Overclock.net › Forums › Industry News › Hardware News › [Computerbase] - 3gb vs 4gb vs 6gb vs 8gb GDDR5 VRAM Frametime Testing
New Posts  All Forums:Forum Nav:

[Computerbase] - 3gb vs 4gb vs 6gb vs 8gb GDDR5 VRAM Frametime Testing - Page 17

post #161 of 175
Quote:
Originally Posted by EightDee8D View Post

No one does tests like that.

There are numerous tests from that indicate or even prove GCN (Not HBM) struggling more than maxwell and pascal when running low on VRAM. 2GB 380 can sometimes be a stuttery mess while 2GB 960 does much better.
U2
(13 items)
 
U1
(8 items)
 
SFF
(18 photos)
CPUMotherboardGraphicsRAM
i7 3770 Z77N-WIFI GTX 970 Turbo Kingston Valueram VLP 2x4GB  
Hard DriveHard DriveCoolingCooling
Kingston UV400 240GB ADATA XPG SX900 64GB Thermalright Macho Rev. B AC Accelero Mono Plus 
CoolingOSMonitorPower
Thermalright TY-147A Windows 8.1 Enterprise 64-bit Dell P2416D Cooler Master V450S 
Case
Jonsbo U2 
MotherboardRAMHard DriveCooling
MSI N3150I ECO Samsung DDR3 SODIMM 2x2GB Samsung Spinpoint F1 500GB CHECKMATE. IT'S PASSIVE 
OSOSPowerCase
Windows 10 Professional Ubuntu 16.04 LTS Chieftec SFX 250VS Jonsbo U1 
  hide details  
Reply
U2
(13 items)
 
U1
(8 items)
 
SFF
(18 photos)
CPUMotherboardGraphicsRAM
i7 3770 Z77N-WIFI GTX 970 Turbo Kingston Valueram VLP 2x4GB  
Hard DriveHard DriveCoolingCooling
Kingston UV400 240GB ADATA XPG SX900 64GB Thermalright Macho Rev. B AC Accelero Mono Plus 
CoolingOSMonitorPower
Thermalright TY-147A Windows 8.1 Enterprise 64-bit Dell P2416D Cooler Master V450S 
Case
Jonsbo U2 
MotherboardRAMHard DriveCooling
MSI N3150I ECO Samsung DDR3 SODIMM 2x2GB Samsung Spinpoint F1 500GB CHECKMATE. IT'S PASSIVE 
OSOSPowerCase
Windows 10 Professional Ubuntu 16.04 LTS Chieftec SFX 250VS Jonsbo U1 
  hide details  
Reply
post #162 of 175
Quote:
Originally Posted by Smanci View Post

There are numerous tests from that indicate or even prove GCN (Not HBM) struggling more than maxwell and pascal when running low on VRAM. 2GB 380 can sometimes be a stuttery mess while 2GB 960 does much better.
And i'm sure you can also find just opposite of that. and please don't tell me those games are GW infested ones where that happens.

And also don't mix amd's framepacing issues with vram, it's different.
post #163 of 175
Quote:
Originally Posted by EightDee8D View Post

And i'm sure you can also find just opposite of that. and please don't tell me those games are GW infested ones where that happens.

You could, of course, try reading those tests before trying to reinforce your opinion by dodging them? rolleyes.gif
U2
(13 items)
 
U1
(8 items)
 
SFF
(18 photos)
CPUMotherboardGraphicsRAM
i7 3770 Z77N-WIFI GTX 970 Turbo Kingston Valueram VLP 2x4GB  
Hard DriveHard DriveCoolingCooling
Kingston UV400 240GB ADATA XPG SX900 64GB Thermalright Macho Rev. B AC Accelero Mono Plus 
CoolingOSMonitorPower
Thermalright TY-147A Windows 8.1 Enterprise 64-bit Dell P2416D Cooler Master V450S 
Case
Jonsbo U2 
MotherboardRAMHard DriveCooling
MSI N3150I ECO Samsung DDR3 SODIMM 2x2GB Samsung Spinpoint F1 500GB CHECKMATE. IT'S PASSIVE 
OSOSPowerCase
Windows 10 Professional Ubuntu 16.04 LTS Chieftec SFX 250VS Jonsbo U1 
  hide details  
Reply
U2
(13 items)
 
