Jump to content
HWBOT Community Forums

mickulty

Members
  • Posts

    544
  • Joined

  • Last visited

  • Days Won

    4

Everything posted by mickulty

  1. I think this thread is predicated on the idea that it would be nice if 3D benching wasn't a case of having to buy thousands of dollars worth of hardware or just 'go bowling' instead. What you have to remember is back in 2010 the GTX 580 was only $500 and you could be reasonably competitive with a 570 at $330. Then come late 2011 we had the 7970 at $550 and the 7950 close to it and at $450. But now we have a situation where some of the time you need to pay $1200 to be competitive, and the rest of the time the sub-titan cards are $650+. This is not how it's always been, the market has changed drastically. The 2D rankings are and continue to be perfectly fine with per-core categories and are doing an excellent job of making everything from old singlecore semprons and athlon/phenom X3s to i3s, i7s and LGA2011 i7s viable. Surely just wanting a comparable situation for 3D rankings is not such a horrifying goal? It would be a problem if better-separated rankings ended up kneecapping Titan scores. Perhaps at the moment a simple solution would be an expansion of WR points so the best Titan users are still properly rewarded? If nothing else though, if it becomes the case that a 1060 is the card you need for the most points, I'm sure a lot more titan users could afford a 1060 than the number of 1060 users who could afford a titan By the by, here's a thought experiment for you. Imagine we had 3D division, based on something like ROP count or launch MSRP (I still think ROP count makes way more sense because it's a tech spec, launch price is a bit arbitrary). Imagine the 1060 and 1080 are currently the two best GPUs for global points. Now imagine someone wades in and says "I think the 3D global ranks should be merged, I paid $1200 for my Titan XP, I want more global points dammit! 1060 users should stop trying to compete and just go bowling!". Imagine how well that would go down
  2. The download link for the HEVC h.265 Decode benchmark is broken. It goes to a onedrive page informing the user that "This link doesn't work any more". Please could it be changed to a working link so I can continue to collect gold cups by benching stuff no-one else can be bothered to bench?
  3. Could you add the be Quiet! Dark Power Pro P10 1200W as well please? As seen in nearly all of my subs: http://hwbot.org/image/1684673.jpg
  4. I hadn't thought of the effect of people going for globals on hardware points. That makes the prospect of all cards still becoming obsolete with new releases in their price range a bit more palatable. I don't envy whoever has to collect price information for every GPU on hwbot though and I'm still not sure what would happen to the rebrands like 7970 ($549) -> 280X ($300) that are currently having their categories merged - do you screw over the 'new' card by keeping it in the high category, or give an unfair advantage to the old card by dropping it to the price point of the rebrand? A 680 has 32 ROPs, so the only more recent nvidia cards it'd be against are 960s and 950s and that's a pretty fair fight ROPs may be a bit technical but it's easy to look them up, they're in wikipedia's lists of nvidia and ati stuff. In fact, once people overcome the innate fear of a vaguely technical term it's at worst equally easy to find out as launch MSRP and in nearly all cases (lol 970) a figure that carries more certainty than MSRP does with modern shenanigans like 4GB unicorns and FE cards.
  5. I wish I'd known this before I bought geekbench.
  6. It'd definitely be a massive improvement over £1200 but it would still kinda suck having to chose between getting a decent amount of hardware to play with and only benching one thing every few months but getting more points. I'm also concerned that vendor division relies on having a situation where AMD don't have a superexpensive halo card which I'm not sure will always be the case.
  7. Guess I should have another go at in-os overclocking, my asrock A88M-G/3.1 won't boot above 106mhz (which is a setting of 105mhz in the bios because asrock).
  8. I really really don't think an AMD ranking would be a complete solution. The cheapest I can find a Fury X in the UK right now is £375 on ebay which really isn't all that accessible so if the goal is to make it possible to get global points without massive outlay then that wouldn't work. Maybe in the inevitable Matrox, SiS, 3DFX, S3 etc rankings but most of those will only run aquamark and the super early 3dmarks (not sure where 3dmark01 went?), plus that hardware is getting rare and it becomes a test of who can get the drivers to work. The problem with launch price is right now it's not clear what the launch prices are on some cards. Looking forward to the arguments when some card has an MSRP of $499 but an FE price (nvidia) or price it's actually available at (AMD) of $549? I wouldn't Also things like the 7970 and 280X are literally the same card to the point their rankings are in the process of being merged and launched at totally different prices. Also ROP count would help older cards stay relevant and worthwhile OC'ing because the lower ROP counts are obselete, same as how no-one makes a 1-core CPU. Launch price would make every single older card obselete the moment something else launched in that price range. It would still be better than the current situation and something I'd be happier to see than just a vendor split but I think it has issues. Maybe I'm biased because I own a radeon 7000, which is the fastest 1-ROP model, but really I just really like the idea of people being able to buy a truckload of dirt-cheap old cards because that was how I started out. Nice and relaxing voltmodding a £5 card that you don't give a crap about, rather than a £50+ card Cheapaz chips is good for that reason, but having a system that gave globals for ancient cards would be like having a constant, even cheaper cheapaz chips.
  9. I defy you to think of a better way way of doing that than ROP count Other than manual categorisation which would become the subject of endless arguments. The thing is there's no reason to think AMD might not also release an overpriced halo card, and the moment that happens we're back to square one. The Fury X still isn't all that cheap.
