Overclock.net › Forums › Industry News › Hardware News › [Videocardz]AMD Radeon R9 Nano confirmed to feature 4096 Stream Cores (up to 1000MHz)
New Posts  All Forums:Forum Nav:

[Videocardz]AMD Radeon R9 Nano confirmed to feature 4096 Stream Cores (up to 1000MHz) - Page 25

post #241 of 273
Quote:
Originally Posted by Colonel Gerdauf View Post

Just a few notes then:

1) Consumers do not buy products based on the engineering marvel, that is the job of the enterprise. Consumers value products based its practicality for the build they are making, which is usually standard ATX. Having a small card attempt to shoot for the big leagues, as shown by the Fury X launch, ends up being a catch 22 with its perceived power. You can bring in the discussion about the in-depth prosumers, but the sad cold truth is that they are statistically irrelevant.

2) As impressive as HBM is, and the change of VRAM landscape being nothing short of beneficial, what did it do for the consumers in the meantime? VRAM speed was not a bottleneck to begin with; it was never the weakest link in the chain. AMD is very much beta-testing the HBM tech, a divisive decision to make, with the technical limitations that HBM-1 had. Now NVidia and AMD have concrete plans for HBM-2 with Pascal and Greenland, respectively. One of the key tangible benefits this has over GDDR5, is that it can effortlessly beat the reigning Titan X in regards to the VRAM capacity, going for 32GB as the new maximum volume.

1) Consumers don't buy products based on engineering, and i don't believe i said they did. At least i never intended to. The engineering marvel benefit comes in the fact that AMD has it on the market now, while NV is yet to prove they can in any meaningful way. As for perceived power, AFAIK, reviewers are saying that it went down, no? So who is saying it went up then, online marketing trolls? Not exactly AMD's fault. I'd rather they didn't engage in an online forum trollfest with NV's paid marketing shills.

2) What it did for the benefit of consumers is enable products like Nano/Fury and AMD never had to make major changes to the core to get the perf/watt (efficiency lead). I'm not sure what technical limitations you are talking about? Especially if as you say, VRAM speed wasn't a bottleneck to begin with. So that rules out bandwidth. How about VRAM size? Nope again, as reviews have showed that 4GB is not a limiting factor either. What other limitations are there with HBM1? I can't think of any but there seem to be many advantages such as size and power consumption. Again, HBM enables Nano to be the worlds most efficient card, which is the number one consideration now.

The R&D for HBM by AMD is completed, we've yet to see if NV can accomplish the same. The only divisive thing about it is AMD have proven products on the market and are well on the way to refine it to HBM2, and NV has nothing proving they can accomplish the same. And BTW, Titan X is not leading anything in regards to VRAM capacity. That belongs to the FirePro 9100 with 16GB VRAM, so AMD has the ability to bring it to the consumer market as well. It is however completely unessesary for consumers as is 12GB, that's just a marketing stunt to suck in gullible fanboys.
post #242 of 273
Thread Starter 
lol Nvidia won`t have any issues getting HBM to work with Pascal tongue.gif

Hint 1 AMD didnt work 7 years on building Fiji, but 7 years co-developing HBM. Once Hynix got that done, it is now open for anyone to sample HBM and use it with their own products.
Getting HBM to work on GPUs take way way shorter time, and Nvidia have probably been working on Pascal for many years now

Hint 2: Nvidia have way bigger R&D for engineering GPUs, and their driver team is much bigger. Do the math

