Oh, so X570 chipsets are mandatorily positioned closer to ram slots? Does that influence the ram timing ability? I assumed it was all up to the mem controller in the CPU itself, but I'm no expert. Here i'm using a daisy-chained X570 motherboard (aorus elite). Couldn't find any t-topology mobo within my budget, which would probably have been better in my case. This ram has samsung dies (vengeance lpx), which got me the best timings our of the 2 rams i tried (the other had hynix dies). Gigabyte was responsive and provided me with several bios betas, but all failed to improve the timings unfortunately. The ryzen RAM calculator tool was useless in my experience too. So I settled with these. OC-wise I'm not touching anything other than memory stuff, I don't want overclocking other than the advertised XMP clocks and latencies. In the brief testing from yesterday, it managed 28 pickups on gridmap at High graphics settings in real time. After that it was falling below 20fps, even if I reduced gfx quality.
Yes, x570 boards not only feature pci-express 4.0 to the video board and upper PCI-E NVME slot, they offer 4.0 on the chipset as well, so the down-link to the chipset is faster. They also have the RAM trace paths shortened on the board for tightest timings VS 3xx and 4xx chipset boards. Make sure to try A2 and B2 slots first, as they're supposedly the best for tight timings. Make sure that is what your motherboard manual tells you, it's easily overlooked and is the opposite of intel setups. 28 pickups, that's pretty good, that's a 50% gain over my 3700x here, stock speed, stock cooler (I just did myself the favor and took a store discount card and scraped off ALL the factory thermal paste, and put a quarter grain of rice worth of liquid metal on it, any more and it'd squeeze out as I didn't use alcohol to clean off the surfaces first either). I only used XMP as you did. The processor will overclock itself as-needed, so you don't have to touch it. Overclocking it will actually cost you your single-core high boost which can net 7~10% extra burst speed on single core. Each 8-core die in your CPU (you have two on 12 or 16-core chips, I have one as do all 8 core or 6 core or less chips) has ONE strong binned core programmed from the factory, and other cores are less fast (4.1~4.3ghz binning depending on chips, 3950x and 3800x have the better bins with 3950x being the strongest bin of the mainstream AM4 3000 series, only Threadripper is better). As-per RAM and Ryzen 3000 this is the skinny on it, from best to worst ram is MICRON, SAMSUNG, HYNIX. On Ryzen *2000* or *1000* series chips best to worst RAM was was Samsung, then Micron or Hynix. So entirely, if you don't have success in sourcing some now-out-of-production B-die, get Micron stuff. They're every bit as good as Samsung 'Sammy' RAM on the 3000 series. Crucial Ballistix is often Micron sourced, as are several others. There's threads devoted to RAM and Ryzen out there on computer hardware forums. I entirely am not prepared here (don't have a link handy), but I've seen them. I'm sure if I had any semblance of issue on my end, I'd surely have the link handy. Sorry for being unprepared here, entirely my failing. I would however, entirely tell you to look up what kits the reviewers used when reviewing your motherboard (best) or Ryzen 3000 series processors (almost as good but can differ by board!). If you stick to what the reviewers used, or find what works good via the plentiful hardware forums on the internet, you should have decent luck. I wish I could help more but I hope this has steered you in the right direction to help find a better solution that just 'plug and pray', which we're unfortunately still dealing with in PC's from time to time in 2019.
Recently I bought a laptop with a i5-9300H and was surprised with it's performance: It was ~21% better than my desktop with a 2600k @ 4.6ghz But sadly I'm thinking on returning the laptop, it wasn't exactly what I was expecting... Sure the performance is absurd, better than I expected really, but the "portable" part of it is disappointing I came from a laptop with a i7-4510u + gt840m, and it was COLD AF, I could play beamNG, GTA 5, with it on my lap without problems This new laptop on windows without any open program is hotter than my old laptop was while playing games And it's not only hot, it's also very noisy since the coolers enter some kid of "jet taking off" mode as soon as you open a game
I would recommend looking into whether any other laptops will be better in that regard before returning. You may find that everything with that level of performance and form factor will get rather warm to the touch. The i7-4510u is a low power processor with a 15w tdp, the i5-9300h has a 35w tdp. So depending on how the manufacturers configured the devices, your new processor may use over twice as much power as your old one did, hence it gets warmer. To make it run cooler you need either more thermal mass, or more airflow, which generally means bigger and heavier, or louder. However, it is worth noting that some laptops (certain Dell XPS models come to mind) do have insulation fitted inside them, it will have a slight negative effect on cooling (I'm sure Dell accounted for that in the design though), but it will keep the heat away from your legs a bit. But you may well find that another laptop will have quieter fans.
Yeah, I was aware of the TDP difference and stuff, but I chose this model (Samsung Odyssey NP850XBD-XG1BR) because I read in a lot of places that this model had the best cooling on this price, I saw people that returned dell laptops (with same specs) to buy this model and they said this model had much better cooling, this made me think that it would have acceptable thermals, but it seems I was wrong doing some research today I found a laptop with i5-10210u that surprisingly is supposed to have the same performance of the 9300H while having only 15w tdp, sadly it was paired with a mx250, the combination of CPU+GPU tdp would go from 105w to 40w, but it's more expensive than my Odyssey with much worse performance in games... I'm still thinking if it's worth expending more money to lose a lot of performance in games, just to improve portability and thermals (and of course there's still the possibility that this i5-10210u laptop can have terrible cooling despite the low overall tdp, resulting in still bad temps...) I really wish they gave more freedom to customize stuff, for example I wouldn't mind shutting down completely 2 cores while not in game, also lowering the frequency of the remaining cores could still be possible without a massive performance hit (my i7-4510u had 2 cores @ 2.0ghz and it was not slow in daily tasks) Another thing that i would like to customize is the fan speed curve, it seems that until 50ºC it keeps the fans completely stopped, and when they kick in, they start already in a audible speed...