U1
(8 items)
 
SFF
(18 photos)
CPUMotherboardGraphicsRAM
i7 3770 Z77N-WIFI GTX 970 Turbo Kingston Valueram VLP 2x4GB  
Hard DriveHard DriveCoolingCooling
Kingston UV400 240GB ADATA XPG SX900 64GB Thermalright Macho Rev. B AC Accelero Mono Plus 
CoolingOSMonitorPower
Thermalright TY-147A Windows 8.1 Enterprise 64-bit Dell P2416D Cooler Master V450S 
Case
Jonsbo U2 
MotherboardRAMHard DriveCooling
MSI N3150I ECO Samsung DDR3 SODIMM 2x2GB Samsung Spinpoint F1 500GB CHECKMATE. IT'S PASSIVE 
OSOSPowerCase
Windows 10 Professional Ubuntu 16.04 LTS Chieftec SFX 250VS Jonsbo U1 
  hide details  
Reply
post #164 of 175
Quote:
Originally Posted by Smanci View Post

You could, of course.
Thanks wink.gif
post #165 of 175
Quote:
Originally Posted by EightDee8D View Post

No one does tests like that , but i have used both and talking from experience. ( mostly in WD,rom, fc4, som, on 1440p/4k mixed ultra) there are scenarios where 980 stutters but 290x doesn't.

Regardless, we have lots of people with 380x and 960 on this forum right ? so why don't we just test this and end this argument ?

run GTX 960 2GB on default, run same games and settings on GTX 960 4GB with vram frequency lowerd to half. same with 380x. basically same gpu but with different vram config.

there. if 2gb version ( faster vram one ) doesn't improve anything than it means faster vram doesn't solve low capacity problem even 1%. what kind of enthusiast community is this who instead of doing testing on their own, just happy being keyboard warriors ? why don't we just test this simple thing ?

Why are you assuming that people haven't done such tests?

The results of these sorts of tests is precisely where my stance is coming from...learning the why came later. Analyses regarding the amount of VRAM needed and the effects of not having enough go all the way back to the dawn of AGP and Win32 gaming (the first time where not having enough local memory didn't just mean a crash or refusal to run).

I did crap loads of tests in the GeForce 6800 days with things like ShaderMark and Doom 3. Exceeding necessary local VRAM capacity by even a moderate amount and performance dropped precipitously; not a measly half (the most performance you can ever lose by cutting memory frequency in half), but by 80-95%. I've also tested things like the effect of AGP clocks vs. multiplier (interestingly 100MHz AGP 4x is faster than 66MHz AGP 8x on a 6800GT...of course AGP performance really only made much difference at all when you didn't have enough VRAM, and it never made enough difference to make things playable). I repeated such tests again during the 8800 (PCI-E) days and found several cases where my 8800GTS 320 was far less than half as fast as an 8800GTS 640 (Doom 3 ultra again, LotRO ultra high texture quality, etc) at identical clocks (and worse memory timings on the 640).

Every few years, up until very recently, quite a few people were testing the need for various levels of VRAM. However, these tests have been complicated by how cheap memory has become and how texture buffer size (rather flexible in it's use, as modern games and GPUs are demonstrating) has outpaced frame buffer (which, while compressible, absolutely cannot be evicted without annihilating performance) size, which makes it very difficult to precisely control VRAM utilization or account for increasing differences in how memory is managed.

Still, the test you propose would work, provided the game or application used really did make significant use of more than 2GiB of assets at a time, and in this case, it's almost certain that more, but slower (even much slower) VRAM would produce a more playable result.

I've personally seen Elite: Dangerous on both 4GiB and 8GiB 290X (I own three of the former) and the experience at higher IQ settings is appreciably better on the latter (much less hitching when loading planet textures or entering orbital cruise, and much reduced texture pop-in.

