mickulty
-
Posts
544 -
Joined
-
Last visited
-
Days Won
4
Content Type
Profiles
Forums
Events
Blogs
Posts posted by mickulty
-
-
not sure how this fixes the issue at hand honestly.
Lets take the last generation of cards as an example, the 980 and 970 would most probably be in the same price category as the 980Ti since their prices were slightly inflated due to them being the fastest products on offer at the time. Then Titan X comes along, is realised for a stupid price. Is in it's own category, great. But then 980TI is released and is in the same category as the 980 and 970 no doubt so the bunnying will probably start all over again as those with money can capitalise on the easy points with the Titan X being in its own category and the 980Ti dominating its category easily.
TL:DR This doesn't seem like a solution but more of a hornets nest to a whole new era of "money wins in overclocking" since your rich OC'ers will LN2 everything, bench all of the new GPU categories with their golden CPUs so the little guy can't even get close anyway.
I agree that the 980Ti ending up categorised alongside the 980 would be bad. ROP count would suffer less from this issue
You're 100% right about the top still being ruled as much by dollars as dewars but you'd have new low-midrange categories that should hopefully get a lot of participation and would be somewhat insulated. Hopefully they'd be less CPU dependent in some benchmarks too since the GPUs would be slower.
Also if we want to split AMD and Nvidia apart then we might as well do it on the CPU side too.I think people have decent ideas but are not really thinking them fully through here. Yes what we have is not perfect, but that is simply because AMD are not competitive right now. Most of the issues we have with leaderboard rankings etc all come down to the fact that Nvidia and Intel have a near monopoly in their respective areas so the price of the products are obscene now.
This is an argument that I've seen put forward in a lot of places, starting in 'why we need amd' blog posts and moving into intel and nvidia fanboys on reddit upset about 6950x/titan prices ranting at AMD for allowing it to happen rather than blaming the company actually setting the prices despite the fact that they have no intention to actually consider an AMD product.
The thing is, it's a duopoly. AMD started the overpricing shenanigans with original Athlon FX and it's a pattern that's been followed on and off since. I guess arguably Asus started it on GPUs with ares/mars? But anyway, back in the days of 'small chip' with the 4870 and 5870 prices were reasonable enough even though amd weren't at all competitive for the high end, more recently with the 600 and 7000 series amd were if anything slightly ahead (in the market, say what you will about unreleased big kepler but that's not the point) but prices climbed noticeably. To be fair there's also an element of silicon just getting more expensive. So, overall, the distortion isn't AMD's fault and AMD having faster products won't magically bring the high end and therefore globals down to $200-300.
-
What superpi scores did it get? Memory timings?
-
if you cant play globals be a hardware master.
The difference between the current hardware #1 and #2 is globals, FYI.
please no talk of globals based on something like ROP count or launch MSRP or brand,then its not global.The same could be said of globals based on CPU cores, surely? I think maybe it's WR points you're thinking of that aren't divided by any aspect of hardware specs.
seems some are just hurt they cant afford a top end cardget a used 980ti you can still do well for about 350usd
if you want a cheaper hobby then blowing is good right?
I'm not sure the "you're just bitter you can't buy a fair chance" attitude is especially helpful. The cheapest 980Ti on ebay.com that doesn't use the reference PCB is $400 but even if it was only $350 to compete that's still a lot for just the cost of entry.
\it guys that only would bench a piece of hardware for points that's what we should get changed .benchmarking is for fun not points when some learn that it will fix itself.
Of course points matter - if they don't why are you defending the status quo, surely you shouldn't care?
A big part of the issue is right now if you're not loaded but like 3D you have to choose between a lot of old hardware (fun) or putting everything into only stuff that gets globals (points). That's not a great tradeoff for people to have to make, and surely if it can be made so you don't have to sacrifice fun to do well on points that's worthwhile?so you have a 480 or 1060 yet?seem like ok cards to play with maybe see points in a few years or so,
No, I don't - see above. My recent GPU purchases are a 4850, 570, 6450, 8800 ultra, 2900 pro, 2600 XT, pair of 4890s, 6800 gt, GTX 275, Radeon 7000, Radeon 9250 and a 9800 GT. Together they add up to the same price as a 4GB RX 480. I could have just got the RX 480 instead of all that, but then I'd have only had one piece of hardware to play with and I'd have been scared of breaking it. This way I have 13 different cards to play with.
but if you want points today from a new card you have to pay.don't get me wrong the best cards are priced way too high but is life.
If there was no better way I'd agree. This thread isn't here to whine and complain, though. This thread is here to try and find a better way. And if we can find one - IMO we have - then the idea that the status quo is 'just the way it is' is the worst reason imaginable not to go for it because if we had the proposed system already it'd be equally applicable against a move to the system you're defending.
If you have any constructive feedback on the issues - whether SPs should be involved, cost vs ROPs, how best to futureproof any grouping system, or for that matter any arguments in favour of the current system that are based on its merits and not its position as the current system - I'd be really interested to hear it
-
-
Cheers
-
Awesome, thanks
-
why is it when a new high end piece of hardware comes out craps on the older scores with ln2 with just water or air people call for change?
whats next? ?
they have hardware points just forget global till a few years and your 480 will have some hardware weight
all the money you spend on vga's and gp's is always short lived what else is new.
its just as bad as giving more points because the high end is more sxpensive.
so ok a get a 1200usd vga I want more points then not less .
see that's stupid to use cost at all,cant afford the big toys then go bowling or go play with older hardware
stop screwing with gpu global points we all know its for high dollar cards always has been
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
-
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?
-
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
-
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?