Hint 3: Nvidia may have played the smart game, they let someone else order HBM and allow yields to increase. They didnt lose anything doing this, the 980Ti is still the fastest card and it comes with GDDR5. And HBM may have contributed to higher price for all of AMDs Fiji cards.
If anything, it is AMD that risked it all with this move
post #243 of 273
Just want to point out that more money for R&D does not automatically mean the product will be sucessful. Last time Nvidia moved to a new node, arch and memory arch at once, they faceplanted hard (GTX 480 being massively delayed, yields in the low single digits, initially clocked half as fast as expected, with much higher power-consumption). That being said, they've probably learned their lesson.
post #244 of 273
Quote:
Originally Posted by iLeakStuff View Post

lol Nvidia won`t have any issues getting HBM to work with Pascal tongue.gif

Hint 1 AMD didnt work 7 years on building Fiji, but 7 years co-developing HBM. Once Hynix got that done, it is now open for anyone to sample HBM and use it with their own products.
Getting HBM to work on GPUs take way way shorter time, and Nvidia have probably been working on Pascal for many years now

Hint 2: Nvidia have way bigger R&D for engineering GPUs, and their driver team is much bigger. Do the math

Hint 3: Nvidia may have played the smart game, they let someone else order HBM and allow yields to increase. They didnt lose anything doing this, the 980Ti is still the fastest card and it comes with GDDR5. And HBM may have contributed to higher price for all of AMDs Fiji cards.
If anything, it is AMD that risked it all with this move

Nvidia's plan isn't flawless in regards to HBM. AMD had exclusivity with HBM1 and the availability (or more precisely LACK of availability) of HBM2 next year virtually guarantees that AMD will actually get a bit of a head start on the next generation, at least IMO. I still believe that Hynix is likely to give AMD first dibs on the low supply of HBM2 and Nvidia will likely end up waiting on Samsung anyway which puts big Pascal well into 2017. If AMD uncharacteristically shows even an inkling of competence with their marketing and design of their flagship sku next year they could conceivably beat Nvidia to the punch with a card that destroys the 980Ti and leaves Nvidia to twist in the wind for half year or more before they can even respond with a flagship of their own. I know, slim chance there but it is conceivable...
post #245 of 273
Quote:
Originally Posted by Majin SSJ Eric View Post

Nvidia's plan isn't flawless in regards to HBM. AMD had exclusivity with HBM1 and the availability (or more precisely LACK of availability) of HBM2 next year virtually guarantees that AMD will actually get a bit of a head start on the next generation, at least IMO. I still believe that Hynix is likely to give AMD first dibs on the low supply of HBM2 and Nvidia will likely end up waiting on Samsung anyway which puts big Pascal well into 2017. If AMD uncharacteristically shows even an inkling of competence with their marketing and design of their flagship sku next year they could conceivably beat Nvidia to the punch with a card that destroys the 980Ti and leaves Nvidia to twist in the wind for half year or more before they can even respond with a flagship of their own. I know, slim chance there but it is conceivable...

The problem with this is that AMD hasn't been coordinated in respect to marketing and product placement for a long time now. AMD barely even has a foothold on gaming as we speak! Meanwhile, NVidia is primarily focused on the tough battle against Intel in the enterprise space.

This is my personal bias speaking, but every time I see AMD's marketing materials, it screams to me "L33T HAx0RZ MLG 3SWAD5LYFE".
My PC
(14 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i5 6600K Gigabyte Z170MX-Gaming 5  Gigabyte GeForce GTX 1070 - G1 Gaming Corsair Vengeance LPX 
Hard DriveHard DriveCoolingOS
Samsung 850 Evo - 2.5" Seagate Desktop Swiftech H240-X Windows 10 Pro 
MonitorKeyboardPowerCase
Samsung S22D390H Gamdias HERMES Ultimate Seasonic 660W 80-Plus Platinum xp2 Corsair 350D 
MouseAudio
Logitech G502 Proteus Spectrum Beyerdynamic Custom One Pro Plus 
  hide details  
Reply
My PC
(14 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i5 6600K Gigabyte Z170MX-Gaming 5  Gigabyte GeForce GTX 1070 - G1 Gaming Corsair Vengeance LPX 
Hard DriveHard DriveCoolingOS
Samsung 850 Evo - 2.5" Seagate Desktop Swiftech H240-X Windows 10 Pro 
MonitorKeyboardPowerCase
Samsung S22D390H Gamdias HERMES Ultimate Seasonic 660W 80-Plus Platinum xp2 Corsair 350D 
MouseAudio
Logitech G502 Proteus Spectrum Beyerdynamic Custom One Pro Plus 
  hide details  
Reply
post #246 of 273
Quote:
Originally Posted by Colonel Gerdauf View Post

The problem with this is that AMD hasn't been coordinated in respect to marketing and product placement for a long time now. AMD barely even has a foothold on gaming as we speak! Meanwhile, NVidia is primarily focused on the tough battle against Intel in the enterprise space.

