marco.is.not.80 Posted June 28, 2019 Posted June 28, 2019 (edited) Guys/Girls, I now have a tick because of this benchmark. Can someone please share with me what the revolutionary frackin' secret is to getting some of the epic scores from this comp: https://hwbot.org/competition/Challenger 2019 Division I round 2/stage/4616_unigine_heaven_-_basic I've put in 4 days of trying everything you can imagine and I get the same score no matter what within 30 to 40 points. Realizing that this is a severely CPU limited benchmark I have tried: - maximum memory frequency all the way up to around 4700 Mhz and got the same score. - lowest possible timings at 12-12-12-28 CR1 with good subtiming configuration and got the same score. - spent an entire day on BCLK although I noticed none of the of guys I'm competing against were doing that but what the hell. 102 x 49, 105, etc. made no difference. - must be timer related, right? Windows 7, Windows 10 pre 1809 with 3 mhz TSC timer and Windows 10 post 1809 with the 10 mhz FrankenTimer - doesn't matter - same score. - maybe some 16 GB Double Sided sticks will do some magical hocus pocus - ya... no. The only thing that would move my score was upping the core of my processor past 5 ghz which is over the amount that the competition allows. So what the fuck is the magic sauce that is allowing shit like 9650.75 DX9 Marks? I tried all the LOD hacks, etc. just in case somehow they made a difference but like I said - CPU limited but literally have run out of ideas and could use some insight before I pluck my eyes out and feed them to my turtle. Marco Edited June 28, 2019 by marco.is.not.80 Quote
Crew Leeghoofd Posted June 28, 2019 Crew Posted June 28, 2019 LOD via Supersampling or replay mode ? Quote
marco.is.not.80 Posted June 28, 2019 Author Posted June 28, 2019 Tried both of those - had no bearing on my score because as already mentioned - the benchmark is CPU bound - but I did try both of those methods to see if they had any impact. Quote
Guest Digg_de Posted June 28, 2019 Posted June 28, 2019 (edited) D3D Overrider? How much Pts you get? Ok found your Score: 9283.14 Edited June 28, 2019 by Digg_de Quote
marco.is.not.80 Posted June 28, 2019 Author Posted June 28, 2019 Digg: My submission is here: https://hwbot.org/submission/4183075 D3D Overrider? I mean, if it's for LOD isn't using nvidia profile inspector the same thing or ... did you just give me the insight I needed :-) ? Marco Quote
Guest Digg_de Posted June 28, 2019 Posted June 28, 2019 (edited) You can find D3D Overrider in Rivatuner Folder if isntalled, or download from here: Global->High detection Lvl->Force Triple Buffering on->Vsync off. Edited June 28, 2019 by Digg_de Quote
marco.is.not.80 Posted June 28, 2019 Author Posted June 28, 2019 But It is for triple buffering and vsync right? Those are already forced off by my driver... Quote
Guest Digg_de Posted June 28, 2019 Posted June 28, 2019 I would try to see, if it is the same or not. I've never tested the driver thing i guess. Quote
yosarianilives Posted June 28, 2019 Posted June 28, 2019 1 hour ago, marco.is.not.80 said: But It is for triple buffering and vsync right? Those are already forced off by my driver... Forcing triple buffering from d3doverrider isn't the same as through driver for some reason, anyways basically what it does is let your GPU render a few more frames in really high fps benches where the gpu isn't holding it back much. Essentially caches extra frames. Similarly using a higher refresh rate monitor will also help score in high fps benches. Other than that if you're cpu bound best thing is maximum cpu core and cache (duh) and then highest frequency you can run with bench timings (4k 12-12-28 1t) also make sure to test subtimings, some do more than others and you may find that a few subtimings need to be cranked more than others. Quote
Splave Posted June 28, 2019 Posted June 28, 2019 (edited) priority? affinity? HPET vs RTC? Removememory instead of maxmem? Edited June 28, 2019 by Splave 1 Quote
yosarianilives Posted June 28, 2019 Posted June 28, 2019 26 minutes ago, Splave said: priority? affinity? HPET vs RTC? Removememory instead of maxmem? The master shares the sekrit sauce Quote
marco.is.not.80 Posted June 28, 2019 Author Posted June 28, 2019 3 hours ago, Splave said: priority? affinity? HPET vs RTC? Removememory instead of maxmem? Splave and Yosarian, Thanks for responding. Changing priorities from normal, high and realtime made no difference for me. HPET VS RTC: I wish I had more control over this but my Apex XI doesn't have a BIOS settings for HPET but my comments in my initial post about Windows 7, Windows 10 pre-1809 and Windows 10 post-1809 covered this correct me if I'm wrong. Don't know anything about removememory so I will read up on that. Obviously I use the maxmem with the 12-12-12-28 and lower timings. Affinity is an interesting question. Let me play with it and see if I can find any measurable difference... RE: higher refresh rate monitor - mine is 144. I never thought that maybe the guys scoring higher than me might be using 240 or higher... Man, that would suck. Getting beat because of refresh rate. LOL. I'll post back when I have run some more tests. Quote
