Jump to content
HWBOT Community Forums

Monabuntur

Members
  • Posts

    22
  • Joined

  • Last visited

  • Days Won

    1

Everything posted by Monabuntur

  1. If pyprime was a multithreaded benchmark that CPU would jump out of the socket! Anyways, great score!
  2. absolutely, zen2 is rather limited in that regard, and PYPrime's affinity for low memory latency definitely isn't helping
  3. It depends, Zen3 and Comet Lake are basically on par (architecture wise), the main problem there is that most of these AMD CPUs are limited to about 4000MHz on the ram (because of infinity fabric limitations), sure, you can get it higher, but that will increase latency which will seriously hinder performance; your main avenue for increasing performance is tightening the timings and overclocking the CPU itself. Older zen based architectures don't do as well, this is mainly due to the relatively high memory access latency and the lower maximum infinity fabric clock, this is also the reason I think Rocket Lake won't be as fast as Comet Lake and zen3 in this benchmark. I hope this helped!
  4. same for me, the aspect ratio seems to be somewhat odd on newer phones
  5. I have this problem too It continues as a blank page Some time ago the error was slightly different, but unfortunately I don't remember the details
  6. Hi, I was looking through the submissions to see how well the benchmark scales on other systems with different ram, and I noticed you got a relatively low score for your setup (which is basically the fastest PC you can buy at the moment); Since this is a new benchmark and people might not know yet what to prioritize, here are some suggestions to get a better results: You might want to get the FCLK/UCLK (infinity fabric and memory controller clock) to a 1:1 ratio (I'd start with 1800MHz, then try to get it to 1900MHz), then lower the DRAM clock to 2x that of the FCLK/UCLK. After that you can start tightening the timings, CL20 is quite bad, I think you should be able to get it to at least CL16, the results here will depend on your memory though, but generally the lower the better! Sorry for being annoying, good luck with your benchmarking!
  7. PYPrime 2.0 is a synthetic benchmark for ram, when overclocking keep in mind that it scales very well with RAM clock and timings, but at a certain point you might have to overclock the CPU too when that becomes the bottleneck! To try it out you'll need the latest BenchMate version (0.10.7.1), Good luck and have fun and with your OC!
  8. The cheaper one? The 10900KF is a 10900K without the iGPU, and unless you need it it's gonna be the better buy, many disable it anyways when overclocking To be completely honest though I would wait for the 11th gen, it should come out fairly soon (about 2 months) and unless you need a 10900K right now it will be the better buy
  9. So, I can easily implement the screenshot (I hope at least), python can take screenshots, but I'll have to see if I can integrate it with the hwbot API For the manual option, it's a bit more complicated, you can't manually input it unless it's not in the database (it happens often with motherboards), what I can do, is ask at the end of the benchmark if you want to automatically fill in the hardware and generate a file with a different structure, but this complicates the usage a bit.
  10. This benchmark is very easy, you have to run the benchmark and submit the verification file. There are no restrictions
  11. So, after some work I got it to work. To be able to submit a result to hwbot you need a specific version, that runs only on windows as it's an .exe (if you don't want to submit anything you can use the .py script that doesn't have the security features). Then to submit it you need to upload the encrypted .hwbot file (glorified xml, literally , it's an xml encrypted with AES 128) with score, os, motherboard, ram and all this snazzy stuff. I also compiled a version for linux and macOS just because I can
  12. So after a good day of me basically wanting to die I figured out the problem, the hwbot api wants the key and initialization vector encoded in hex notation, while I have to use it in base64 inside python. That's why the iv is 32byes long in hwbot while it has to be 16 bytes long in python!
  13. I'm updating my benchmark to use encrypted datafiles, and I get this error when I try to upload something. Without encryption the datafiles seem to work just fine, but when I add it I get this error. I am using AES 128 CBC. this is the unencrypted .xml file datafile.xml and this is the encrypted one datafile.hwbot when it goes public I will change the key, so here it is 3778214125442A472D4B615064536756 and the initialization vector 6E5A723475377821 The iv has to be 16 bytes otherwise cryptodome, the library I use, launches an exception. this is the function that creates the .xml and encrypts it def datafile(second): time = datetime.datetime.now() submission = ET.Element('submission') application = ET.Element('application') name = ET.Element('name') version = ET.Element('version') score = ET.Element('score') points = ET.Element('points') timestamp = ET.Element('timestamp') name.text = 'PYPrime' version.text = '1.4.1' points.text = str(second) timestamp.text = str(time) submission.set('xmlns:xsi', "http://www.w3.org/2001/XMLSchema-instance") submission.set('xmlns', "http://hwbot.org/submit/api?client=pyprime&clientVersion=1.4.1") submission.append(application) application.append(name) application.append(version) submission.append(score) score.append(points) submission.append(timestamp) file = open("datafile.hwbot", 'wb') file2 = open("datafile.xml", 'wb') key = b'3778214125442A472D4B615064536756' iv = b'6E5A723475377821' cipher = AES.new(key, AES.MODE_CBC, iv=iv) file.write(cipher.encrypt(pad(ET.tostring(submission, xml_declaration=True, encoding="utf-8"), 16))) file2.write(ET.tostring(submission, xml_declaration=True, encoding="utf-8")) file.close() I've been on stuck on this thing for about 10 hours, I hope someone can help me.
  14. So usually these linux topic don't end well, but is the Linux subsystem for windows allowed in cases like Geekbench? I can use CPU-Z and everything but the benchmark itself technically runs on linux, and the performance is slightly better.
  15. I'aware of this flaw (it's stupidly easy to modify the code) and I'm working on it, the easiest way would be compiling it to .pyc, but it would limit compatibility (as only the version on which it has been compiled will work), but this isn't the only way to do it... Anyways, thanks for the feedback, this really helps me make this better
  16. You can use whateaver os you want, but Linux (or the Linux subsystem for windows) would be a better overall choice, you can boot off a USB stick or use the aforementioned Linux subsystem for Windows Good luck!
  17. Is there a suggested API to get all specs in the computer? I am using python to make a multi platform benchmark (yes, python) and I have sorted most of the basic things out, in fact you can try it out (its called PYPrime, you can find it in the benchmarks section) but there are some, SOME, security issues I'm working on, being open source it's way too easy now to cheat, so I am now implementing datafiles to add some security, but I still can't reliably get the system specs (like memory timings, memory speed, motherboard vendor) ,heck, even the CPU is a bit hard to find in some cases. I would like to avoid using external libraries to make the installations as simple as possible, that's why I used this kind of things: def get_processor_info(): if platform.system() == "Windows": print('CPU: ', platform.processor()) elif platform.system() == "Darwin": print('CPU: ', bytes(subprocess.check_output(['/usr/sbin/sysctl', "-n", "machdep.cpu.brand_string"]).strip()).decode()) elif platform.system() == "Linux": command = "cat /proc/cpuinfo" print('CPU: ', subprocess.check_output(command, shell=True).strip()) return "" The worst part is that I can't seem to get access to the API documentation because this weird problem <response> <status>error</status> <message>Unable to process data request, please contact HWBOT if this issue persists.</message> </response> that appears every time I try to open the link, sometimes It happens when I try to view some benchmark submissions
×
×
  • Create New...