This is my personal bias speaking, but every time I see AMD's marketing materials, it screams to me "L33T HAx0RZ MLG 3SWAD5LYFE".

Lol it does seem a bit ridiculous at times and is almost always over exaggerated. The only time I remember their claims being true in like the last 5 years was with Piledriver performance increase compared to Bulldozer.. which wasn't much.

Ever since Bulldozer it seems to have been pretty downhill.

R9 nano AMD slides say faster than 290X. Watch it not even be able to match the 290X because of like throttling/issues meeting power target or something like that.
Celerity
(20 items)
 
  
CPUMotherboardGraphicsRAM
Intel i7-6850K @ 4.5GHz / 1.41v Asus X99 Deluxe II EVGA GTX 980 Ti Classified  Corsair Vengeance LPX 4x8GB 3000MHz 
Hard DriveHard DriveCoolingCooling
Samsung 850 Pro 256 GB Western Digital Purple 2TB XSPC Raystorm Pro XSPC RX 360 
CoolingOSMonitorMonitor
XSPC D5 Photon 170 Windows 7 Professional BenQ GW2765 ViewSonic XG2401 
KeyboardPowerCaseMouse
Corsair Strafe RGB MX Silent EVGA SuperNOVA G2 750W Fractal Design Define S Logitech G900 
Mouse PadAudioAudioOther
SteelSeries QCK+ Sennheiser HD 598 FiiO E10K Cablemod Sleeved Cables (Black/White) 
  hide details  
Reply
Celerity
(20 items)
 
  
CPUMotherboardGraphicsRAM
Intel i7-6850K @ 4.5GHz / 1.41v Asus X99 Deluxe II EVGA GTX 980 Ti Classified  Corsair Vengeance LPX 4x8GB 3000MHz 
Hard DriveHard DriveCoolingCooling
Samsung 850 Pro 256 GB Western Digital Purple 2TB XSPC Raystorm Pro XSPC RX 360 
CoolingOSMonitorMonitor
XSPC D5 Photon 170 Windows 7 Professional BenQ GW2765 ViewSonic XG2401 
KeyboardPowerCaseMouse
Corsair Strafe RGB MX Silent EVGA SuperNOVA G2 750W Fractal Design Define S Logitech G900 
Mouse PadAudioAudioOther
SteelSeries QCK+ Sennheiser HD 598 FiiO E10K Cablemod Sleeved Cables (Black/White) 
  hide details  
Reply
post #247 of 273
Thread Starter 
Quote:
Originally Posted by Majin SSJ Eric View Post

Nvidia's plan isn't flawless in regards to HBM. AMD had exclusivity with HBM1 and the availability (or more precisely LACK of availability) of HBM2 next year virtually guarantees that AMD will actually get a bit of a head start on the next generation, at least IMO. I still believe that Hynix is likely to give AMD first dibs on the low supply of HBM2 and Nvidia will likely end up waiting on Samsung anyway which puts big Pascal well into 2017. If AMD uncharacteristically shows even an inkling of competence with their marketing and design of their flagship sku next year they could conceivably beat Nvidia to the punch with a card that destroys the 980Ti and leaves Nvidia to twist in the wind for half year or more before they can even respond with a flagship of their own. I know, slim chance there but it is conceivable...

