|
SwissArmyDruid posted:Tweaktown rumor: AMD 400 series has taped out. Names are "Ellesmere" and "Baffin", both names of islands well inside the Arctic Circle. Maybe not rebrands but die shrinks?
|
#
?
Oct 23, 2015 15:37
|
|
|
|
| # ? Apr 26, 2024 00:32 |
|
|
I would really like a Hawaii die shrink.
|
|
#
?
Oct 23, 2015 16:24
|
|
|
I'd probably replace my 290 with a pair of die shrunk 390s if they dropped to 175 W each. And if I had the money of course.
|
|
#
?
Oct 23, 2015 16:28
|
|
|
cisco privilege posted:Later models of the 6950 either didn't have the switch or they used it for fan profiles since the dual-BIOS was removed, so that explains why it's not present. You might be able to force-flash the BIOS to the card though to fix it - pretty sure the flashing tool has a force option but it's been awhile.
|
|
#
?
Oct 23, 2015 20:14
|
|
|
numberoneposter posted:It doesn't even post so how would I go about doing this though? I don't have easy access to a dual PCI-E mother board either. Can you get into the bios to select integrated graphics and use that?
|
|
#
?
Oct 23, 2015 20:16
|
|
|
Seamonster posted:Maybe not rebrands but die shrinks? Well.... okay. Let's say that AMD die shrinks the current GCN 1.2 silicon of Fiji and Tonga, and gets rid of everything else. When they bring it to market, what are they going to designate it as? What would it mean for AMD to bring out new 300-series products at the same time as their launching 400-series?
|
|
#
?
Oct 23, 2015 20:35
|
|
|
I don't know because being on the 28nm process node for 4 goddamn years (by the time we can buy 14/16nm parts anyway) and all the rebrands and bullshit has warped everything I assume about product tiering/pricing. Die shrunk Fiji replacing Pitcairn based parts would be quite awesome and they wouldn't even need the 8GB boards.
|
|
#
?
Oct 23, 2015 20:59
|
|
|
Subjunctive posted:Can you get into the bios to select integrated graphics and use that? numberoneposter posted:It doesn't even post so how would I go about doing this though? I don't have easy access to a dual PCI-E mother board either. http://www.overclock.net/t/593427/how-to-unbrick-your-bricked-graphics-card-fix-a-failed-bios-flash (if you're using onboard, change the PC's BIOS to initialize onboard/PCI graphics first instead of PEG/PCI-E graphics, or it won't boot with the bad card installed). future ghost fucked around with this message at 21:24 on Oct 23, 2015 |
|
#
?
Oct 23, 2015 21:22
|
|
|
Seamonster posted:I don't know because being on the 28nm process node for 4 goddamn years (by the time we can buy 14/16nm parts anyway) and all the rebrands and bullshit has warped everything I assume about product tiering/pricing. Yeah, being entirely overclocked/memory-enhanced rebrands kinda hosed up the labelling on the 300 series. My personal bet is that they abandon the 300-series number, and it goes down in history as one of those weird generations like the OEM-only 8000 series. The price premium over the 200-series equivalents will probably collapse over time, because performance didn't jump much when compared to doing the overclocking yourself. Maybe something you'll pay an extra 10% for but not the 30-50% that currently stands. I'm betting we see die-shrinks released as the 400 series, and maybe a new letter in the R3/R5/R9 numbering scheme to distinguish the HBM cards. Like maybe "RF3/RF5/RF9" as HBM moves down the product line, for "Radeon Fury" or something. Paul MaudDib fucked around with this message at 23:48 on Oct 23, 2015 |
|
#
?
Oct 23, 2015 23:44
|
|
|
Uh, these cards aren't coming out anytime soon. What's the problem with them being the next generation?
|
|
#
?
Oct 24, 2015 00:18
|
|
|
The smart thing to do would be to use the X designator, BUT SANELY. X would be new silicon, non-X would be refreshes, like how we THOUGHT it was going to work for the 200-series.
|
|
#
?
Oct 24, 2015 00:21
|
|
|
SwissArmyDruid posted:The smart thing to do would be to use the X designator, BUT SANELY. X would be new silicon, non-X would be refreshes, like how we THOUGHT it was going to work for the 200-series. With the non-X getting replaced by the Xes?
|
|
#
?
Oct 24, 2015 01:07
|
|
|
SwissArmyDruid posted:The smart thing to do would be to use the X designator, BUT SANELY. X would be new silicon, non-X would be refreshes, like how we THOUGHT it was going to work for the 200-series. That could work, maybe just keep the Fury name with the number as a performance designation instead. E.g. Fury X- high end, Fury 9xx - midrange, Fury 8xx - low end or something similar. Make the Nano a low profile/SFF card as part of the 4xx series geared more for smaller builds while keeping good performance. Or, make the Fury line the high end, then go back to an R designation for the mid/low range - like Fury X/Nano - high end, then Radeon R1xxx - midrange, Radeon R9xx low end.
|
|
#
?
