Jump to content

Mahatma Gandhi

Member
  • Content Count

    87
  • Joined

  • Last visited

  • Medals

Everything posted by Mahatma Gandhi

  1. Woud really love to see whether or not ddr5 performs well with ArmA3 Just got contacted elsewhere by a dude who apparently bought a 12900k with speedy ddr5. No direct comparison to ddr4 but at least we'll learn about alder lakes arma3 performance. @Groove_C i told him to contact you regarding the avx2 malloc, as i dont find the file and armaholic is gone unfortunately 😪
  2. I have certainly not been living under a rock and as said and as you agree, there are CPUs that are artificially reduced in core count which are just the same as the top tier models and then there are the ones where a core is defect and where those are used to produce lower tier SKUs. But there is certainly none which are selected based on the IMC or uncore frequency capabilities. The max spec is 3200MHz and all of them work at that. OC is OC and they dont bin for RAM OC. Neither of them, AMD or Intel. Core Voiltage is also identical between 5600x and 5800x. Two CCD Models might be selected at that end, but certainly not the lower ones.
  3. There is no logical explanation for such a difference. They disable two cores because one might have an error, but that does not necessarily mean that other parts of the chip are affected. Same with Intel. I don't see a reason why lower tier models should not reach the same memclock or uncore freq. In fact, I had a 3300x CPU that had no problems running FCLK 1900+ and 3800MHz RAM. And in most cases, they disable cores not because they are broken but just to diversify the products. There are more 5600x sold than 5800x, yet there are for sure more fully functional dies coming from the fab then ones with defective cores. Last not least, the fastest on the current list is this one: https://imgbb.com/0tSC7XB And I want to pick up your statement again that Ryzens are slower at high quality settings. Do you make that up or do you have proof? Its the first time I hear that.
  4. Sure, I fully agree. My point was, the 102FPS documented have not been a particularly obszene result for a 5800x as I get 97-100 with a bad sample, SR RAM, an ITX box, and a 30 Euro cooler. That was the only thing I wanted to add to the discussion. Min FPS for me are about ~71/72 on average. And by the way, I am hard limiting mine to 95W power draw. EDIT: Also the lower the settings (resolution/quality), the "better" Ryzen is, due to it's cache. But nobody plays at such settings. Once you go for 24/7 settings, the picture is different. Thats the first time I hear that. Do you think so or do you have data on that?
  5. But you know my results, so it is proven that 97-100 FPS average is where you end up with a properly tuned poor mans 5800x system with good but not great RAM. And my 5800x I got cheap, because it was the worst in a binning done by a few folks I know. But you know all that. We talked about it.
  6. This screenshot for the 102FPS 5800x was very good no doubt. But it wasn't one in a million. I have a 5800x myself meanwhile. 2x8gb Single Rank RAM at 3733MHz CL15, mild CPU overclock (101MHz BCLK), no curved PBO, Beta Bios on a X370 Board, 30 Euro Air Cooler in a mITX System and I reach 100FPS myself. So for sure this Sys was fast, but the statement it was a golden Sample for ArmA3 and a lucky run that is out of range otherwise is not true. What the RKL can do we will have to wait and see. So far I haven't seen one that reached 100FPS+
  7. Mahatma Gandhi

    We need AMD Ryzen Benchmarks! Share your toughts!

    It seems that AMD did the evil thing and integrated a ID check into their Firmware. To make it short, any BIOS newer than AGESA 1.1.0.0 Patch C does not seem to work with Zen3 on X370 Chipsets. That said, crossflash worked very well with Zen2 and gives all the nice features including rBAR and per CCX overclock on the X370 Board. Certainly worth it if you are willing to take the risk but it won't help with Zen3
  8. Mahatma Gandhi

    We need AMD Ryzen Benchmarks! Share your toughts!

    @Valken The system specifics: Mainboard: Asrock Fatal1ty X370 Gaming-ITX/ac Motherboard - I had bought it back in 2017, originally paired with a Ryzen 1700. AsRock leaked some beta BIOSes for their old B350 X370 lineup, you can find the deitails with google. For this board it has been the BIOS 6.61. But this only works with a couple of AsRock Boards and the BIOS is surely beta. For B450 and X470 Boards, there is official Zen3 support, so for a budget build, those could be used or upgraded if already there. CPU: the mentioned 5800x. For now, I only borrowed it from a friend who was screening several for high IF capabilities. This one failed to achieve 1900MHz IF without WHEAs and he was nice enough to send it over for testing on this board,. This is why I also run it at 1866MHz and 3733MHz RAM. Cooler: Scythe Kotetsu Mark II, a really cheap but not bad cooler. However, I should mention that I realized before, that the cooler is somewhat convex which is why I sanded it a bit to make it fit better on my previous 3300x ryzen heat spreader. No magic, I just used my thumb and some sandpaper to make it a bit more flat in the middle of the plate. With this cooler, at the end of a cinebench R20 run (multithread followed by a single thread run) HWinfo gives a peak temperature of about 83 °C. I think that's perfectly fine given the load which will probably never be the same during gaming. RAM: 2x 8GB TridentZ 3600MHz CL15, so single rank B-Die RAM. To be honest, the kit has a great reputation and has been very pricey but I never got it working with true low latency. It gives a reliable CL15 up to 3800MHz. 3733 or 3800MHz CL14 is definitely out of reach at 1.45V. Subtimings go ok with a tRFC of ~147ns. Not great nor bad either. At least with my CPU-Board combinations, i can't get further, but it never really mattered. CL15 vs CL14 is not a dramatic difference. Yet of course... it could be better particularly given the price I paid back then. GPU: MSI 1070Ti 8gb --- nothing special. PSU: 400W BQ power supply Chassis: Node 304
  9. Mahatma Gandhi

