Go Back   Rage3D » Rage3D Discussion Area » Graphics Technology Forums » AMD Radeon Discussion and Support
Rage3D Subscribe Register FAQ Members List Calendar Mark Forums Read

AMD Radeon Discussion and Support General discussion, tweaking, overclocking and technical support questions about discrete Radeon graphics products.

Reply
 
Thread Tools Display Modes
Old Dec 30, 2017, 12:26 PM   #31
Advertisement (Guests Only)

Login or Register to remove this ad
Treeckcold57
Good ol' ATI
 
Join Date: Sep 2004
Posts: 17,021
Treeckcold57 knows why the caged bird singsTreeckcold57 knows why the caged bird singsTreeckcold57 knows why the caged bird singsTreeckcold57 knows why the caged bird singsTreeckcold57 knows why the caged bird sings


Default

Quote:
Originally Posted by badsykes View Post
For AMD is urgent to bring Vega stock and also price the partner cards lower than what are they now.
Good luck with that. People is still all over cryto mining. Unless AMD able to create Navi to separate two different architectures that designed for compute while other one that designed for gaming. But honestly, I doubt AMD would do that. AMD tend to go all in one round for GCN architecture.
__________________

AMD Phenom II X2 555 @ stock clock
Xigamtek Knight cooler
ASUS M4A79XTD EVO
G.Skill 8GB DDR3 1333 (4x4GB)
Intel 530 240GB SSD
XFX ATI Radeon 4870 1GB
Antec Truepower 750W
NZXT Source 210
Windows 7 x64



AMD FX-8350 @ stock clock
Gigabyte GA-990FX-UD5 R5
G.Skill Sniper 16GB (8x2) DDR3 1866
Arctic Freezer 7 Pro 7 rev. 2
Gigabyte Windforce 7950 3GB Ghz Edition
Samsung 840 Pro 128GB SSD
EVGA SuperNova 650W
NZXT Source 210 w/ two Noctua F-12 fans
Ubuntu MATE 64-bit
Intel i5 3570K @ stock clock | G.Skill 16GB (8GBx2) DDR3 1866 | Silicon Power 60GB SSD | Win 10 Pro x64 | NZXT Source 210
Treeckcold57 is offline   Reply With Quote
Old Dec 30, 2017, 12:53 PM   #32
badsykes
Radeon Arctic Islands
 
Join Date: May 2007
Location: Romania Romania, Constanta
Posts: 5,046
badsykes once held a door open for a complete strangerbadsykes once held a door open for a complete strangerbadsykes once held a door open for a complete stranger


Default

Oh ... wow...Now i see.The biggest hardware shop in my country has no vega in stock.No reference or third party...I thought vega is not profitable for crypto because of the high price/high power ... Now i realisead that crypto on gpu is not dead yet ...
Haha

https://www.ebay.co.uk/sch/i.html?_f...c&LH_PrefLoc=3

On ebay vega56 is 850 pounds // over 900$ ... If i sell my card i get a profit ...
The funny thing is Vega frontier is the same price on second hand market )
AMD hit the jackpot with vega outside gaming...Only if they can sustain the demand...
__________________
"There is no beggining, and there is no end.There is no alpha, and there is no omega.You never began, and you will never end."
badsykes is offline   Reply With Quote
Old Dec 30, 2017, 01:00 PM   #33
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by bill dennison View Post
since vega 64 lq was released at the same 699 price of a 1080 ti fe it should have been as fast or at most 5% less .
to be that price and only match the 1080 non ti sorry it is a fail .


One thing to keep in mind though, it that NVidia was sufficiently worried about Vega to even release the GTX1080 TI featuring big Pascal in January 2017 at that same 700$ price point, exactly when AMD had the first unveiling of Vega.


The GP 104 ( little Pascal ) was released in may ~ june 2016 and Nvidia also wanted 700$ for it before cutting it's price significantly with the release of the TI version, and the one thing to keep in mind is that Vega seriously outguns the GP 104 in shading power, memory bandwidth, texturing..... 9 Teraflop vs 13 teraflop is not a small difference and even outguns Big Pascal on that front, even if to a smaller degree.


Said it earlier that games still aren't demanding enough....
shadow001 is offline   Reply With Quote
Advertisement (Guests Only)
Login or Register to remove this ad
Old Dec 30, 2017, 01:05 PM   #34
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by Treeckcold57 View Post
Good luck with that. People is still all over cryto mining. Unless AMD able to create Navi to separate two different architectures that designed for compute while other one that designed for gaming. But honestly, I doubt AMD would do that. AMD tend to go all in one round for GCN architecture.

Nvidia did the same thing with the latest Titan release based off the Volta compute chip, and where the gains in gaming considering the amount of hardware on that chip, just aren't that huge but if you want the power you pay the 1200$ price tag, straight from Nvidia's shop like all previous titan releases.


