|
FaustianQ posted:Actually, speaking of future revisions, AMD is tacitly acknowledging that Polaris is a bit hosed up, likely do to process and fab (as if it's a design issue then whoa boy is that a lot of money to dump). I've been reading this thread for about a day and I can already identify Faustianposts without looking, by these links that are formatted to contain an entire awful paragraph.
|
#
?
Jun 30, 2016 23:00
|
|
|
|
| # ? Apr 25, 2024 09:09 |
|
|
THE DOG HOUSE posted:what 970 did you have? So far it looks like they're all about the same size as before except for ones specifically made small, but I can see going from one brand to another being startling. it's a zotac, not sure which one exactly - I got it in exchange for RMAing a 770 (great deal for me, I happily put up with the coil whine). new MSI 1070 overclocked great: +150 core, +700 mem, turning furmark into heat and fps with no issues. The other notable difference - others have mentioned - is the fans stopping completely until the card starts to heat up.
|
|
#
?
Jun 30, 2016 23:06
|
|
|
Messing around a bit more with the dual fan Gigabyte GTX 1070, specifically in regards to coil whine. Much like my old GTX 770, it does exhibit some coil whine at extremely high FPS, we're talking like 1000+ FPS like in the old DX9 3D Mark benchmarks or that screen at the end of Heaven benchmark which is actually a 3D image and runs at like 4k FPS. It however exhibits none in any sort of real scenario where FPS isn't in the multiple hundreds.
|
|
#
?
Jun 30, 2016 23:09
|
|
|
FormatAmerica posted:it's a zotac, not sure which one exactly - I got it in exchange for RMAing a 770 (great deal for me, I happily put up with the coil whine). Yeah twin frozer is imposingly big looking on top of being big. Super tall
|
|
#
?
Jun 30, 2016 23:10
|
|
|
FaustianQ posted:AMD is tacitly acknowledging that Polaris is a bit hosed up nothing in your link suggests that.
|
|
#
?
Jun 30, 2016 23:11
|
|
|
wicka posted:nothing in your link suggests that. There is no reason to plan revisions if you consider your first set of steppings successful.
|
|
#
?
Jun 30, 2016 23:27
|
|
|
FaustianQ posted:There is no reason to plan revisions if you consider your first set of steppings successful. Or they consider them successful but think they can get to where they can reliably do better. That'd fit well with heavily volting early chips to improve parametric yield issues while they get them worked out and the result selling well.
|
|
#
?
Jun 30, 2016 23:36
|
|
|
Do you guys have any idea if Palit is any good? I'm eyeing this 1070, which is going for the "reasonable" price of 480 to 520 euros: http://www.palit.com/palit/vgapro.php?id=2629&lang=en Apparently they're the same company as Gainward, if that helps.
|
|
#
?
Jun 30, 2016 23:41
|
|
|
FaustianQ posted:There is no reason to plan revisions if you consider your first set of steppings successful. agreed, progress is not real
|
|
#
?
Jun 30, 2016 23:49
|
|
|
wicka posted:agreed, progress is not real Don't be a revisionist 
|
|
#
?
Jun 30, 2016 23:55
|
|
|
Lol, if you like AMD you better well hope that wasn't a "best result" of any kind.
|
|
#
?
Jul 1, 2016 00:01
|
|
|
THE DOG HOUSE posted:Lol, if you like AMD you better well hope that wasn't a "best result" of any kind. i don't "like AMD," i like competition. what i don't like are fanboys attempting to poison fairly successful launches. everyone should dislike that unless you love spending more money than you need.
|
|
#
?
Jul 1, 2016 00:06
|
|
|
wicka posted:i don't "like AMD," i like competition. what i don't like are fanboys attempting to poison fairly successful launches. everyone should dislike that unless you love spending more money than you need. Boy are you in the wrong thread. Although plenty of people will jump in saying they are not fanboys and gosh they would rather buy AMD cards if they were better and competition is good for everyone. But drat if they aren't the same people posting RIP and LOL emoticons in reaction to every single thing AMD does.
|
|
#
?
Jul 1, 2016 00:09
|
|
|
xthetenth posted:Or they consider them successful but think they can get to where they can reliably do better. That's what I am reading from the article, 4X5 series cards will be released alongside Vega in 2017 after they get poo poo sorted out at the foundry level. I can get that maybe AMD thinks this is a good stepping but it can be even better, but for 14nmFF it kind of looks like a dumpster fire. wicka posted:agreed, progress is not real So basically drop money forever on new steppings to gently caress the 99% optimization chicken when you're at 95% instead of R&D for a new µarch, because that's progress.
|
|
#
?