Anyway, I'm all for an updated test, though it's hard to find modern titles that don't dynamically adjust LODs to account for insufficient texture buffer issues, and there is no practical way to make the frame buffer not fit in a modern GPU's VRAM. We'd need to compare frame times/hitching, and texture IQ at fixed points, not just average frame rates.
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
post #166 of 175
Quote:
Originally Posted by Blameless View Post

Why are you assuming that people haven't done such tests?

The results of these sorts of tests is precisely where my stance is coming from...learning the why came later. Analyses regarding the amount of VRAM needed and the effects of not having enough go all the way back to the dawn of AGP and Win32 gaming (the first time where not having enough local memory didn't just mean a crash or refusal to run).

I did crap loads of tests in the GeForce 6800 days with things like ShaderMark and Doom 3. Exceeding necessary local VRAM capacity by even a moderate amount and performance dropped precipitously; not a measly half (the most performance you can ever lose by cutting memory frequency in half), but by 80-95%. I've also tested things like the effect of AGP clocks vs. multiplier (interestingly 100MHz AGP 4x is faster than 66MHz AGP 8x on a 6800GT...of course AGP performance really only made much difference at all when you didn't have enough VRAM, and it never made enough difference to make things playable). I repeated such tests again during the 8800 (PCI-E) days and found several cases where my 8800GTS 320 was far less than half as fast as an 8800GTS 640 (Doom 3 ultra again, LotRO ultra high texture quality, etc) at identical clocks (and worse memory timings on the 640).

Every few years, up until very recently, quite a few people were testing the need for various levels of VRAM. However, these tests have been complicated by how cheap memory has become and how texture buffer size (rather flexible in it's use, as modern games and GPUs are demonstrating) has outpaced frame buffer (which, while compressible, absolutely cannot be evicted without annihilating performance) size, which makes it very difficult to precisely control VRAM utilization or account for increasing differences in how memory is managed.

Still, the test you propose would work, provided the game or application used really did make significant use of more than 2GiB of assets at a time, and in this case, it's almost certain that more, but slower (even much slower) VRAM would produce a more playable result.

I've personally seen Elite: Dangerous on both 4GiB and 8GiB 290X (I own three of the former) and the experience at higher IQ settings is appreciably better on the latter (much less hitching when loading planet textures or entering orbital cruise, and much reduced texture pop-in.

Anyway, I'm all for an updated test, though it's hard to find modern titles that don't dynamically adjust LODs to account for insufficient texture buffer issues, and there is no practical way to make the frame buffer not fit in a modern GPU's VRAM. We'd need to compare frame times/hitching, and texture IQ at fixed points, not just average frame rates.

I don't assume nor i claim anything. that's the reason why suggested doing those tests. i have only done limited amount of testing last year ( i didn't know i'll be arguing this one year later here otherwise i could have done extensive tests with videos etc.). and from my experience 290x was smoother than 980, although only in limited scenarios and same in almost all other games. now i don't know if it was just those games or some bug. i want to know this too. but imo faster vram ( like fury'hbm) does help with capacity issues. it doesn't completely fix it. but it does help.
post #167 of 175
Question still stands, have any of the naysayers actually used a card with hbm? Im still gonna say no on that
Yokai Onmyoji
(18 items)
 
  
CPUMotherboardGraphicsRAM
I7-4930k @ 4.5GHz ASUS P9X79 Deluxe  Sapphire RX Vega 64 Liquid  G. Skill Ripjaws Z 16GB (4x4) 2400 Mhz  
Hard DriveHard DriveOptical DriveCooling
Samsung 840 Pro 256GB (OS) Samsung 840 Pro 512GB (Games) Lite-on Corsair H100 
OSMonitorKeyboardPower
Windows 7 Ultimate Wasabi Mango UHD490 IPS 49" Freesync 33-60hz Corsair K90 Corsair Ax1200i 
CaseMouseMouse PadAudio
Corsair 500r Razer Naga Dx Racer Arm wrest  Asus Xonar DX 
AudioOther
Altec Lansing 5.1 Razer Nostromo 
  hide details  
Reply
Yokai Onmyoji
(18 items)
 