Oct 24, 2015 04:00
|
|
|
I'm wondering if Greenland is HBM2, Baffin HBM1.5 and Ellesmere GDDR5. That's a set of Furies, 490/480 and 470/460. 450 and below might not be worth it, but they could just keep cutting up Ellesmere. Highly doubt AMD is going to update the 300 series, it makes no sense. Really the big question now is how AMD will name the new Furies.
|
|
#
?
Oct 24, 2015 06:01
|
|
|
I hope these new designs increase the ROP to shader ratio, the lack of ROPs has clearly been holding AMD's GPUs.
|
|
|
#
?
Oct 24, 2015 06:23
|
|
|
THE DOG HOUSE posted:Nice. Although does this sentence make sense to anybody "These aren't the successors to the Fury range at all, but more the Radeon 300 series that was rebranded earlier this year from the Radeon 200 series. We should expect more news on AMD's new cards in the coming months." Dunno how I missed this earlier. AMD needs the R5/R7 parts to pan out first, so they can start integrating them into the APU dies. FaustianQ posted:I'm wondering if Greenland is HBM2, Baffin HBM1.5 and Ellesmere GDDR5. That's a set of Furies, 490/480 and 470/460. 450 and below might not be worth it, but they could just keep cutting up Ellesmere. Whatever cores they're taping out now, I think is a good chance that will be the ones that will go into APUs. So yeah, if Ellesmere/Baffin are the R7/R5s, there you go, chopped up Ellesmere in APUs. (I'm aware APUs are one single silicon die, I'm just saying.) SwissArmyDruid fucked around with this message at 06:51 on Oct 24, 2015 |
|
#
?
Oct 24, 2015 06:48
|
|
|
I think it'd be really cool if AMD used R8/6/4/2 for APU/mobile with full naming convention, because "R7" or "R5" is pretty nebulous. Strictly using the same naming system would be confusing as well, so odd for dGPU and even for mobile is helpful and makes it look less like a sad attempt to ape Intel. So R8 is up too 12 CU unit with 1gb HBM, R6 is up to 8 CU with 512mb HBM, R4 is 4 CU with 256mb HBM and R2 would be a 2 CU product and no HBM. Something like that R4 would German suplex the poo poo out of current tablet offerings. Actually nothing is really stopping HBM from serving as general system memory so imagine an SOC with 2GB HBM, 4-6 CU and a baytrail class processor competing against even proposed note/Chromebook, tablet name phone offerings for a solid 2 years.
|
|
#
?
Oct 24, 2015 07:47
|
|
|
FaustianQ posted:Actually nothing is really stopping HBM from serving as general system memory so imagine an SOC with 2GB HBM, 4-6 CU and a baytrail class processor competing against even proposed note/Chromebook, tablet name phone offerings for a solid 2 years. So more or less an NVIDIA Jetson TK1 / TX1 upgrade?
|
|
#
?
Oct 24, 2015 08:24
|
|
|
AVeryLargeRadish posted:I hope these new designs increase the ROP to shader ratio, the lack of ROPs has clearly been holding AMD's GPUs.
|
|
#
?
Oct 24, 2015 10:22
|
|
|
Paul MaudDib posted:So more or less an NVIDIA Jetson TK1 / TX1 upgrade? In theory yes, although leaning on "more" due to x86 processor flexibility of Zen and thus more marketable. I can't really guage how potent K12 would end up being and what AMD could leverage from x86 that wouldn't just make K12 run hotter while being babby Zenlite in performance.
|
|
#
?
Oct 24, 2015 15:34
|
|
|
VelociBacon posted:So you're not gaming but want to watch movies in 4K? What processor are you using? I would be surprised if you even needed a discreet video card if you had a modern Intel processor with HD 4600 or newer. HalloKitty posted:It would need to support DisplayPort 1.2 or HDMI 2. He wanted 60Hz. Ok, asked my uncle. He has an Acer M3985 PC with a Core i5 processor (3xxx) using integrated graphics (I believe it's HD 2500). I can't find any really detailed spec sheet listings, but I don't think the HDMI port on the back of his PC will output 4K at 60Hz. Assuming that's correct, what would be the cheapest videocard he could buy to drop in his PC to use a 4K display at 60Hz?
|
|
#
?
Oct 24, 2015 19:41
|
|
|
Gtx950
|
|
#
?
Oct 24, 2015 20:22
|
|
|
Don Lapre posted:Gtx950 Is 750 ti same price?
|
|
#
?