    We need AMD Ryzen Benchmarks! Share your toughts!

    @Valken The funniest part is that this is all done with an unofficial beta BIOS from AsRock on my old AsRock X370 Fatal1ty mITX Board with a 35 Euro Scythe Kotetsu Mark II Cooler. So... this isn't quite a costly high end system
  10. Mahatma Gandhi

    We need AMD Ryzen Benchmarks! Share your toughts!

    I finally got a 5800x to play with. Its not a great overclocker RAM wise and i only reach 3733MHz RAM stable, but its already very fast. The mid 70 fps results were from my 3300x.
  11. Hey Dwarden, if you wonder whether you would ever see 100 FPS happening in YAAB, you should check this out: As a teaser: the min FPS is 71

     

  12. A friend in the German Ryzen RAM OC Community (spawn off the Computerbase.de Forum) going by the name ZeroCoolRiddler, gives some pretty impressive results from a 5800x running Micron-E Dies @4000MHz 1:1 with the FCLK. To my knowledge, this is the first system to pass the 100 FPS sound barrier. Interestingly, his min. FPS hit a mind boggling 71FPS which is pretty much exactly what I proudly presented as the max. FPS my 3300x could achieve a week ago. The system ran the cma_x64 AVX2 malloc and a curved PBO, a new Zen3 AGESA function which allows a manually adjustable Boost behaviour in terms of frequency and voltage settings allowing to customize the boost behaviour with individual (and non escalating) voltage settings per core making curved PBO apparently the cpu optimizing method of choice on Zen3.
  13. As a final assessment of the possibilities of a 3300x: This is 3300x @45W cTDP + PBO (+150MHz); 2x8gb TridentZ 3600MHz @3800MHz CL15 + Subtimings; XFX RX470 4gb @stock - CMA_AVX2 malloc - 1080p Standard Preset+S-key By the way, the board is an old school AsRock x370 Not sure. There is a whole lot of performance hidden in the subtimings and in particular for the Micron E Dies, there are settings that basically always work out of the box.
  14. Thanks @Valken and thanks @Groove_C for detailing the malloc installation. I got it working! So here is the 3300x @stock + 2x8gb SR 3800MHz CL15 (IF1900) + Radeon RX470 4gb @stock 1080p standard settings + S Key; cma_x64 enabled Min frame rate was: worst run 40 best run 42
  15. Let's wait for some self made bechmarks on a 5600x/5800x from one of us. I think those ones will be the ideal for most. The 5900x is great but for pure gaming somewhat oversized. By the way, my 3300x @stock +3800MHz CL15 and default malloc gives ~65 FPS average (min=39; max=109) in yaab @1080p (GFX=RX470 with 4gb RAM) I still couldn't figure out how to get the CMA AVX2 malloc to work. Looks like I am to stupid to do that. If someone has an idiot proof 'how to' at hands (or can make one), I will try again and report. Best, Mahatma
  16. Mahatma Gandhi

    We need AMD Ryzen Benchmarks! Share your toughts!