  
CPUMotherboardGraphicsRAM
I7-4930k @ 4.5GHz ASUS P9X79 Deluxe  Sapphire RX Vega 64 Liquid  G. Skill Ripjaws Z 16GB (4x4) 2400 Mhz  
Hard DriveHard DriveOptical DriveCooling
Samsung 840 Pro 256GB (OS) Samsung 840 Pro 512GB (Games) Lite-on Corsair H100 
OSMonitorKeyboardPower
Windows 7 Ultimate Wasabi Mango UHD490 IPS 49" Freesync 33-60hz Corsair K90 Corsair Ax1200i 
CaseMouseMouse PadAudio
Corsair 500r Razer Naga Dx Racer Arm wrest  Asus Xonar DX 
AudioOther
Altec Lansing 5.1 Razer Nostromo 
  hide details  
Reply
post #168 of 175
Quote:
Originally Posted by EightDee8D View Post

but imo faster vram ( like fury'hbm) does help with capacity issues. it doesn't completely fix it. but it does help.

It only helps to the extent that faster VRAM helps with performance irrespective of capacity issues. When capacity becomes the primary bottleneck, the same general sort of things (reduced performance and/or texture streaming issues) will happen no matter how fast the memory is; until the primary bottleneck is alleviated, gains are either going to be minimal, or won't address the are of greatest concern (speeding up average frame rate while not reducing hitching, for example).
Quote:
Originally Posted by Tgrove View Post

Question still stands, have any of the naysayers actually used a card with hbm? Im still gonna say no on that

I don't own any cards with HBM, but I've used them (I always try before I buy). Fury still hitches in Elite: Dangerous (the only game I have that really pushes VRAM needs to the point 4GiB cards show issues), more so than an 8GiB 290X, despite average frame rate being higher.

If past tests/precedents, logical deductions based on fundamental principles of how memory works, and AMD's own statements, don't sway you from the fallacy that 4GiB of HBM is miraculously able to hold more than 4GiB of GDDR5 (or GDDR2, or DDR4, or RDRAM, or SRAM, or magnetic core memory, or a billion sticky notes written in crayon for that matter), or that memory performance can influence what is not in memory, nothing will.

The current Fury, Fury Nano, and Fury X parts are very likely to be the only 4GiB HBM parts ever released (certainly with Fiji GPUs), so there is no way to directly compare them to parts with more or less memory on perfectly even footing. So, if only seeing is believing, fire up Elite: Dangerous Horizons with texture resolutions set to 4k fly around for a while then go land on a planet...your Fury Xes will stutter more than slower cards (within reason) with more memory. If you watch VRAM utilization, you'll see it climb gradually until the GPU is forced to start evicting old assets and bring in new ones, at which point you'll hit a major stutter, see VRAM use decline significantly, then start to rise again. 8GiB cards virtually never reach this point, VRAM use generally plateaus and eviction events become far more subtle.
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
Primary
(15 items)
 
Secondary
(13 items)
 
In progress
(10 items)
 