BUTmore cards could be worth more for longer! Using the same classification criteria! Why is that being ignored? People still game using (example) GTX680 because they are still saving for GTX1060 or whatever comes next. They would still be told that what they have is only good for hardware points.
....and that isn't how real-life PC hardware is used.
I'm not saying the idea is bad, i'm saying that it can do more than the original suggestion
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.
-
LPT: Browse to installation folder and replace the _32.exe with the _64.exe file.
I wish I'd known this before I bought geekbench.
-
Standard top-end cards were £400 for several years. Longer than I expected them to be. From 6800 series to Fermi 5! Calling a £375 card a poor price for global is wrong, IMO. I would be so happy if prices went back to that.
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.
-
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).
-
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. -
I have previously suggested we could categorize into low end, mid end, high end and ultra high end. Might be easier and more "fair" as it reflects real world performance at the time they are released.
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.
Regarding the AMD/NVIDIA split up.The reason why @Strong Island's idea resonates with me is because it's a separation which is more closer to how people segment graphics cards in real life. No one evaluates or buys a card based on the SP or ROP count. Mainly because it's not something that is really comparable across vendors and architectures. It would be like using a metric like pipeline length for processors (or any other metric related to architecture design).
People do separate AMD and Nvidia in their segmentation. I agree it's less 'clean' or 'technical', but in this case I would prefer a more real-world segmentation. Just my opinion though, @Christian Ney has final word on this matter.
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.
-
It's not impossible, but this will create a LOT of new global rankings. I just did a quick check with 3DMark03 and without verifying if all the graphics cards have their SP filled out correctly, there would be ~80 global rankings (instead of the 4 we have now). We'd also have to find a solution to group the older architectures that don't have the ROP/SP layout.
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.
-
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...
-
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.
-
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.
-
TIL: yes, this runs on XP, but not well.
-
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.
-
Here we are
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. -
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
-
stage 2 is hwbotprime and stage 3 is geekbench multi core full out with core division as well which gives another playing field again if you know which cpus can do best at this.
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...
How to fix the rankings :)
in HWBOT Development: bugs, features and suggestions
Posted
Quotes are from K404s earier comment but I ended up with quite a long comment addressing arguments I've seen all over the thread, having seen the post elsewhere about email notifications I've removed the link back to the comment I'm quoting.
I'm not entirely sure what you're saying here - I would assume it's not that you want ranked overclocking to be F1/NBA levels of expensive and inaccessible? If you're thinking people should just put up the current system and work around it maybe you should check the subforum title, it's not like anyone is going on strike here, it's a suggestion that a fair number of people think could improve hwbot.
WR points are the ones that are without conditions. Global points are not without conditions, on the 2D side they're split by cores and it works pretty well (i5s get a bit screwed but as far as I can tell there isn't a good solution to that). Wprime 1024M, for example, has global points for 31 different categories depending on the class of hardware being used as determined by number of CPU cores! How can you possibly look at that and say "nope, global means global, titan should compete against a 710"? It's absurd and I'm honestly flabbergasted that anyone is defending the current situation.
It's not about trying to make a midrange card beat a titan, in fact as I said earlier I think it'd be good to see any changes come alongside an expansion of 3D WR points to keep cards like the Titan XP relevant. It might also be sensible to delay any changes until the next hardware cycle to avoid screwing over anyone who's bought a titan this time around. It's about trying to make 3D globals as inclusive as 2D globals are. The target here is something as close as possible to the system that already exists and is working just fine on the 2D side.
Something I should probably add is, I don't think it's likely I'd benefit personally from any of these changes (at least in terms of points, if they mean we get better oc-focused midrange kit that'd be awesome). As much as I love working with GPUs 15 good subs with 1C and 3C CPUs, maybe 2C as well, seem much more achievable than a bunch of 3D globals. The group I do think it would benefit are new members who are just finding out about competitive overclocking and want to have a go at something with their i5 4460+R9 380 gaming rig. Unlike certain other rather vitriolic individuals (and btw, I have yet to bench XTU at all, I certainly didn't come here through it) I think it'd be great to get more people interested and involved. OC-esports is a big part of that and is really helpful but it's still a problem when however much work someone has put in, however much they've learnt and applied, however well they've modded their card, the hwboints engine just tells them "lol, nice score, here's your two hardware points". For a lot of people that's gonna be pretty demotivating. I happened to already have a huge (by non-ocer standards, it wasn't measured in kg) collection of old hardware for no good reason so for me it was more like "wow, 0.4 points, if I keep this up then with all the hardware points from the hundreds of subs I plan to make I'll be top of rookie league within a couple of months!" (I hadn't figured out how the rankings actually worked at the time, but still). I am the exception, most people with potential won't find themselves in that position.
An expansion of oc-esports could help a lot but I gather it's not easy to set up and run competitions so it'd be nice if there was a solution that didn't involve constant extra pressure on staff who already work extremely hard.
To be honest at the end of the day the current system for 3D globals is clearly indefensible - we can tell this because none of the people who have come here to defend it have given any reason why it's better than the alternatives suggested. There hasn't been a single argument put forward that would be applicable if the question was whether we should move from a ROP-based system to a per-card system (well, ok, one - ROPs are bit technical. But I'm confident a single explanatory page with a table could overcome that). It's all been "that's just the way it is", "points shouldn't matter anyway" (ok, so let's get rid of them?), "yeah but OC-Esports" (oc-esports does soften the blow of an imperfect system but doesn't make it less imperfect) and the like. And a few people have pointed out that there would be imperfect quirks like the Fury X for an ROP-based system and the 980Ti being with 980s in an MSRP-based system, both of which are true and a better alternative would be nice if possible but it's a bit like refusing to stop drowning because the seat in the life raft is a bit damp.