes from the GPU meaning that a simple AIC will get the ProArt to have the same capability, and IMO if you actually care about audio you'll buy an external DAC rather than using onboard. 
     
    Yes, the on board slots are actually just a downgrade over the dual x8 slots since those you could use for more than just fast SSDs. 
     
    If it wasn't for the lack of the POST code I'd be 100% recommending the ProArt. The Z790-E Strix is just a really weird board, the feature set is pretty lacking compared to competing vendors, and the only advantage of it is that it's ASUS's cheapest Z790 board with a POST code, and for a system that needs to be easily troubleshooted if something goes wrong that's a very useful feature. 
     
    Software wise they're not great (though no worse than Gigabyte or ASUS IMO, and it's not like you need to use any of the in-Windows tools they make for RGB and stuff), performance wise they're in line with everyone else (ASRock does consistently have their motherboard own world records, partially because they sponsor Splave), durability wise their high end offerings like the Taichi series are generally built really well, and BIOS wise they have come a long way in the past few years. I've used their boards for the past while, they really aren't bad, and realistically I've had a better experience with them than I have with some of the ASUS boards I've used (I still despise the Z690 Apex I used a while back, and I had less BIOS problems with the Z170 OC Formula I have than with the Maximus X Apex I've got). The X570 Taichi I used to use was a really fantastic AM4 board, the only reason I sold it was because I was bored and the X570 Master I replaced it with was better at memory overclocking (plus I wanted dual BIOS). 
     
    I would be going Taichi if you need the Thunderbolt support, ProArt if you need the 10GbE on-board and the ATX sizing. 
  15. Like
    Melodist got a reaction from Tan3l6 in PSU-Tier list with ATX 3.0?   
    The list isn't complete and very small because there are a bunch of new ones certified missing and not yet tested and on the list, which says 3.0 compatible, only the RMx and RMe are 2x6 compatible the rest has the 12VHPWR. 
     
    https://www.corsair.com/de/en/p/psu/cp-9020259-eu/hx1000i-fully-modular-ultra-low-noise-platinum-atx-1000-watt-pc-power-supply-cp-9020259-eu
     
     But then the question is whether the old version was already compatible:
     
    https://www.corsair.com/de/en/p/psu/cp-9020214-eu/hx1000i-fully-modular-ultra-low-noise-platinum-atx-1000-watt-pc-power-supply-eu
     
    Here's another one that isn't on the list: 
     
    https://www.corsair.com/de/de/p/psu/cp-9020246-eu/sf-l-series-sf1000l-fully-modular-low-noise-sfx-power-supply-cp-9020246-eu
     
    Corsair's website is horrible though, I'm getting better results from my price listing site and links than theirs lol.
     
    The list is misleading because most of them have the 12VHPWR and that is not an option, only the RMe, RMx and Be Quiet Pure Power 12 are 3.0 certified and don't have that stupid 12VHPWR.
     
    The ones I found above aren't even on the list.
     
    There is only one EVGA PSU left on the German market 🤧
  16. Funny
    Melodist got a reaction from Mark Kaine in 5959X is throttled - need assistance   
    Hello there,
     
    Friend of mine runs his 5950x on an Asus X570 Proart motherboard and he scores about 24.000 in Cinebench R23, which tells me he's not hitting the CPU at full performance.
     
    What does he have to activate in the BIOS to run the CPU at full speed?
     
    I know it's probably PBO and I have 0 experience with X570 and since I'm helping him over the phone I can't test settings myself and I've seen like 8 different type of PBO and Performance settings in the motherboard BIOS manual which I can't connect.
     
    It's not necessarily about OC but running the CPU at it's boost frequency because it looks like it only does the base clock.
     
    Can somebody tell me how to most basically unlock the 5950X so that it runs at it's advertised speeds because in BIOS default, it doesn't do it.
     
    Motherboard is the Asus Proart X570 Creator.
  17. Like
    Melodist reacted to RONOTHAN## in Z790 motherboard and ram for 13600K?   
    Oh, the Z790 ProArt is fine for those chips, the VRM on it is plenty for a 13900K. Don't know why they only put the 4 pin on instead of the 8 pin for the secondary EPS connector, though it won't cause issues either. 
  18. Funny
    Melodist reacted to AI_Must_Di3 in Quality case with very good airflow?   
    Lian Li PCV2010B https://www.newegg.com/p/N82E16811112170
    😉  You find one that is actually in stock let me know!
  19. Funny
    Melodist reacted to freeagent in Quality case with very good airflow?   
    I like the new Torrent line.
     
    Looks great, and is a top performer.
     