Yes, it is entirely possible that AMD indeed beat Nvidia to the punch with the upcoming new 16nm new architecture cards. AMD was first out with GCN back in 2012 with 7970 while Nvidia launched GTX 680 2 months after.
I have to question the whole HBM2 exclusive rumors though. I don`t think Hynix want to make a bad relationship with Nvidia and say "sorry, AMD get first". Nvidia have the marketshare to their advantage. Nvidia is like Apple while AMD is like Altera. Both make SOC`s, but in the eyes of a manufacturer, it would for sure cater for Apple and make sure they get their chips before allocating chips to Altera. Both are important for the manufacturer, but in number of silicon ordered, the value of getting a good relationship with Apple is extremely important.
Right now we are at 80% for Nvidia and 20% for AMD. For every HBM GPU AMD sell, Nvidia could sell 4 HBM GPUs.
Back in 2012 AMD had a 40% market share while Nvidia was at 60%. Now we are at 80/20.

We shall see. You might be right, you might be wrong. I`d say the recent rumors about GP100 already being manufactured, the Pascal HBM prototype from Nvidia they showed during Maxwell presentation, the news about transistor count for the upcoming Pascal chip, atleast give certain hints that Nvidia is getting closer and closer to a launch.

Whats also interesting imo is that we have no dual Maxwell GPU on the market. Fermi had one, Kepler had one. Are they skipping this because they are closer to a GP100 launch that will match the performance of such a chip?...
Edited by iLeakStuff - 8/30/15 at 1:47am
post #248 of 273
Quote:
Originally Posted by iLeakStuff View Post

I have to question the whole HBM2 exclusive rumors though. I don`t think Hynix want to make a bad relationship with Nvidia and say "sorry, AMD get first". Nvidia have the marketshare to their advantage. Nvidia is like Apple while AMD is like Altera. Both make SOC`s, but in the eyes of a manufacturer, it would for sure cater for Apple and make sure they get their chips before allocating chips to Altera. Both are important for the manufacturer, but in number of silicon ordered, the value of getting a good relationship with Apple is extremely important.
Right now we are at 80% for Nvidia and 20% for AMD. For every HBM GPU AMD sell, Nvidia could sell 4 HBM GPUs.
Back in 2012 AMD had a 40% market share while Nvidia was at 60%. Now we are at 80/20.

AMD co developed HBM with hynix though, plus it will be in much lower production numbers than DDR memory. If AMD promises to buy everything they can produce then why wouldn't they sell to AMD only?

Samsung is mass producing HBM in the first half of 2016 so I'm sure nvidia will be using them as their source. They are new to working with HBM as is nvidia.

http://www.fudzilla.com/news/graphics/38555-samsung-to-make-hbm-2-memory-in-2016

These reasons are why AMD have a head start when it comes to actually implementing HBM into products.
Edited by Casey Ryback - 8/30/15 at 2:00am
Another 3570K
(14 items)
 
  
CPUMotherboardGraphicsRAM
i5 3570K @ 4.6ghz Gigabyte Z77X-D3H  GB 7970 @ 1100/1500 8GB Corsair V LP 1600mhz  
Hard DriveCoolingMonitorPower
Intel 330 series 240GB Noctua NH-D14 1080p 24" + 27" Antec HCG 620W 
CaseAudioAudioAudio
HAF XM Yamaha RX-V373 Accusound 8.8XD 180W Signature 8" 3-way  Auzentech meridian 7.1 2G 
Audio
Audio Technica AD700X 
  hide details  
Reply
Another 3570K
(14 items)
 
  
CPUMotherboardGraphicsRAM
i5 3570K @ 4.6ghz Gigabyte Z77X-D3H  GB 7970 @ 1100/1500 8GB Corsair V LP 1600mhz  
Hard DriveCoolingMonitorPower
Intel 330 series 240GB Noctua NH-D14 1080p 24" + 27" Antec HCG 620W 
CaseAudioAudioAudio
HAF XM Yamaha RX-V373 Accusound 8.8XD 180W Signature 8" 3-way  Auzentech meridian 7.1 2G 
Audio
Audio Technica AD700X 
  hide details  
Reply
post #249 of 273
Out of topic question : when the arctic island gpus are supposed to launch ???
Number 3
(16 items)
 
