Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016
New Posts  All Forums:Forum Nav:

[TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016 - Page 65

post #641 of 724
Quote:
Originally Posted by Mand12 View Post

Then you should restrict your rants to optimization support, and not include things like FreeSync disinformation and overblown conspiracy theories about 970.

There's a legitimate grievance on optimization support, but it does nobody any good to pollute your message with obvious lies and spin. It just makes it easier for people to dismiss your entire post, because you include extra stuff that is both wrong and irrelevant to your actual point.

What? kookoo.gif ... I am making an observation, and the only one confused appears to be you...
Edited by provost - 3/1/16 at 7:57am
Simplicity
(11 items)
 
Apotheosis
(10 items)
 
 
CPUMotherboardGraphicsRAM
4770k Asus Z87 Pro TBD Corsair Vengeance (2x8GB) DDR3 1600 RAM 
OSMonitorKeyboardPower
Windows 7 Pro Dell U2713HM Alienware TactX gaming Seasonic 850W Gold  
CaseMouse
Cooler Master HAF XB Alienware TactX premium mouse 
  hide details  
Reply
Simplicity
(11 items)
 
Apotheosis
(10 items)
 
 
CPUMotherboardGraphicsRAM
4770k Asus Z87 Pro TBD Corsair Vengeance (2x8GB) DDR3 1600 RAM 
OSMonitorKeyboardPower
Windows 7 Pro Dell U2713HM Alienware TactX gaming Seasonic 850W Gold  
CaseMouse
Cooler Master HAF XB Alienware TactX premium mouse 
  hide details  
Reply
post #642 of 724
Quote:
Originally Posted by Olivon View Post

Man, don't listen too much what AMD say. If not, you'll have big surprises.

During their Fury-X Crossfire review, hardware.fr faced problems related to 4GB limitation in 2 games, Dying Light and Evolve.
They had to lower the details in order to have a correct behaviour again.
http://www.hardware.fr/focus/111/crossfire-radeon-r9-fury-x-fiji-vs-gm200-round-2.html

I mentioned that on the other page.
Quote:
Ps.
In theory, varying on PCIe bus saturation, CrossfireX configurations could suffer however. This would require further testing

That would be due to the PCIe bus saturation of using XDMA for CrossfireX + swapping data in/out of system ram over the PCIe bus. We'd need more recent benchmarks in Crossfire in those games to see if AMD have introduced further optimizations for those titles.

The kicker is that this shouldn't affect the Fury X2 due to the onboard PCIe PLX switch chip which would tackle the XDMA transfers between both GPUs leaving the PCIe bus open for these frame buffer optimizations.
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
post #643 of 724
Quote:
Originally Posted by Mahigan View Post

Ugh, I think you're probably the only person who doesn't get what I'm saying.

Let me break this down for you:

1. Drivers control what goes into the framebuffer and what doesn't.
2. Driver optimisations can be used to store bandwidth sensitive data into the frame buffer and non-sensitive data into system memory.
3. HBM can thus be compared to a sort of embeded cache in the way AMDs drivers treat it.
4. The result is thus a 0 performance penalty from using system memory for non bandwidth sensitive data.
5. You get the full performance benefit of using HBM even if you're buffering more data than its capacity would allow.

Do you understand?

What the...? In my very first post to you, I directly pointed out and mentioned AMD's driver optimizations myself. You're not even comprehending what I'm posting, evidently. For the umpteenth time, what I'm questioning you on is this:
Quote:
HBM is different than GDDR5. Its bandwidth is used like an embeded cache.

What about HBM or its bandwidth makes it different from GDDR5 in a way that makes it a better "embedded cache" or in any way different from GDDR5 in regards to capacity limitations? I've said it a million times already, but I'll answer it for you again. All VRAM is used for the purpose of storing local data that needs to be accessed quickly and all dGPUs, regardless of memory technology, can access system RAM at equal speeds for less immediately-necessary data.