Jul 1, 2016 00:10
|
|
|
Honestly it seems to be selling fine on its own. Right now it's a solid product that you should probably wait a little bit for aftermarket versions (see also every single card after the invention of aftermarket coolers) at a great price point, but it's got some characteristics that are a bit suboptimal and are indicators that other products in the stack aren't going to be great. However, there's a lot of variables floating around that could easily mean that by the time those products come out they'll be fine/great. I prefer that to what's been coming out the rest of this generation, that's for sure.
|
|
#
?
Jul 1, 2016 00:11
|
|
|
FaustianQ posted:So basically drop money forever on new steppings to gently caress the 99% optimization chicken when you're at 95% instead of R&D for a new µarch, because that's progress. incremental updates of graphics cards are not a new or controversial thing, faustian, and their existence doesn't indicate any inherent architecture issues. FaustianQ posted:That's what I am reading from the article, 4X5 series cards will be released alongside Vega in 2017 after they get poo poo sorted out at the foundry level. I can get that maybe AMD thinks this is a good stepping but it can be even better, but for 14nmFF it kind of looks like a dumpster fire. the article uses the words "when yields are better." the fact that you must turn that into "when they get their poo poo sorted at the foundry level" speaks volumes about the kind of person you are. wicka fucked around with this message at 00:13 on Jul 1, 2016 |
|
#
?
Jul 1, 2016 00:11
|
|
|
Please don't quote FaustianQ tia
|
|
#
?
Jul 1, 2016 00:14
|
|
|
wicka posted:incremental updates of graphics cards are not a new or controversial thing, faustian, and their existence doesn't indicate any inherent architecture issues. I wasn't arguing there is an architectural issue in Polaris 10 though, rather it's a foundry issue with GloFo that's producing dies with horrible enough quality/variance AMD needs to ramp up voltages to get volume. That requires a new stepping to solve. Maybe my ignorance is working against me here, but are you also lumping moving a design to a smaller process node into this as well? I don't think that counts as a revision, that seems to be more a new generation.
|
|
#
?
Jul 1, 2016 00:17
|
|
|
FaustianQ posted:I wasn't arguing there is an architectural issue in Polaris 10 though, rather it's a foundry issue with GloFo that's producing dies with horrible enough quality/variance AMD needs to ramp up voltages to get volume. That requires a new stepping to solve. Maybe my ignorance is working against me here, but are you also lumping moving a design to a smaller process node into this as well? I don't think that counts as a revision, that seems to be more a new generation. waiting for the process to mature and releasing an incremental update with higher clocks and maybe a better reference cooler to reenergize sales late in the product cycle does not equate to AMD tacitly admitting that polaris is hosed up. that's what i'm saying.
|
|
#
?
Jul 1, 2016 00:19
|
|
|
I want to record some 120 fps video and compare it with 60 fps video on my 144hz display. Maybe try and a/b test. I think I should be able to do that with my iphone. I need more evidence that i'm Eyeneto
|
|
#
?
Jul 1, 2016 00:23
|
|
|
wicka posted:waiting for the process to mature and releasing an incremental update with higher clocks and maybe a better reference cooler to reenergize sales late in the product cycle does not equate to AMD tacitly admitting that polaris is hosed up. that's what i'm saying. My impression was less that the process wasn't allowing higher clocks, rather it's an issue with ASIC quality resulting in needing higher power draw for given performance. At least from my perspective, the only flaw with the RX480 is it's rather high power consumption and very poor performance per watt compared to Nvidia's µarchs so far, and thus they're falling further behind in extracting performance for a given node. But I'll concede and say that may just be me projecting, and that Polaris 10 is drawing the correct amount of power at the correct voltages and the only issue is making 1.5-1.6Ghz clock speed more normal.
|
|
#
?
Jul 1, 2016 00:27
|
|
|
400+ posts since AMD card was released. Can someone give me the tl;dr of where it fits in the current generation of high-end GPUs and it's performance in VR?
|
|
#
?
Jul 1, 2016 00:27
|
|
|
SweetBro posted:400+ posts since AMD card was released. i really like the performance summary charts techpowerup puts together: https://www.techpowerup.com/reviews/AMD/RX_480/24.html
|
|
#
?