Oct 24, 2015 21:15
|
|
|
750ti is $100-120, 950 is 130-150
|
|
#
?
Oct 24, 2015 21:29
|
|
|
The 950 is better for a 60hz 4K display, it just needs a six-pin connector.
|
|
#
?
Oct 24, 2015 22:07
|
|
|
Ok cool, GTX 950 it is. Thanks thread!
|
|
#
?
Oct 24, 2015 22:09
|
|
|
teagone posted:Ok cool, GTX 950 it is. Thanks thread! The 2GB frame buffer is small, but if it's only driving a 4K display/monitor strictly for video, that's fine.
|
|
#
?
Oct 24, 2015 22:17
|
|
|
Anime Schoolgirl posted:750ti is $100-120, 950 is 130-150 750ti doesn't have HDMI 2.0
|
|
#
?
Oct 24, 2015 22:22
|
|
|
Don Lapre posted:750ti doesn't have HDMI 2.0 Anime Schoolgirl fucked around with this message at 02:53 on Oct 25, 2015 |
|
#
?
Oct 24, 2015 23:46
|
|
|
BIG HEADLINE posted:The 2GB frame buffer is small, but if it's only driving a 4K display/monitor strictly for video, that's fine. Yeah, nothing graphic intensive other than browsing the web and watching video. I think he should be good.
|
|
#
?
Oct 25, 2015 01:59
|
|
|
Yo I heard that the GPU market is hosed because nVidia is a piece of poo poo company that keeps bribing devs to not optimize for AMD cards while also simultaneously dropping support for their old cards (or even actively gimping them with new drivers) to get people to buy new cards that will become obsolete in like 2 years anyway because of the previous business practices How much of that stuff is true?
|
|
#
?
Oct 25, 2015 16:01
|
|
|
So a guy on twitch spent 13 hours building a pc last night and it didn't work
|
|
#
?
Oct 25, 2015 16:25
|
|
|
Minorkos posted:dropping support for their old cards (or even actively gimping them with new drivers) Huh? They are still pushing out new drivers and adding Vulkan/DX12 support for Fermi cards, which are pretty old at this point. Anything further back is too old for new games or workstation applications anyways so a lack of constant updates isn't really a big deal. On the other hand, AMD has basically dropped support for pre-GCN cards. The 5000 and 6000 series cards (from the same time-period as Fermi) have bare-bones support in the current drivers, but there is no support for new APIs or features.
|
|
#
?
Oct 25, 2015 16:26
|
|
|
Don Lapre posted:So a guy on twitch spent 13 hours building a pc last night and it didn't work Some people just aren't meant to play with legos.
|
|
#
?
Oct 25, 2015 16:29
|
|
|
The_Franz posted:Huh? They are still pushing out new drivers and adding Vulkan/DX12 support for Fermi cards, which are pretty old at this point. Anything further back is too old for new games or workstation applications anyways so a lack of constant updates isn't really a big deal. On the other hand, AMD has basically dropped support for pre-GCN cards. The 5000 and 6000 series cards (from the same time-period as Fermi) have bare-bones support in the current drivers, but there is no support for new APIs or features. Yeah I assumed it was all bullshit but it made me worry
|
|
#
?
Oct 25, 2015 16:32
|
|
|
Minorkos posted:Yeah I assumed it was all bullshit but it made me worry The bit about bribing devs to include blobs that are hard to optimise for is basically true though
|
|
#
?
Oct 25, 2015 16:36
|
|
|
In fairness that little pcie slot latch that holds the graphics card probably shouldn't exist... I mean when you actually put in the thumbscrews at the IO plate, the card isn't going anywhere.
|
|
#
?
Oct 25, 2015 16:36
|
|
|
Zero VGS posted:In fairness that little pcie slot latch that holds the graphics card probably shouldn't exist... I mean when you actually put in the thumbscrews at the IO plate, the card isn't going anywhere. I don't know, I'm guessing it stops the back end of a very large card from shifting at all. I agree that it can be something that trips people up, but that's some kind of perverse rite of passage.
|
|
#
?
Oct 25, 2015 16:38
|
|
|
Zero VGS posted:In fairness that little pcie slot latch that holds the graphics card probably shouldn't exist... I mean when you actually put in the thumbscrews at the IO plate, the card isn't going anywhere. Modern cards can definitely be heavy enough to pull out of a slot if not latched
|
|
#
?
Oct 25, 2015 16:42
|
|
|
|
| # ? Apr 26, 2024 00:32 |
|
|
Don Lapre posted:So a guy on twitch spent 13 hours building a pc last night and it didn't work So what even happened. Did he break something or just have some dead parts?
|
|
#
?
Oct 25, 2015 16:45
|
|