HBM/GDDR5's bandwidth figures refer only to GPU <--> VRAM communication and don't do a thing to assist with VRAM capacity limitations (no particular memory capacity optimizations benefit from HBM's bandwidth) since all the data that must be stored locally will have to be regardless due to PCI-e limitations. A diagram with crudely-written bandwidth figures (512 GB/s for Fiji and HBM, 16 GB/s for PCI-e 3.0):




Data taken from system RAM into VRAM across PCI-e 3.0 is going to be that slow no matter what and therefore the same amount of data for any given scene needs to be stored in VRAM no matter whether the card is using HBM or GDDR5. Additionally, any optimizations involving prioritization and segmentation of data across VRAM and system RAM will be equally as effective regardless of GDDR5 or HBM for the same reason. VRAM bandwidth figures don't mean what some people, including yourself, think they mean. It has no bearing on capacity limitations because those limitations exist due to external factors regardless (like PCI-e).

HBM is not different from GDDR5 in any respect involving effective capacity or optimizations via the separation of data across the two primary memory areas.
Ice-nine
(17 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K Asus Z170-DELUXE Gigabyte G1 GTX 980 Ti Corsair Vengeance LPX DDR4 3000 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 500 GB Seagate Barracuda 7200 RPM 2 TB Asus DVD Drive DRW-24B1ST/BLK/B/AS Phanteks PH-TC14PE 
OSMonitorKeyboardPower
Windows 8.1 Pro AOC Q2770PQU Corsair K95 RGB - Cherry MX Blue EVGA Supernova G2 750W 
CaseMouseMouse PadAudio
NZXT Phantom 410 Gunmetal Edition Logitech G502 Proteus Core Razer Goliathus Speed Edition Small Cyber Acoustics Satellite CA-3602 
Audio
Oppo PM-3 Closed Back Planar Magnetic Headphones 
  hide details  
Reply
Ice-nine
(17 items)
 
  
CPUMotherboardGraphicsRAM
i7-6700K Asus Z170-DELUXE Gigabyte G1 GTX 980 Ti Corsair Vengeance LPX DDR4 3000 
Hard DriveHard DriveOptical DriveCooling
Samsung 850 Evo 500 GB Seagate Barracuda 7200 RPM 2 TB Asus DVD Drive DRW-24B1ST/BLK/B/AS Phanteks PH-TC14PE 
OSMonitorKeyboardPower
Windows 8.1 Pro AOC Q2770PQU Corsair K95 RGB - Cherry MX Blue EVGA Supernova G2 750W 
CaseMouseMouse PadAudio
NZXT Phantom 410 Gunmetal Edition Logitech G502 Proteus Core Razer Goliathus Speed Edition Small Cyber Acoustics Satellite CA-3602 
Audio
Oppo PM-3 Closed Back Planar Magnetic Headphones 
  hide details  
Reply
post #644 of 724
Quote:
Originally Posted by Mand12 View Post

Then you should restrict your rants to optimization support, and not include things like FreeSync disinformation and overblown conspiracy theories about 970.

There's a legitimate grievance on optimization support, but it does nobody any good to pollute your message with obvious lies and spin. It just makes it easier for people to dismiss your entire post, because you include extra stuff that is both wrong and irrelevant to your actual point.

here's a fact.
none of those tech sites test a game using settings which i use. like ultra textures,medium shadows and everything else is off/low. that way i can get 50-60 avg and 30 minimum fps even at 2k.
now with 970. what i know was it has same 4gb 256bit memory just as 980 with less shaders. so i thought it would be better choice. now when i tried running games at those settings it stuttered like hell. why ? because with ultra textures at 2k (fc4/watchdogs) u'll hit that 3.5gb wall. those tech sites didn't show that nor nvidia stated it has weird memory subsystem. so whatever info they gave + my previous experience with 670 i guessed it should work fine. but it never did. and after 1 month or so we knew why. so please don't say it's my fault of buying a gpu based on specification. because that's what you do when tech sites don't use same settings as you do for gaming. so no, there are no lies and spin but an actual experience i had.
post #645 of 724
Quote:
Originally Posted by Serandur View Post

What about HBM or its bandwidth makes it different from GDDR5 in a way that makes it a better "embedded cache"

It's wider and thus offers more bandwidth. So it's a faster "cache".
Quote:
or in any way different from GDDR5 in regards to capacity limitations?
I never made that claim. Where in my quotation do I mention its capacity? Nowhere. What you fail to grasp is that you misunderstood what I said and have created this enormous strawman that you then write 5 paragraphs slaying.
Quote:
HBM is not different from GDDR5 in any respect involving effective capacity or optimizations via the separation of data across the two primary memory areas.