You might be able to adjust things a little more to your liking: https://www.notebookcheck.net/Intel-Extreme-Tuning-Utility-XTU-Undervolting-Guide.272120.0.html I presume you could also use something like EVGA precision X to lower your GPU's power target? But I don't know if that works on laptops.
I was messing with XTU already yesterday, I can do some vcore offset, but every clock related option is grayed out for me
Wow, i would be really p*ssed... Lets hope the RMA isn't as long as it is to get a new one. I was loosing maybe 25 - 30 % of my bananabench score when I forgot to turn on XMP profile. Going from the base 2133 to 3600 mhz did make quite a difference. Though your lower than expected score may be caused by your bios. In my case I have to stick to an older version because of the 1.0.0.4 implementation by Asrock being borked and unfixed yet. It caused my cores to never "go to sleep" and stay at 3.8 ghz minimum, and big overall performance reductions (10 to 30 % depending of apps). Don't know if it's a thing for other brands but here's that. edit : Yeah, you need 1.0.0.4 anyway for the 3950X lol. On another note, as the last update changed the pickup truck, I figured out that it also changed banana results. So heres my new scores on the latest update : 3900X Stock : 409 CCX OCed (4.4 to 4.5 Ghz) : 427
Here's my new personal midrange build.. 3600X at 4.3 all core and reusing 1080Ti from my mac pro. Not bad for a $180 USD chip. Stock, wraith prism cooler. (Took from my buddies 3900x build) Got the board for $60 on black friday (ASUS B450F.)
An "ending" to my saga with the new Laptop: By adding a offset of -0.145v on the CPU via XTU and lowering the maximum processor state to 90% on windows power options, I was able to lower the CPU power consumption from 45w to 13w while gaming, this lowered the temps from more than 80ºC to around 49ºC (according to the benchmarks with the underclock it's still around 2 times faster than my old 4510u) On the GPU I did an undervolt from 1.05v to 0.68v and underclock from 2100mhz to 1500mhz, this lowered the temps from more than 60ºC to around 52ºC (according to benchmarks even with the underclock it's still around 3 times faster than a MX150/MX250) This changes definetly lowered the overall performance in games, but now at least I'm able to play on it while on my lap without getting first degree burns and the fan noise is also greatly reduced I was able to play GTA 5, minecraft (shaders) with my wife and Assetto Corsa with it on my lap, all of them ran pretty well above 60fps most of the time Sadly BeamNG was the only game I tested that became almost unplayable, it's still able to surpass 60fps with ease, but for some reason the framerate now is all over the place, with stutters happening all the time, I'm still trying to think how to fix it EDIT: fixed the stuttering by changing the "minimum processor state" to 89% and it's still below 50ºC Still, I decided to keep it, the only other option I would have would be expend more on another laptop with a GPU at least 3 times slower, and I can't be sure that it would really be colder
Since version 0.18 brought the revamped D-Series, scores changed a bit, so I added a 4th page (New Banana Bench (Dynamic Collision ON) Ver.0.18+) in the spreadsheet with the results after 0.18. There has been a 9% loss of performance particularly on a CPU, the R9 3900X of @Patate_Folle and a 3% loss in my case, going from 145,145 to 140,031 Mbeam/s on a non-overclocked i5-7600K running during the bench at 3,98 Ghz.
Ryzen Master OC CCX0 is 4.350 and CCX1 is running at 4.300 GHz.. if OCed in the BIOS will not let me pass a 4.250 all core OC.. it will crash instantly (cinebench) at a 4.250 all core OC no matter what is done. Thats the reason why ryzen master is being used.
I'm getting almost 100 points less compared to 4 months ago. The CPU fan doesn't really start spinning and it feels like this bench/version doesn't tax my CPU like it did before. Clock speeds hit about 4Ghz (but mainly stay at 3.8) with the temperature at around 52°C. Cinebench R20 still gives me 7250 points like 4 months ago so I doubt it's an issue of the CPU or throttling. @stenyak maybe this was causing you lower scores as well? Someone posted a 400 score before with a 3900X at stock, so I'm wondering what could be the cause of the issue here.
So since I'm still on my old haswell processor (i5-4690K) in 2020, I've ran the bench again. 140.603 Mbeams/s at 4.5 GHz, not bad for an almost 6 year old processor I think.
Spreadsheet updated with a few more results. BIG UPDATE: the tabs are now organised from newer to older results (first tab has now the newest results, not the oldest), as suggested by @Specht77.
Thanks @Alex [ITA] ! And to celebrate, I decided to make a new run on my desktop (2600k) Here's the result: Not bad for a 9 years old CPU
Is the list of results current? Edit: looks like no for the one on the first post. Is there a current running list? Edit: Did some basic testing. I ran it once after a bios reset and got around 450Mbeams/s. Setting the bios to the auto overclock gave it around 600Mbeams/s. I also accidentally disabled a stick of ram for the second run. At stock everything 71 vehicles is around 15fps. I havent tried it yet with the overclock. Cpu usage sits around 85%-90%.
Yes, this is still kept updated by me with this spreadsheet. I have to contact the forum moderators in order to add the updated spreadsheet link to the first post or redirect the people to page #50 where I've first posted the updated version.