    And its stiffer than I am.
  20. Informative
    Melodist reacted to RONOTHAN## in Z790 motherboard and ram for 13600K?   
    ASRock does weird things with their BIOS, and it's the most differently laid out. In the past they were also known to have a bunch of glitches in it, but nowadays their worst offenses are putting settings in odd spots and copy-pasting their BIOS from high end boards to low end boards, meaning you end up with settings that do nothing on the lower end boards a lot of the time. Personally, I can get around an ASRock board just fine, and on AMD I do actually kinda like their BIOS, though on Intel it does take a bit of getting used to (I still think it's relatively intuitive for the settings I use, but it's much different than ASUS, Gigabyte, and MSI). 
     
    The Lite is the exact same PCB as the normal Taichi, just with less RGB, so the BIOS file is likely exactly the same. The Taichi is their top end board for Z790, so it should get pretty good BIOS support from them. 
     
    Yeah, the Z790-F and Z790-H are almost the same board from what I can tell, only minor differences between the two. 
     
    No, the ProArt is also an 8 layer board. My main problem with the ProArt is just the price, since you can get the Ace for the same price, and if you're spending that much on a motherboard I'd want it to have a POST code. It is still a fine board though and if you need on-board 10GbE and Thunderbolt at the same time it's usually your best pick. 
     
    No, it all comes down to the individual board's capability. The only difference between the two is that 8 of the PCIe Gen 3 lanes on Z690 are Gen 4 on Z790. Other than that they're the exact same chipset, and it comes down to the individual boards how they implement those lanes whether it actually makes a difference or not. 
     
    Just saying, on LGA 1700 I'd actually consider a Gen 5 M.2 slot a turnoff because of the way to get there. The only way to get a Gen 5 M.2 on Z690/Z790 is to have the GPU's x16 slot drop down to x8 wired, and the couple of workloads I can think of that need Gen 5 M.2s they also need the GPU with as much PCIe bandwidth as you can get. 
     
    I've used a couple ASRock boards before, they've all been fine. The X570 Taichi I used to run was a great board and a fantastic value for the feature set. Main reason you're probably not seeing the Taichi lite in many YouTube builds is A) it's really new and has only recently started shipping, B) the regular Taichi was not really a good value in the past, and C) most YouTube builds factor in aesthetics, and in that regard the normal Taichi is better. 
     
    I'd still rather have the Z690 Ace than the Z790 Taichi (Lite), it's just a better motherboard overall, though neither is a bad option. If you're worried about the VRM on the Ace, it's the exact same VRM on my Unify-X, and I ran that board with no VRM heatsink and an overclocked 13700K pulling 400W with no issues back when it was on test bench duty. The Taichi might technically have a better VRM, but practically they're identical. 
  21. Like
    Melodist reacted to RONOTHAN## in Z790 motherboard and ram for 13600K?   
    My relationship with G.Skill on DDR5 is complicated. Their PCB is rather good, and from what I've seen they consistently do tighter timings than their green PCB counterparts and other brands, but it's only 1-2 of the subtimings that are consistently better so not really that big a deal. Their heat spreaders themselves isn't actually that bad, but the problem with them is they didn't put a thermal pad between the PMIC (the main heat generator on the sticks) and the heat spreaders, so memory temps end up being rather high. 
     
    I don't avoid their kits, if they were a good price I'd be buying (hence why I currently own two different G.Skill DDR5 kits), it's just that I either limit the voltage I run into them for long periods (my 24GB sticks I don't run above 1.4V and my 16GB sticks I don't generally run above 1.5V with active airflow as they end up with voltage rollover) or just remove the heat spreaders and run them bare as my Flare X5 6000 CL30 kit currently is (this kit I got used so the warranty was non-existent, hence why I felt comfortable doing this). With the same amount of cooling bare the Flare X5 kit can handle 1.65V without rolling over, and even more for benchmarking itself. 
     
    If you aren't overvolting or overclocking your RAM, none of this is really that big an issue, they do stay below dangerous temps and are stable at the XMP, it's just that the memory itself is rather high compared to everyone else and thus if it's the same price as someone else, I'd go for the other brand. 
     
    Their DDR4 was pretty good, yeah. If I wanted a good B die kit and didn't want to deal with the silicon lottery, I'd be buying a higher rated G.Skill kit like 3600 CL14 or 4000 CL16, as their kits generally had the best binning. Surprisingly though the best kit of DDR4 I ever owned was from Adata of all places, in a 3600 CL17 bin (the 2nd worst guaranteed Samsung B die bin). That kit had a terrible, all plastic heat spreader, but it scaled with voltage indefinitely and would run insane settings (the best I ever got was 4800 CL15, and I'm pretty sure it would go higher if I had a better CPU). 
     