Jul 1, 2016 00:30
|
|
|
SweetBro posted:400+ posts since AMD card was released. TL;DR - It's basically a GTX 970 or R9 390 at stock, with little overclocking headroom due to a poor cooler and use of a 6-pin connector. It's an entry level VR experience.
|
|
#
?
Jul 1, 2016 00:31
|
|
|
That's unfortunate to hear. I guess I'll return to refreshing the Newegg page hoping a 1080 gets in stock.
|
|
#
?
Jul 1, 2016 00:34
|
|
|
Yeah then it isn't even close to targeting the segment you're interested in.
|
|
#
?
Jul 1, 2016 00:36
|
|
|
SweetBro posted:Can someone give me the tl;dr of where it fits in the current generation of high-end GPUs and it's performance in VR? VR: meets current minimums (ie 970), but maybe won't stand up long term if nvidia's new tech that renders binocular views for VR becomes the standard Power & Heat: 150 watts, terrible blower cooler on first wave reference cards, issue with drawing too much power from the pcie slot especially if overclocked. wait for models with good heatsinks and possibly 8-pin power connectors. edit: lol $200 card or $700 card, what a weird comparison Klyith fucked around with this message at 00:40 on Jul 1, 2016 |
|
#
?
Jul 1, 2016 00:38
|
|
|
fozzy fosbourne posted:I want to record some 120 fps video and compare it with 60 fps video on my 144hz display. Maybe try and a/b test. I think I should be able to do that with my iphone. I need more evidence that i'm Eyeneto everyone I have ever tested can tell the difference between 120 and 60, they just never bothered to find out if they could. They cant tell that its double, but its noticeable smoother. Just go to a best buy and compare sports footage on the 120hz tvs compared to the 60hz. I dont know what they broadcast but most of the cameras used record in 120 any higher than that is when people start having trouble seeing much improvement Fauxtool fucked around with this message at 00:48 on Jul 1, 2016 |
|
#
?
Jul 1, 2016 00:38
|
|
|
wicka posted:i don't "like AMD," i like competition. what i don't like are fanboys attempting to poison fairly successful launches. everyone should dislike that unless you love spending more money than you need. So if a company releases a hot turd for a good price, nobody is allowed to say anything at all, ever, otherwise now you're a fanboy? No thanks. There was some  due to disappointment, but that'll pass. Flat out ignoring and denying stuff IS the problem, which is what you've been doing. Who have been the people generally insulting people who buy the other brand's cards? Who are the people generally insulting the company itself? AMD has survived 3 years without making a drat thing until two days ago and priced it low enough to be attractive. But nobody is allowed to hate on them for this, even a little? It's not all sunshine and rainbows in AMD land and you'd have to blind to not see that, and its getting kind of obnoxious to not be able to talk about it without having to think "hmm how do I be neutral about this [bad thing] without someone flipping out and scoffing how'd anybody buy a nvidia card?". Yes, I'm trying to poison a video card launch  . Buy whatever you want, trust me thats all anybody actually says here at the end of the day. A 480 is a $200 reference 970, and thats pretty fine, and its definitely the price everybody likes. If you don't like somebody's argument against one aspect of the tech behind or another, if its wrong or out of perspective or whatever, tell them. But this whole "you cant NOT like it or youre terrible" attitude is growing plain as day. I am not impressed with the card for all the clear reasons that have been repeated over and over again. But that doesn't matter, because its $200. The press is okay (hint: if you want to use "read the reviews" as an argument, careful there, 99% of gpu reviews are positive), it will sell well and I'm sure it already has. And to rephrase what I said, if you like competition, then you better hope this isn't their best effort. Guess what? Most of us like competition, because thats the only reason we have any of this poo poo. In the interest of competition and progress, this card is not comforting at all.
|
|
#
?
Jul 1, 2016 00:39
|
|
|
SweetBro posted:400+ posts since AMD card was released. It doesn't fit in the current generation of " high end" GPUs at all, as it is very solidly mid-range. It also appears effectively impossible to overclock due to already being at the edge of its power envelope. It is probably comparable to a GTX970 in VR (give or take vagaries like DX12 and each vendors VR secret sauce features) for $200, which is to say "adequate for current gen products". I'm not sure I understand the appeal for VR, though, since that is pretty much by definition not the type of audience who would otherwise be shopping for a $200 GPU. If you are already paying $600-800 for a VR headset, why spend $250 on something "adequate" for current gen when you can something with substantially more headroom (possibly enough for next gen headsets as well) for a small marginal cost increase. Anyways, definitely don't buy a reference RX480 as it has established power issues and the cooler is both loud and bad, which aftermarket cards should greatly help. Of course by the time those are available there might be 1060s in the market...
|
|
#
?