If people really want and need the power for their gaming, that's the only place to shop but the price tag on those things makes them think twice most likely....


Edit: My bad, Titan V featuring Volta are 3000$ a pop.... O_O

Last edited by shadow001 : Dec 30, 2017 at 01:13 PM.
shadow001 is offline   Reply With Quote
Old Dec 30, 2017, 02:29 PM   #35
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

The reason I said earlier that Nvidia tipped it's hand with Volta is that if one looks at it's spec sheet, all they really need to do is disable / remove double precision math at half clock, remove those specialised tensor cores for deep learning which would make for a smaller and cheaper to make chip, and the power savings from the removal of said hardware allows to clock the chip higher while still remaining a 250 Watt part.


There's your Volta gaming chip right there.
shadow001 is offline   Reply With Quote
Old Dec 31, 2017, 01:30 AM   #36
badsykes
Radeon Arctic Islands
 
Join Date: May 2007
Location: Romania Romania, Constanta
Posts: 5,046
badsykes once held a door open for a complete strangerbadsykes once held a door open for a complete strangerbadsykes once held a door open for a complete stranger


Default

Quote:
Originally Posted by bill dennison View Post
since vega 64 lq was released at the same 699 price of a 1080 ti fe it should have been as fast or at most 5% less .
to be that price and only match the 1080 non ti sorry it is a fail .



I don't think it is going to happen at this point

the vega 56 needs to be 399
vega 64 air needs to be 449


and vega 64 lq needs to be 549 to to match a 1080 non ti price

and I don't think they can make that many to over come the crap coin miners even if they could sell them at that price

hell the vega LQ is now selling at 1099 on newegg and OOS almost 300 bucks more than a asus 1080 ti strix at 809 and the strix is almost a 100 over it's msrp

From what i read AMD is working to bring a DDR6 controller to vega.THe achile hill of the vega is the hbm2 price/performance.The use of this memory made the Vega very inflexible from value perspective.Is easy target for nvidia.I speak strictly if used for gaming purporse..If one can see put the scientific performance into the basket too then you have a great value card even at 700$ (vega 64).It still needs tweaks to get the most out of it and i am saying you have to go beyond just ocing.
For vega 56 a vega 64 bios is needed for HBM2 tweaking..An aftermarket cooler is needed for ocing and also tweaks in software...All this steps are not for begginers.You need to mess with warranty unless you have to 2 years to get out of warranty.Plenty of stock.At that point may or may not matter anymore ..
Nvidia made things simpler...The gtx 10x0 is very gaming friendly.Ocing friendly and easy.Nice temps and power use.Nice performance out of box.No bios messing.Also nvidia have resources to throw at devs for making their cards as pain free as possible...All in all easy to achieve for begginers.
It really needs an enthuziast eye to see deeply and even may or may not matter for the person.With AMD is like a lottery a bit.Some people sold their 290 for the same price or more as they bought it.With Nvidia this didn't happen too often...With AMD value comes from exceptional circumstances.
The vega is in the same boat now.Looking at ebay i can sell my card for profit.
__________________
"There is no beggining, and there is no end.There is no alpha, and there is no omega.You never began, and you will never end."
badsykes is offline   Reply With Quote
Old Dec 31, 2017, 11:04 AM   #37
pax
Rage3D Spammer
 
Join Date: Apr 2001
Location: Canada Grand Falls, New Brunswick, Canada
Posts: 26,536
pax knows why the caged bird singspax knows why the caged bird singspax knows why the caged bird singspax knows why the caged bird singspax knows why the caged bird sings


Default

Where did you read that about gddr6 on vega? I doubt it but maybe for the 12nm refresh if hbm2 is still problem to manufacture...

Theres no point in amd segmenting their gpus and making compute only cards like nvidia. And they do in a way actually where they dont have display ports on them. Both lines sell out anyway.

Volta 2080 will be 1500$ us. Making a gaming chip wont make it that much cheaper or faster from what we've seen.
__________________
I talked to the tree. Thats why they put me away!..." Peter Sellers, The Goon Show
Only superficial people cant be superficial... Oscar Wilde

Piledriver Rig 2016: Gigabyte G1 gaming 990fx. FX 8350 cpu. Asus R290 DCUII Cats 19.9.3, SoundBlaster ZXR, 2 x 8 gig ddr3 1866 Kingston. 1 x 2tb Firecuda seagate with 8 gig mlc SSHD. Sharp 60" 4k 60 hz tv. Win 10 home.