And I never said it was. I said "HBM is different than GDDR5". <--- notice the period? Then I mentioned "Its bandwidth is used like an embeded cache". <--- notice the period.

Is HBM different than GDDR5? Yes, it ia stacked DRAM with many connection points over an interposer. These connection points or "pins" are in greater number than GDDR5 and thus allow for a wider memory interface (4,096 bit). This translates into higher bandwidth. So yes HBM is different than GDDR5.


"Its bandwidth is used like an embeded cache", The 4GB of HBM is used as an embeded cache by AMDs drivers in order to store bandwidth sensitive data. Thus AMD, through driver otimizations, gains the full benefit of HBM even when data spills over its capacity. So yes, AMD drivers use HBM as a cache.

What's so hard to understand????
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
post #646 of 724
Quote:
Originally Posted by Mahigan View Post

It's wider and thus offers more bandwidth. So it's a faster "cache".
I never made that claim. Where in my quotation do I mention its capacity? Nowhere. What you fail to grasp is that you misunderstood what I said and have created this enormous strawman that you then write 5 paragraphs slaying.
And I never said it was. I said "HBM is different than GDDR5". <--- notice the period? Then I mentioned "Its bandwidth is used like an embeded cache". <--- notice the period.

Is HBM different than GDDR5? Yes, it ia stacked DRAM with many connection points over an interposer. These connection points or "pins" are in greater number than GDDR5 and thus allow for a wider memory interface (4,096 bit). This translates into higher bandwidth. So yes HBM is different than GDDR5.


"Its bandwidth is used like an embeded cache", The 4GB of HBM is used as an embeded cache by AMDs drivers in order to store bandwidth sensitive data. Thus AMD, through driver otimizations, gains the full benefit of HBM even when data spills over its capacity. So yes, AMD drivers use HBM as a cache.

What's so hard to understand????

You are still limited by the PCI-e bus no matter how fast the "cache". That's what he is saying.
Super P's rig
(20 items)
 
  
CPUMotherboardGraphicsRAM
5960x ASUS X99-A II Asus GTX 1080 Ti Corsair Vengeance DDR4 3000 
Hard DriveHard DriveHard DriveOptical Drive
MyDigitalSSD BPX NVMe Samsung 850 EVO Seagate Momentus XT 500 GB External DVDRW 
CoolingCoolingOSMonitor
EK-XLC Predator 240 Swiftech 240mm Radiator Windows 10 Samsung 40" 4K - UN40KU6290 
KeyboardPowerCaseMouse
G710+ EVGA SuperNOVA 850G2 Fractal Design Define S G700s 
Mouse PadAudioAudioAudio
Vipamz Extended XXXL Asus U7 M-Audio AV40 Sennheiser HD 439 
  hide details  
Reply
Super P's rig
(20 items)
 
  
CPUMotherboardGraphicsRAM
5960x ASUS X99-A II Asus GTX 1080 Ti Corsair Vengeance DDR4 3000 
Hard DriveHard DriveHard DriveOptical Drive
MyDigitalSSD BPX NVMe Samsung 850 EVO Seagate Momentus XT 500 GB External DVDRW 
CoolingCoolingOSMonitor
EK-XLC Predator 240 Swiftech 240mm Radiator Windows 10 Samsung 40" 4K - UN40KU6290 
KeyboardPowerCaseMouse
G710+ EVGA SuperNOVA 850G2 Fractal Design Define S G700s 
Mouse PadAudioAudioAudio
Vipamz Extended XXXL Asus U7 M-Audio AV40 Sennheiser HD 439 
  hide details  
Reply
post #647 of 724
I read somewhere the wider the bus, 4096bit in HBM1's case, the faster the memory swaps because you got more pages open. Any truth to this?
post #648 of 724
Quote:
Originally Posted by criminal View Post

You are still limited by the PCI-e bus no matter how fast the "cache". That's what he is saying.

Yeah you're not. Because the data kept in system memory is non bandwidth sensitive. Basically information is exchanged between the framebuffer and system memory prior to it being accessed by the GPU. So the information accessed by the GPU is always being accessed in the framebuffer. The driver does this.