Jul 1, 2016 00:39
|
|
|
THE DOG HOUSE posted:So if a company releases a hot turd for a good price, nobody is allowed to say anything at all, ever, otherwise now you're a fanboy? i definitely read this
|
|
#
?
Jul 1, 2016 00:40
|
|
|
SweetBro posted:400+ posts since AMD card was released. It is performance in VR? Syntactically confusing. Anyway, the 480 is alright, but it performs more like a 970, instead of the 980 that most people had hoped for. Some say it draws too much power from the PCIe slot, but overall power usage isn't that high; that being said, it's not very spectacular for a process shrink (it draws similar power to a 970). The 4GB reference version has slower RAM than the 8GB reference version. Essentially, it's around the performance of a Geforce 970, gaining in the edge in newer/DX12 titles, with more VRAM, for less money. Edit: oh, and since it essentially matches a 970, that means, that yes, a 480 is good enough for a reasonable (read: baseline) VR experience. HalloKitty fucked around with this message at 00:50 on Jul 1, 2016 |
|
#
?
Jul 1, 2016 00:44
|
|
|
HalloKitty posted:The 4GB reference version has slower RAM than the 8GB reference version. Essentially, it's around the performance of a Geforce 970, gaining in the edge in newer/DX12 titles, with more VRAM, for less money. This isn't set in stone apparently, or maybe it's mislabeling in stores.
|
|
#
?
Jul 1, 2016 00:49
|
|
|
My only issue with the 480 is how tantalizingly close it is to being right while still missing the mark. All it needs is an AIB with an 8 pin PCIE, decent cooler and 4 gb of 8 Ghz ram. Not exactly a tall order at $225. If I could pair that with the salvaged 4670k whose mobo just crapped out on me, I'd be set at 1080p for an HTPC. Here's hoping the non ref boards come in August. E: wicka posted:i definitely read this ZobarStyl fucked around with this message at 00:54 on Jul 1, 2016 |
|
#
?
Jul 1, 2016 00:50
|
|
|
Lots of whining about the 480 on SA, but it's OOS everywhere now and price bumped. AMD seems to have sold out of their initial offering.
|
|
#
?
Jul 1, 2016 00:51
|
|
|
ZobarStyl posted:My only issue with the 480 is how tantalizingly close it is to being right while still missing the mark. All it needs is an AIB with an 8 pin PCIE, decent cooler and 4 gb of 8 Ghz ram. Not exactly a tall order at $225. If I could pair that with the salvaged 4670k whose mobo just crapped out on me, I'd be set at 1080p for an HTPC. you may get them earlier than that, powercolor's devil contest ends july 11th, i think, and that's when people think we'll get the non-reference boards.
|
|
#
?
Jul 1, 2016 00:58
|
|
|
jm20 posted:Lots of whining about the 480 on SA, but it's OOS everywhere now and price bumped. AMD seems to have sold out of their initial offering. Paper launch 
|
|
#
?
Jul 1, 2016 01:07
|
|
|
FaustianQ posted:This isn't set in stone apparently, or maybe it's mislabeling in stores. AIBs will be free to use 8 GHz memory, to my understanding.
|
|
#
?
Jul 1, 2016 01:09
|
|
|
Can anyone confirm the current Gigabyte 1070/1080 Windforce designs with 3 fans will run at 0 rpm if the card is cool enough? This was a hugely advertised feature on previous gens, but I noticed it's suspiciously absent on their marketing page. Also, I noticed that the backplate on these cards can get extremely hot. Guru3D's review has it reaching 60C over a very large surface area. Would adding some heatsinks to the backplate using thermal glue help core temps at all? I have so many old HSFs laying around I could dremel up as an experiment. Chuu fucked around with this message at 01:19 on Jul 1, 2016 |
|
#
?
Jul 1, 2016 01:15
|
|
|
|
| # ? Apr 25, 2024 09:09 |
|
|
Paul MaudDib posted:AIBs will be free to use 8 GHz memory, to my understanding. Yeah, they are, but I'm not surprised that it's pretty rare.
|
|
#
?
Jul 1, 2016 01:17
|
|