marco.is.not.80 Posted June 29, 2019 Author Posted June 29, 2019 (edited) Well, 17 hours later I've increased my score to 9325... A whopping 50 or so points not even worth resubmitting. I've tried hyperthreading off with auto-affinity - no increase. I've tried hyperthreading enabled with auto-affinity obviously - that was default for me. No increase. I've tried assigning 2 core/hyper/no-hyper, 3 core/hyper/no-hyper, etc. The benchmark tanks with only two cores hyper/no-hyper and it seems like 4 cores is needed to achieve a CPU limited situation. Anyways, the point is - I've tried almost every possible configuration there is and the best I could was a 4 core non-hyperthreaded run that got me 9325 after the third run. So obviously I'm not going down the right path considering the highest entry right now is 9650.75 which nothing is indicating that I'm even close to capable of making a run like that. Considering this is an event where we are all limited to 5 ghz it makes no sense to me other than there is a damn good tweak out there that I'm simply not aware of and can't find. Obviously don't expect anyone to tell me either so I'm going to order a monitor today and have it FedEx'd overnight delivery and it was made in a science lab at NASA and can do a fucking 9,000,000 mhz refresh rate and jack me off with robot arms at the same time you fuckers... SIGH... I'll keep pushing. Marco Edited June 29, 2019 by marco.is.not.80 Excuse my french Quote
techjesse Posted June 29, 2019 Posted June 29, 2019 Marco, Your running Heaven in Win7 .... Run it in Win 8.1! With LOD via Supersampling ...... Quote
marco.is.not.80 Posted June 29, 2019 Author Posted June 29, 2019 57 minutes ago, techjesse said: Marco, Your running Heaven in Win7 .... Run it in Win 8.1! With LOD via Supersampling ...... Seriously? Ok, at this point I'll pretty much try anything but it doesn't look like any of the guys above me (and just noticed you and I are ranked together) and you are using Win8? Man, I gotta find a Win8 image somewhere. This will be the first time I've ever even installed it. Quote
techjesse Posted June 29, 2019 Posted June 29, 2019 The best scores are achieved running Win 8.1... https://hwbot.org/submission/4175922_ 1 Quote
techjesse Posted June 29, 2019 Posted June 29, 2019 Marco, don't feel bad for your scores, there honest! Here's how some of the guys are doing it: First set TurboV to 5.2Ghz, run the bench...set TurboV to 5.0Ghz Boom you have it... it's a cheat! They need to load a vid to check for sure. Quote
marco.is.not.80 Posted June 29, 2019 Author Posted June 29, 2019 Ok, I do feel better actually because I don't mind losing to an impossible score and spending lots of time trying to do something I never would be able to actually accomplish. Thank you for unburdening me! The thing is, when they submit the hwbot file from the benchmark won't it show that they were running past 5 ghz (in your example, 5.2)??? Leeg? Quote
techjesse Posted June 29, 2019 Posted June 29, 2019 (edited) I saved the file, I'll make the sub.... https://hwbot.org/submission/4185027_techjesse_unigine_heaven___basic_geforce_gtx_1080_ti_9455.1_dx9_marks/ Edited June 29, 2019 by techjesse Quote
techjesse Posted June 29, 2019 Posted June 29, 2019 Also some overclocking tools will not reduce the vcore volts... easy way to spot a cheat.... Quote
bigblock990 Posted June 29, 2019 Posted June 29, 2019 Since I had 2080ti setup testing other benches, thought I would jump in here and see what all the fuss is about. This is cpu on aio, and card on stock air cooler. I just did a couple runs quickly using the tweaks I know, and my inspector profile from a different gpu. So I'm sure with some actual effort I could improve this. Don't get frustrated, tweaking for 3d takes some time 1 Quote
marco.is.not.80 Posted June 29, 2019 Author Posted June 29, 2019 Damn you Big Block. ? Is that Windows 7? Anyways, enough people are scoring like that which means somewhere out there is the information since I highly doubt everyone magically discovered the tweak on their own but now that I know it's not actually a cheat I will eventually stumble on the solution during my testing. All I can think of is that there is one specific LOD setting out of the many combinations that this benchmark responds to and there is no hint at which number that is because if you go one over or one under you are going to get the same score as if you never turned it on. Anyways, back at it. Thanks for encouragement - you can have dibs on the robotic NASA monitor that will be in the for sale section once I've gained the world record in this benchie ? Marco Quote
ksateaaa23 Posted June 30, 2019 Posted June 30, 2019 (edited) more or less same tweaks discussed here are used. consistently i was doing around 9600 with my gtx 1080. yesterday had half a dozen scores ranging from 9570 to 9610 but couldn't improve my best score of 9617.88 for some reason my gigabyte ultra gaming scores slightly better than asrock gaming k6. https://hwbot.org/submission/4163773_ksateaaa23_unigine_heaven___basic_geforce_gtx_1080_9477.6_dx9_marks/(without maxmem) https://hwbot.org/submission/4181264_ksateaaa23_unigine_heaven___basic_geforce_gtx_1080_9617.88_dx9_marks/(with maxmem=4gb) https://hwbot.org/submission/4163779_ksateaaa23_unigine_heaven___basic_geforce_gtx_1080_9890.37_dx9_marks/(processor @5.2ghz) on asrock board run right after the score of 9477.6 was done. at 5.2ghz cpu, 10000 is possible. and will run with gigabyte board. i don't suspect any scores there. Edited June 30, 2019 by ksateaaa23 1 Quote
ksateaaa23 Posted June 30, 2019 Posted June 30, 2019 (edited) this is i7 8700k@5.2ghz https://hwbot.org/submission/4185333_ksateaaa23_unigine_heaven___basic_geforce_gtx_1080_9999.2_dx9_marks @5ghz, still trying to beat a score which is better by only 1.4 points for 3 days and can't beat it so far.(9619.28>9617.88). edit: finally after almost 15 runs got 9626.3 https://hwbot.org/submission/4185337_ksateaaa23_unigine_heaven___basic_geforce_gtx_1080_9626.3_dx9_marks Edited June 30, 2019 by ksateaaa23 Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.