CPUMotherboardGraphicsRAM
5820K @ 4.2/3.5GHz core/uncore, 1.175/1.15v Gigabyte X99 SOC Champion (F22n) Gigabyte AORUS GTX 1080 Ti (F3P) @ 2025/1485, 1... 4x4GiB Crucial @ 2667, 12-12-12-28-T1, 1.34v 
Hard DriveHard DriveHard DriveCooling
Plextor M6e 128GB (fw 1.06) M.2 (PCI-E 2.0 2x) 2x Crucial M4 256GB 4x WD Scorpio Black 500GB Noctua NH-D15 
OSMonitorKeyboardPower
Windows 7 Professional x64 SP1 BenQ BL3200PT Filco Majestouch Tenkeyless (MX Brown) Corsair RM1000x 
CaseMouseAudio
Fractal Design Define R4 Logitech G402 Realtek ALC1150 + M-Audio AV40 
CPUMotherboardGraphicsRAM
X5670 @ 4.4/3.2GHz core/uncore, 1.36 vcore, 1.2... Gigabyte X58A-UD5 r2.0 w/FF3mod10 BIOS Sapphire Fury Nitro OC+ @ 1053/500, 1.225vGPU/1... 2x Samsung MV-3V4G3D/US @ 2000, 10-11-11-30-T1,... 
RAMHard DriveHard DriveHard Drive
1x Crucial BLT4G3D1608ET3LX0 @ 2000, 10-11-11-3... OCZ (Toshiba) Trion 150 120GB Hyundai Sapphire 120GB 3x Hitachi Deskstar 7k1000.C 1TB 
CoolingOSPowerCase
Noctua NH-D14 Windows 7 Pro x64 SP1 Antec TP-750 Fractal Design R5 
Audio
ASUS Xonar DS 
CPUMotherboardGraphicsRAM
i7-6800K @ 4.3/3.5GHz core/uncore, 1.36/1.2v ASRock X99 OC Formula (P3.10) GTX 780 (temporary) 4x4GiB Crucial DDR4-2400 @ 11-13-12-28-T2, 1.33v 
Hard DriveHard DriveCoolingOS
Intel 600p 256GB NVMe 2x HGST Travelstar 7k1000 1TB Corsair H55 (temporary) Windows Server 2016 Datacenter 
PowerCase
Seasonic SS-860XP2 Corsair Carbide Air 540 
  hide details  
Reply
post #169 of 175
Quote:
Originally Posted by Blameless View Post

It only helps to the extent that faster VRAM helps with performance irrespective of capacity issues. When capacity becomes the primary bottleneck, the same general sort of things (reduced performance and/or texture streaming issues) will happen no matter how fast the memory is; until the primary bottleneck is alleviated, gains are either going to be minimal, or won't address the are of greatest concern (speeding up average frame rate while not reducing hitching, for example).
I don't own any cards with HBM, but I've used them (I always try before I buy). Fury still hitches in Elite: Dangerous (the only game I have that really pushes VRAM needs to the point 4GiB cards show issues), more so than an 8GiB 290X, despite average frame rate being higher.

If past tests/precedents, logical deductions based on fundamental principles of how memory works, and AMD's own statements, don't sway you from the fallacy that 4GiB of HBM is miraculously able to hold more than 4GiB of GDDR5 (or GDDR2, or DDR4, or RDRAM, or SRAM, or magnetic core memory, or a billion sticky notes written in crayon for that matter), or that memory performance can influence what is not in memory, nothing will.

The current Fury, Fury Nano, and Fury X parts are very likely to be the only 4GiB HBM parts ever released (certainly with Fiji GPUs), so there is no way to directly compare them to parts with more or less memory on perfectly even footing. So, if only seeing is believing, fire up Elite: Dangerous Horizons with texture resolutions set to 4k fly around for a while then go land on a planet...your Fury Xes will stutter more than slower cards (within reason) with more memory. If you watch VRAM utilization, you'll see it climb gradually until the GPU is forced to start evicting old assets and bring in new ones, at which point you'll hit a major stutter, see VRAM use decline significantly, then start to rise again. 8GiB cards virtually never reach this point, VRAM use generally plateaus and eviction events become far more subtle.

QFT
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 #170 of 175
Quote:
Originally Posted by Blameless View Post

It only helps to the extent that faster VRAM helps with performance irrespective of capacity issues. When capacity becomes the primary bottleneck, the same general sort of things (reduced performance and/or texture streaming issues) will happen no matter how fast the memory is; until the primary bottleneck is alleviated, gains are either going to be minimal, or won't address the are of greatest concern (speeding up average frame rate while not reducing hitching, for example).
I don't own any cards with HBM, but I've used them (I always try before I buy). Fury still hitches in Elite: Dangerous (the only game I have that really pushes VRAM needs to the point 4GiB cards show issues), more so than an 8GiB 290X, despite average frame rate being higher.

