yosarianilives

Stages look good! Excited to get started this weekend.

If you go to the hwbot comp page it says that agp, pci, integrated, hybrid, pcie meaning that you could have a lineup that doesn't use pci at all. If it's intended to use pci then will have to up to 5 subs or drop a gpu socket. If a socket is dropped pcie is probably the most balanced to drop since with no restriction it could be a rtx titan, whereas the max the next fastest socket, integrated, could be is a 3400g.

Crazy score on crazy setup

Yeah I definitely would agree that the best possible score is gonna be to bin ddr3 boards for high bclk. Also remember that there are some subtimings that you still can't set in bios so certain boards have slightly higher ipc than others. For example C4E gets higher 32m scores than C4F, could end up mattering for something like GB3 or even r15.

I was testing stuff just to test then I found this team competition for asus rog and that got me started. I really only show up for the team comps usually but I do like to follow some of the other interesting comps, especially like cheapaz and legacy comps. Other than that benching is honestly kind of a escape from depression/anxiety in much the same things like video games or most other hobbies are for some. When I'm benching things are pretty straightforward with a clear goal and enough challenge to be interesting. If I get frustrated hopefully I end the session before damage is done to HW

This is great info! I definitely have found that usually you can get higher fsb on 45nm chips around -100c but you also need more cold for more core sometimes so definitely need to bin for good bclk at lower temps, ideally fullpot.

I mean the main post of this thread is calling me out on my stance on Linux by mis conflating what I proposed and creating a sort of a strawman by referring to me as "someone" and putting words into my mouth.

Leeg is a great guy, but saying that Linux is impossibly hard to moderate is silly. If a score is out of line then demand an acronis image of the OS, it's not like the records are set on more than 3 boards anyways. If user won't provide their tweaks to a mod they're cheating and the result is blocked. Ez pz. But that's not the reason for blocking Linux that I read in the conversation that this thread is about. The reason for blocking Linux is a mod personally didn't want to have to do research on Linux os to be able to make competitive results before somebody else does. Or at least that's the way that mods comments read to me. Now it is being claimed that the reason for blocking linux is that as a mod you don't want to be looking at a result wondering if it's legit, which is a legitimate complaint although Imho its solved by requiring a system image from questionable results. Anyways it doesn't matter what I say here cause the members of this thread have predecided that I'm a cock. I'm sad to have apparently contributed to leeg leaving, he's a great mod who really cared and that seems to have been his downfall. He cared enough to let the job get to him and there's far too much abuse in the job unfortunately from those like me and from the many. Way too many people getting pissed off when their win 10 results get removed despite never being allowed within the rules, or their cine score with the render is removed, etc. Definitely far too thankless a job.

Assuming skydiver will be open to all gpus eg rtx titan, 2080ti, 2080, titan v, etc, as is catzilla, as will the pcie score for 3dm06. And for 3dm03 will it be by socket (1151, 1150, 1155, etc) or will it just be 4 unique igps? (9900k, 7700k, 6700k, 6770hq)

Would the same trick also work for those ATI Stream gpus as for intel? I know they wouldn't be very fast but neither is intel igp. Even on lake igp it takes days to get a 32b score.

Also for the 3dmark11 xfire stage on am3+, is it also allowed to use dual gpu cards for both? My plan was to use the same R9 295x2 for both if possible

Was supposed to be dual gpu cards, ie 2 gpu 1 card. However with that the case can we not do timespy? Nothing pre gcn or pre Fermi can run it. Which gives a total list of dual gpu cards of Radeon pro duo (Fiji, Polaris, Vega), R9 295x2, Gtx 690, titan z, Gtx 590, HD 7990. Can we do a dx11 bench so we can allow a few more possibilities in an already scarce playing field of potential cards?

Is there any plan/possibility to add non opencl support for graphics chips that have hardware support for double precision but not opnecl driver support? for example hd 3000 series supports double precision through ATI Stream but not OpenCl or Ivy and Haswell IGP both have double precision and support it through driver calls however there is no opencl driver for them.

The gpupi OR ycruncher stage is per socket. Is it also going to be per core or will it be an absolute score. Ie; 7740x or 9980x on x299 wins?

I think that's actually a good idea, or at least dump pts for geek like we did for XTU. If we dump geek then we'll need a new mt and st bench for global pts. I propose ycruncher 1b and maybe pifast for st. Not too many other ST benches unless we add 1m or testcpu.

I'd be good with ditching all geek on the bot but some people may not like their work for naught. IMHO if we keep geek we keep all geek, kinda like how we have every 3dmark.

Presumably it could be changed to allow submission the same as the other geeks or are we dropping geek all together.

Was surprisingly easy, the android-x86 project is pretty far a long now and comes with mesa 18.6.3 built in meaning any amd gpu except navi should work out of the box. I'd go almost as far as saying easier than windows benching since I didn't strip to OS. Although if I had competition I would need to recompile as currently it's hard locked at 8 threads and vsync is still enabled.

Ideally this stage is by cpu socket, although then you have to do things like ban broadwell and mobile chips otherwise the lineup is all about obscure igps. imho doing it by gen would suck as hd 530, hd 630, and uhd 630 are literally the same IGP and I don't want it to just be getting scores on lake cpus.

The new submitter works really well, although I don't know how it works in relation to comps. Old whycruncher required network connection to submit through the app, new ycruncher makes a datafile that can be saved offline.

x265 swapped from 2-3-4-5 to 1-3-4-5, assuming that's intentional. And assuming that r15 single means single core cpu and not r15 single core score.

Not sure why but won't let me attach all images to sub, the rest hosted here https://imgur.com/a/WB4W6pm

Bench runs without emulation, I see no breaking of rules. The rules merely specifies that android OS must be used, which is what I use. If you're worried about lack of competition then I will tell you that android x86 oreo currently is compiled to use up to 8 threads and has mesa 18.6.3 driver allowing use of AMD gpus up to vega arch. So anyone who wants the team pts can run 9900k ht off and vega VII and easily obliterate my scores.

Not sure why but won't let me attach all images to sub, the rest hosted here https://imgur.com/a/WB4W6pm

Not sure why but won't let me attach all images to sub, the rest hosted here https://imgur.com/a/WB4W6pm