  10. Well, CPU rankings do go up to 96 cores... but more practical would be some grouping. There's no need for, say, a 970 to be in a different category to a 980 IMO. I don't think number of SPs should count towards it because performance per SP varies hugely with different architectures. The SP*ROP figure for a Fury X is 2x that of a 1080 for goodness sake! 5870 vs 480 is a good historical example of the pitfalls - going by ROPs alone the 480 would be one category up from the 5870 which is probably about right, going by SPs and ROPs the 5870 would be in the category above the 480! ROP count alone is not a perfect leveller but it's a pretty good one, it'd be simple to implement and would achieve the goal of opening up 3D globals to people on tight (even extremely tight) budgets. Newer generation products would still have a big advantage in their class but it would still be lessened compared to the current situation and in 1, 2, 3-4 and 5-8 ROP rankings it seems unlikely there will be any new products so they would end up like 1-core and 3-core CPU rankings. As for older architectures, just group them all under whatever the lowest category is (say, 1 ROP)? At that point they're so old, and faster cards are so cheap, it can't possibly be pricing anyone out of global rankings. BTW, IMO there shouldn't be a need for separate AMD and Nvidia rankings. If you classify products by ROPs then while there are some outliers (Fury X beating a 980 in the 64 ROP class, for example) everything is generally pretty close and I'm not sure it'd favour either. Nvidia's halo strategy means they will probably continue to have the fastest card most of the time but once that's kicked to the 65+ ROP rankings the rest of the competition between gpu designers will be very close.
  11. Couldn't not share this - my 2900 XT just got nearly 4x the current hardware record in 3Dmark03. Sadly, I didn't manage to get the core to 6GHz and memory to 4GHz - the card has serious display artifacts in 2D mode and didn't actually render anything at all in 3D mode which I'm pretty sure isn't an allowed tweak The hunt for working R600 continues...
  12. I really like the sound of ROP-based rankings as Sam suggested - per-core rankings obviously work well on the CPU side and ROP count seems to follow a comparable pattern of per-ROP performance increases and a slowish increase in ROP count over time. To avoid an explosion in the number of rankings there would have to be groups - maybe 1, 2, 3-4, 5-8, 9-16, 17-32 etc? Of course one or two cards would still be slightly disadvantaged but never any more than they are by the current system.
  13. I'm 23, and still haven't finished uni - failed the electromagnetics and AC bits of electrical engineering after two years so I changed to computer science which I'll shortly be starting my third and (hopefully) final year of. I've collected old computer hardware, especially graphics cards, for a while already and for years I've also been a cooling enthusiast obsessed with making my system as quiet as possible even when I was running 3-way 4870s or later 6950s just to use my system as a space heater so when I came across buildzoid's streams competitive overclocking became the logical next step for me. What motivates me is really a feeling of achievement. Whether it's improving a score, collecting hwbot cups or improving my ranking (not gonna pretend I don't hope to hit #1 novice league at some point in the next few months, unless I end up jumping to apprentice first), overclocking is full if things you can achieve and in my experience if you can get your hands on a good amount of cheap, old hardware to play with the reward is pretty much proportional to the amount of time and effort you put in which for me is really motivating. As brilliant as it would be to end up working with/for a hardware manufacturer and effectively having overclocking or an aspect of it as part of my job, I'm careful not to kid myself that it's likely. I'm hoping to end up working as a programmer of some sort and avoid web dev if I'm lucky, ideally something involving microcontrollers if I'm really lucky.
  14. Really sorry - looks like it might not be dead. I hadn't expected it to work, just gave it a last try because I knew I'd always wonder otherwise, and it did actually post. Hopefully someone else (maybe in germany?) will have an actually-dead chip available. Again I'm really sorry to have messed you around.
  15. Awesome, thanks I hereby claim my 2 hardware points for reference clock Also left a comment on the hardware page with a link to the PLL datasheet.
  16. PLL is an IDT 9LPR363, datasheet at http://www.datasheets360.com/pdf/-266145806877066915 - can be overclocked using rweverything to change clock control bits if you know what you're doing
  17. Ooh, I'd totally forgotten about that, that's a damn good point. Not sure about geekbench but I can think of a few CPUs that would easily beat a 6700k in hwbot prime per-core for a pittance. Now I'm tempted to enter even though I doubt I'll even be able to submit for XTU...
  18. If you take a skylake i3 to 103mhz+ base clock it disables some power management thing which which is needed to activate the avx part and XTU scores fall off a cliff, that's why there's a massive 742 club. What core divisors do succeed in is stopping people from winning by spending $2000 on X99.
  19. That's no different to the 6600k having newer AVX instructions the 4690k lacks though. Except it's not a 4690k that you own.
  20. To be fair it'd be great to see an AM1 competition. Could even provide a mix of 2D and 3D. (says the guy who hasn't got round to voltmodding his 710 yet...) Guessing radeon 2000 series, what was the second?
  21. I'd be surprised if no-one a bit closer geographically has anything but I have a dead G3258 I could be persuaded to part with if necessary (tbh was planning to ebay it at some point as just "not working" without mentioning that the reason was I stupidly threw too much input and cache voltage at it and the heat put my cheap clc pump into thermal shutdown).
×
×
  • Create New...