If past tests/precedents, logical deductions based on fundamental principles of how memory works, and AMD's own statements, don't sway you from the fallacy that 4GiB of HBM is miraculously able to hold more than 4GiB of GDDR5 (or GDDR2, or DDR4, or RDRAM, or SRAM, or magnetic core memory, or a billion sticky notes written in crayon for that matter), or that memory performance can influence what is not in memory, nothing will.

The current Fury, Fury Nano, and Fury X parts are very likely to be the only 4GiB HBM parts ever released (certainly with Fiji GPUs), so there is no way to directly compare them to parts with more or less memory on perfectly even footing. So, if only seeing is believing, fire up Elite: Dangerous Horizons with texture resolutions set to 4k fly around for a while then go land on a planet...your Fury Xes will stutter more than slower cards (within reason) with more memory. If you watch VRAM utilization, you'll see it climb gradually until the GPU is forced to start evicting old assets and bring in new ones, at which point you'll hit a major stutter, see VRAM use decline significantly, then start to rise again. 8GiB cards virtually never reach this point, VRAM use generally plateaus and eviction events become far more subtle.

I would be glad to try elite dangerous if ever get my hands on it. I used to run sli 980s and there are plenty of scenarios where they tap out on vram where the fury x doesnt. Playing 1 game isnt enough see what the card has to offer.

At a minimum every game i play i disable motion blur, chromatic abberation, film grain, depth of field , and aa (unless sweetfx or smaa). There is also almost always one bugged setting as well that eats a ridiculous amount of performance for what it offers. After turning those settings off i rarely have issues.

After a year of gaming with these cards @ 4k the only games to even come close to making games stutter due to vram was shadows of mordor, far cry primal, and rise of the tomb raider. Even then a switch of a single setting reeled things in.

Just beat doom not long ago with nightmare shadows i dont remember seeing that game cross 3gb

See commentary in this video

https://youtu.be/xKtvDGySoAI
Edited by Tgrove - 9/27/16 at 7:53pm
Yokai Onmyoji
(18 items)
 
  
CPUMotherboardGraphicsRAM
I7-4930k @ 4.5GHz ASUS P9X79 Deluxe  Sapphire RX Vega 64 Liquid  G. Skill Ripjaws Z 16GB (4x4) 2400 Mhz  
Hard DriveHard DriveOptical DriveCooling
Samsung 840 Pro 256GB (OS) Samsung 840 Pro 512GB (Games) Lite-on Corsair H100 
OSMonitorKeyboardPower
Windows 7 Ultimate Wasabi Mango UHD490 IPS 49" Freesync 33-60hz Corsair K90 Corsair Ax1200i 
CaseMouseMouse PadAudio
Corsair 500r Razer Naga Dx Racer Arm wrest  Asus Xonar DX 
AudioOther
Altec Lansing 5.1 Razer Nostromo 
  hide details  
Reply
Yokai Onmyoji
(18 items)
 
  
CPUMotherboardGraphicsRAM
I7-4930k @ 4.5GHz ASUS P9X79 Deluxe  Sapphire RX Vega 64 Liquid  G. Skill Ripjaws Z 16GB (4x4) 2400 Mhz  
Hard DriveHard DriveOptical DriveCooling
Samsung 840 Pro 256GB (OS) Samsung 840 Pro 512GB (Games) Lite-on Corsair H100 
OSMonitorKeyboardPower
Windows 7 Ultimate Wasabi Mango UHD490 IPS 49" Freesync 33-60hz Corsair K90 Corsair Ax1200i 
CaseMouseMouse PadAudio
Corsair 500r Razer Naga Dx Racer Arm wrest  Asus Xonar DX 
AudioOther
Altec Lansing 5.1 Razer Nostromo 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [Computerbase] - 3gb vs 4gb vs 6gb vs 8gb GDDR5 VRAM Frametime Testing