That's what he's not understanding. AMD is optimizing on a per game basis. That's why Rise of the Tomb Raider doesn't suffer a performance loss when going from high quality to very high quality textures. Under RotTR the Fiji cards are basically only accessing the very high quality textures needed for a frame from the HBM. So the HBM is acting like an embeded cache.

AMD have been doing this for all the games which have released and use more than 4GB of framebuffer.

Serandur is assuming that the textures are being pulled by the GPU straight from the system memory pool and that's not the case.
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
post #649 of 724
Quote:
Originally Posted by Mahigan View Post

Yeah you're not. Because the data kept in system memory is non bandwidth sensitive. Basically information is exchanged between the framebuffer and system memory prior to it being accessed by a game. So the informatin accessed by the game is always being accessed in the framebuffer.

That's what he's not understanding. AMD is optimizing on a per game basis. That's why Rise of the Tomb Raider doesn't suffer a performance loss when going from high quality to very high quality textures. Under RotTR the Fiji cards are basically only accessing the very high quality textures needed for a frame from the HBM. So the HBM is acting like an embeded cache.

AMD have been doing this for all the games which have released and use more than 4GB of framebuffer.

Serandur is assuming that the textures are being pulled by the GPU straight from the system memory pool and that's not the case.

Both of you are smarter than I am on the matter, so one of you has to be correct. I just don't see how that's possible despite AMD claims. It just doesn't make sense to me and if I was buying a card TODAY, HBM or not 4GB would not be enough.

Edit: wording
Edited by criminal - 3/1/16 at 8:47am
Super P's rig
(20 items)
 
  
CPUMotherboardGraphicsRAM
5960x ASUS X99-A II Asus GTX 1080 Ti Corsair Vengeance DDR4 3000 
Hard DriveHard DriveHard DriveOptical Drive
MyDigitalSSD BPX NVMe Samsung 850 EVO Seagate Momentus XT 500 GB External DVDRW 
CoolingCoolingOSMonitor
EK-XLC Predator 240 Swiftech 240mm Radiator Windows 10 Samsung 40" 4K - UN40KU6290 
KeyboardPowerCaseMouse
G710+ EVGA SuperNOVA 850G2 Fractal Design Define S G700s 
Mouse PadAudioAudioAudio
Vipamz Extended XXXL Asus U7 M-Audio AV40 Sennheiser HD 439 
  hide details  
Reply
Super P's rig
(20 items)
 
  
CPUMotherboardGraphicsRAM
5960x ASUS X99-A II Asus GTX 1080 Ti Corsair Vengeance DDR4 3000 
Hard DriveHard DriveHard DriveOptical Drive
MyDigitalSSD BPX NVMe Samsung 850 EVO Seagate Momentus XT 500 GB External DVDRW 
CoolingCoolingOSMonitor
EK-XLC Predator 240 Swiftech 240mm Radiator Windows 10 Samsung 40" 4K - UN40KU6290 
KeyboardPowerCaseMouse
G710+ EVGA SuperNOVA 850G2 Fractal Design Define S G700s 
Mouse PadAudioAudioAudio
Vipamz Extended XXXL Asus U7 M-Audio AV40 Sennheiser HD 439 
  hide details  
Reply
post #650 of 724
Quote:
Originally Posted by criminal View Post

Both of you are smarter than I am on the matter, so one of you has to be correct. I just don't see how that's possible despite AMD claims. It just doesn't make sense to me and if I was buying a card TODAY, HBM or not 4GB would not be enough.

Edit: wording

It works smile.gif

http://m.hardocp.com/article/2016/02/29/rise_tomb_raider_graphics_features_performance/4#.VtWl9OTS9oM

That's evidence that it works. I've asked him to post evidence to the contrary and he doesn't have any.

Without evidence to the contrary, he can't refute the claim that it works.

A.k.a he's wrong.

End of discussion
Edited by Mahigan - 3/1/16 at 8:53am
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Rumors and Unconfirmed Articles
Overclock.net › Forums › Industry News › Rumors and Unconfirmed Articles › [TT] NVIDIA should launch its next-gen Pascal GPUs with HBM2 in 2H 2016