Dust Catcher
(9 items)
 
CPUMotherboardGraphicsRAM
i7 6700k 4.5 ghz MSI Gaming M7  1080ti aorus xtreme 2x8 go Gskill 3200mhz C15 
Hard DriveHard DriveHard DriveHard Drive
crucial ssd 128 Go 2x Samsung 840 evo 250 Go Hitachi HDD 2To Toshiba HDD 2To 
Optical DriveCoolingOSMonitor
Old DVD-RW KRAKEN X61 Windows 10 x64 Swift 2k 165hz 
KeyboardPowerCaseMouse
g910 CM V1000 CM Trooper modded g502 proteus 
CPUMotherboardGraphicsRAM
Phenom ii x6 1075t Asus Sabertooth 990fx CF 2x Asus 6950 directcu ii  2x4Gb Gskill 1333 
Hard DriveCoolingOSPower
80go Intel x25 ssd series + 500 go Samsung  cooler master V6 gt Win 7 64bit ultimate edition cooler master silent pro 850w 
Case
HAF 912 advanced 
  hide details  
Reply
Number 3
(16 items)
 
Dust Catcher
(9 items)
 
CPUMotherboardGraphicsRAM
i7 6700k 4.5 ghz MSI Gaming M7  1080ti aorus xtreme 2x8 go Gskill 3200mhz C15 
Hard DriveHard DriveHard DriveHard Drive
crucial ssd 128 Go 2x Samsung 840 evo 250 Go Hitachi HDD 2To Toshiba HDD 2To 
Optical DriveCoolingOSMonitor
Old DVD-RW KRAKEN X61 Windows 10 x64 Swift 2k 165hz 
KeyboardPowerCaseMouse
g910 CM V1000 CM Trooper modded g502 proteus 
CPUMotherboardGraphicsRAM
Phenom ii x6 1075t Asus Sabertooth 990fx CF 2x Asus 6950 directcu ii  2x4Gb Gskill 1333 
Hard DriveCoolingOSPower
80go Intel x25 ssd series + 500 go Samsung  cooler master V6 gt Win 7 64bit ultimate edition cooler master silent pro 850w 
Case
HAF 912 advanced 
  hide details  
Reply
post #250 of 273
Quote:
Originally Posted by Noufel View Post

Out of topic question : when the arctic island gpus are supposed to launch ???

Those details haven't been released yet.
Another 3570K
(14 items)
 
  
CPUMotherboardGraphicsRAM
i5 3570K @ 4.6ghz Gigabyte Z77X-D3H  GB 7970 @ 1100/1500 8GB Corsair V LP 1600mhz  
Hard DriveCoolingMonitorPower
Intel 330 series 240GB Noctua NH-D14 1080p 24" + 27" Antec HCG 620W 
CaseAudioAudioAudio
HAF XM Yamaha RX-V373 Accusound 8.8XD 180W Signature 8" 3-way  Auzentech meridian 7.1 2G 
Audio
Audio Technica AD700X 
  hide details  
Reply
Another 3570K
(14 items)
 
  
CPUMotherboardGraphicsRAM
i5 3570K @ 4.6ghz Gigabyte Z77X-D3H  GB 7970 @ 1100/1500 8GB Corsair V LP 1600mhz  
Hard DriveCoolingMonitorPower
Intel 330 series 240GB Noctua NH-D14 1080p 24" + 27" Antec HCG 620W 
CaseAudioAudioAudio
HAF XM Yamaha RX-V373 Accusound 8.8XD 180W Signature 8" 3-way  Auzentech meridian 7.1 2G 
Audio
Audio Technica AD700X 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [Videocardz]AMD Radeon R9 Nano confirmed to feature 4096 Stream Cores (up to 1000MHz)