    Everything I've seen has shown that TeamGroup and Kingston have some very good heat spreaders on their DDR5 kits, so if I could take any it would be a TeamGroup kit (especially for the 2x24GB sticks where the memory temps get exceptionally high). Generally though I'd just be getting whoever is the cheapest, there really isn't that much difference between all the different brands. TeamGroup's T-Create brand has been a rather good value, at least in the US. 
  22. Informative
    Melodist reacted to Tetras in Recommendation for my 13700k build   
    If:
    Your usage is going to lead you to use both together, regularly, for long periods of load. You plan an upgrade to a 14900 and suitably higher-end GPU. Then, I'd just get the 1000 or higher. The price difference is not that much between them (and relative to the cost of a whole build, trivial), but for you, the additional headroom could be invaluable.
  23. Like
    Melodist reacted to seon123 in PSU-Tier list with ATX 3.0?   
    Always go multi rail if possible. "Single rail" is just the marketing way of saying that they cut corners on the protections by omitting multi rail OCP on the 12V rail. Gullible people unfortunately fall for it, and some of those even think it's better. 
    Does it matter? If you're curious about a specific PSU, just read a review on it. 
    https://hwbusters.com/psus/be-quiet-pure-power-12-m-850w-psu-review/
     
    If you want to compare a handful of PSUs, you can also just ask about those specifically. 
  24. Informative
    Melodist reacted to RONOTHAN## in Z790 motherboard and ram for 13600K?   
    Not really, but the silicon lottery is very much a thing with Intel memory controllers (13th gen was the worst in a while), where there's a range of over 1000MT/s between the best and worst chips. The numbers I've given (6400 for single rank configs, 6000 for dual rank) are based on if you lost the silicon lottery what I'd expect to just work, not the average chip or even a golden sample. 
     
    It depends what platform you're talking about. If you're on Ryzen 1000 or Haswell-E, 3200 CL14 was sketchy at best. If you're on 11th gen or Ryzen 5000 especially, I'd rather have a 3600 CL16 kit, or even faster. Right now we're in a similar place to where Haswell-E was with DDR4, where the speeds that will be considered relatively pedestrian in 3-5 years are hit or miss today. 
     
    If you want comparisons to B die, that would be Hynix 16Gb A die currently. It clocks to the moon (similar to good B die), does very tight subtimings (similar to B die, a lot of the time better), and is generally the first thing that motherboard makers optimize for. The only differences are the tRCD doesn't go tight (B die was an anomaly in that regard) and it's actually consistent (B die was infamous for being all over the place, where I've had kits that do 4800 CL15 and kits that won't even do 3200 CL16), so honestly I'd say it's better. The consistency with it means that even if you buy a 6000 CL30* or 6400 CL32 kit, you can just dial in the settings of higher rated A die kits and expect it to just work. When memory controllers get better, for instance, you should easily be able to enter DDR5 8000 CL38 with one of those kits and expect it to just work. 
     
    The 24Gb M die is the bigger brother to A die, it's basically the same thing but higher capacity and slightly looser timings. Supposedly it should clock higher, but with my experience it doesn't (granted, this might be BIOS and/or board issues). I'd reasonably expect it to just work with a 7200 CL34 speed bin though if you just dial it in. 
     
     
    The 14th gen memory controller is supposedly significantly better than the 13th gen memory controller, to the point where it's actually guaranteed to hit DDR5 6400 (allegedly, well find out for sure Tuesday IIRC). It will support faster, and it might be that 7200 becomes the speed that "just works" on 14th gen, but at the same time those 6400 rated kits will support that speed anyway so you might as well just use them, they aren't slow by any means anyway. 
     
    *6000 CL30 is only 50/50 to be A die, the older Hynix 16Gb M die is also present on these sticks and while it does perform really well with even tighter timings a lot of the time, it clocks no where near as high as A die does. 
  25. Like
    Melodist reacted to RONOTHAN## in Z790 motherboard and ram for 13600K?   
    Flip a coin. One really doesn't have a benefit over the other. I'd probably go for the Corsair kit, just because G.Skill's RGB and overall inadequate heatsinks does cause memory temps to be rather high, but then again I'm a memory overclocker and that type of criticism is only really relevant for people like me, and you may enjoy the aesthetic benefits of the RGB. 
     
    Odds are you don't need more than an 850W, though I'd get a 1000W if the price difference isn't massive. 
×