    Clawhammer, the RAM is your only problem! I get 58FPS in YAAB with standard preset on a 2600x + RX470 + @3600MHz RAM. Either get your RAM frequency up or toss the RAM and get some good one. Slow mem like yours is your worst enemy in the yaab benchmark. I use the 3200MHz CL14 G.Skill FlareX 2x8gb
  17. The only issue hier is: Voila, you spend more money already than I have spent for CPU, Mainboard and Graphics Card alltogether (against the 9700k). If you take the 9900k you spend more on the CPU alone than I did on CPU, GPU, RAM and mainboard. It really depends on what you are willing to invest and where you want to go with it. Are you only playing ArmA3, is you budget high enough to be able to afford fast RAM AND the Intel CPU, do you have to go lower on the GPU because you run out of budget etc etc .... If you do have the budget to buy all these High End components, it is the best solution. If you don't, its not a good solution. In most cases I still see people buying an 8700k/9700k for ArmA3 overclock it to 5GHz and combine it with the cheapest RAM they can find on the Market just to find themselves having best average FPS. e.g.: My 2600x is by the way not overclocked at all... it's actually undervolted and at auto clock. @Valken We will surely test the Ryzen 3k with ArmA3. I will let you (guys) know what to expect.
  18. Thats perfectly true for Intel based Systems also. my favorite for ArmA3 is still the 2600x + 3600MHz CL15 Kombo. Thats very fast and ultra smooth gameplay for little money... But if you're a fan of Intel, i understand that its not only about $ and frametimes.
  19. Seriously considering to buy one of those to have a mobile ArmA3 device to visit friends. https://www.anandtech.com/show/13830/asrock-at-ces-2019-deskmini-a300-worlds-first-amd-ryzen-mini-stx-pc-launched its 8cm x 15cm x 15cm and AsRock annonced to claim 120 USD for it in the US. Thats dirt cheap given it comes with case, board and PSU. It fits the 2400G which according to Youtube does pretty well in ArmA3.
  20. @Groove_C Nice overclock! Does anyone have such a CPU here? @Valken You can look up memory latency ladders of any CPU on Userbenchmark e.g. 7700k, 8700k, 9700k, 2700x and 7900x (I havent checked whether in these runs here specifically the CPU or Ringbus or Mesh etc is overclocked ... one needs to pay attention there of course) But cache performance alone is a poor indicator as of course BP performance and cache sizes play a huge role. Also cache hierarchy and Inclusive vs. non-inclusive vs exclusive cache organization matters. If you want to make performance extrapolations, its save to say that comparisons within one uArch are good but not between uArches (e.g. Coffee Lake vs Skylake-X vs Ryzen) Maybe @Dwarden can give an idea of what to expect from large caches (e.g. 32mb L3 in Zen2 or the eDRAM in 5775c).
  21. @Groove_C Thanks for making this comparison. I made another one based on the same leak of the ZEN2 ES sample from last week. I compared it to the current incarnation of the Zen µArch. Of course this always needs to be taken with a grain of salt. This is an ES with non final AGESA and UEFI Firmwares and it only clocks at about 3.7GHz. Yet, I think it appears that the outsourcing of the memory controler in Zen2 will roughly cause a 10ns Latency penalty over the Zen+. On the other hand, its pretty save now to assume that Zen2 will double L3 Cache and will come with 32mb L3 Cache that performs fast at a latency of only ~10ns. So for ArmA3 there is good and bad. RAM Latency gets worse, double (32mb 10ns L3!) Cache will however for sure absorb many cache misses. Its very hard to predict what the result will be. With some friends we are just about to start some small RAM Review Channel and I am sure we will be able to review Ryzen 3k soon after release. And I will make sure we include ArmA3 ;-) Will let you know! Mahatma
  22. Mahatma Gandhi

    We need AMD Ryzen Benchmarks! Share your toughts!

    Give them 1.4V and try to reduce those timings further. Every bit helps as you can see. You can also check out the Ryzen DRAM Calculator as a tool to get values that should work with your RAM. Here's a tutorial P.S.: Flashing to the latest BIOS might also help to achieve better timings.
  23. Mahatma Gandhi

    We need AMD Ryzen Benchmarks! Share your toughts!

    You simply bought the wrong RAMs. You are lightyears away from what I have with 3600MHz CL15-14-15-14-26-40 + short subtimings. This is how ArmA3 scales with RAM on a 2700x Ryzen: Best option if you want to get max performance is to sell them and get 3200MHz CL14 FlareX and optimize timings.
  24. Mahatma Gandhi

    We need AMD Ryzen Benchmarks! Share your toughts!

    Retrospectively, the RX470 4gb I have was a very decent buy (back in 2016). Last weekend, we have been playing the "A long way" Coop mission and I am on average well above 80 FPS, mostly 90-110FPS with the "Very high preset" + HBMO @ strong and View Distance set to 8000m. The GPU shows a constant 100% load, so I am GPU bound, but at this level... It simply does not matter anymore to me. I am to old meanwhile to see a difference between 80 and 120 min FPS. The most important aspect to me is that it is completely smooth. Fully even frame times matter much more than 15% average FPS. I have always hated these stutters in the OFP/ArmA3 engine and this is the first build I got completely rid of them. So Kudos to AMD for finally bringing an affordable and fast platform with many cores to the marked. If ArmA4 ever comes to market, I might go for the Ryzen 3k if necessary (lets see how the enfusion engine works). Luckily, they announced that it will be a drop in replacement that works on any current AM4 Board and apparently, it will even upgrade the Boards to PCIe Gen4. Nothing else to wish for. By the way, in the other thread, you wrote that the 2600x only gives 35 FPS @ Ultra setting. I checked that yesterday, too. So with Ultra Settings applied and no reduced View Distance (no S-key) but disabled AA (otherwise the RX470 is a hard limit) I end up with 46.8 FPS on average in YAAB but it remains fully smooth. Don't ever be cheap on RAM if you play ArmA ;-) Thats the most common mistake people do when configuring their systems.
×