Ryzen Rig 2017: Gigabyte X370 K7 F40 bios. Ryzen 1700x. 2 x 8 ddr4 3600 (@3200) Gskill. Sapphire Vega 64 Reference Cooler Cats 19.9.3. Soundblaster X Ae5 May 10th 2018 driver. 28" Upstar 4k ips 60hz panel, Intel 600p NVME 512GB. 4 TB HGST NAS HD. Win 10 pro.

Ignore List: Keystone... -My Baron he wishes to inform you that vendetta, as he puts it in the ancient tongue, the art of kanlee is still alive... He does not wish to meet or speak with you...-
"Either half my colleagues are enormously stupid, or else the science of darwinism is fully compatible with conventional religious beliefs and equally compatible with atheism." -Stephen Jay Gould, Rock of Ages.
"The Intelligibility of the Universe itself needs explanation. It is not the gaps of understanding of the world that points to God but rather the very comprehensibility of scientific and other forms of understanding that requires an explanation." -Richard Swinburne


www.realitysandwich.com

www.makepovertyhistory.org
pax is offline   Reply With Quote
Old Dec 31, 2017, 11:35 AM   #38
bill dennison
Radeon Arctic Islands
 
Join Date: Jan 2007
Location: United States phoenix
Posts: 19,251
bill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Diesel


Default

Quote:
Originally Posted by badsykes View Post
From what i read AMD is working to bring a DDR6 controller to vega.THe achile hill of the vega is the hbm2 price/performance.The use of this memory made the Vega very inflexible from value perspective.Is easy target for nvidia.I speak strictly if used for gaming purporse..If one can see put the scientific performance into the basket too then you have a great value card even at 700$ (vega 64).It still needs tweaks to get the most out of it and i am saying you have to go beyond just ocing.
For vega 56 a vega 64 bios is needed for HBM2 tweaking..An aftermarket cooler is needed for ocing and also tweaks in software...All this steps are not for begginers.You need to mess with warranty unless you have to 2 years to get out of warranty.Plenty of stock.At that point may or may not matter anymore ..
Nvidia made things simpler...The gtx 10x0 is very gaming friendly.Ocing friendly and easy.Nice temps and power use.Nice performance out of box.No bios messing.Also nvidia have resources to throw at devs for making their cards as pain free as possible...All in all easy to achieve for begginers.
It really needs an enthuziast eye to see deeply and even may or may not matter for the person.With AMD is like a lottery a bit.Some people sold their 290 for the same price or more as they bought it.With Nvidia this didn't happen too often...With AMD value comes from exceptional circumstances.
The vega is in the same boat now.Looking at ebay i can sell my card for profit.
Quote:
A rumor recently surfaced involving the LinkedIn page of an AMD technical engineer that listed the company working on a GDDR6 memory controller. I reached out to some people and can confirm that AMD is indeed working on the GDDR6 standard and will be actively using it in future graphics cards. As for the question that I am sure everyone will ask next, no AMD will still be sticking to HBM2 for its high end next generation graphics cards in 2018 (aka Navi )
https://wccftech.com/amd-gddr6-memor...raphics-cards/

that is for mid range cards from what I have read

the would have to respin vega and add a new memory controller
not going to happen they are working on Navi at this point and if it is a MCM-GPU it will be HBM also
bill dennison is online now   Reply With Quote
Old Dec 31, 2017, 02:25 PM   #39
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by bill dennison View Post
https://wccftech.com/amd-gddr6-memor...raphics-cards/

that is for mid range cards from what I have read

the would have to respin vega and add a new memory controller
not going to happen they are working on Navi at this point and if it is a MCM-GPU it will be HBM also

Vega itself as it is right now, doesn't have memory controllers at all for the memory...They're located at the base of each HBM memory stack, and I don't see GDDR6 even being competitive since it can't reach the raw bandwidth figures of HBM 2.0.....Nvidia's Volta in their 3000$ workstation card uses 3 HBM stacks rather than 2 on Vega, and just that alone allows for about ~650GB/sec of memory bandwidth.


The 10 000$ Volta card used for Compute which is basically the same chip as the workstation card but with all hardware enabled, has all 4 HBM stacks working as well, allowing for pretty close to 1 terabyte / sec of memory bandwidth......GDDR6 isn't even a option to compete, never mind the much lower latency of HBM 2.0 since the chips are right beside the die, and don't need to run at such high clock speeds, so the memory timings are lower and that's never a bad thing.
shadow001 is offline   Reply With Quote
Old Dec 31, 2017, 02:54 PM   #40
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

The whole Multi chip GPU package does make sense, but only when fabrication processes really have hit a wall that is insurmountable, and we're not there just yet even though mighty Intel continues to delay their 10nm product line, primarily because the entire chip is made at 10nm, not just certain parts of the chip that are easier to mass produce, which is what both TSMC and Global's 12nm and 7nm are all about.


Maybe the generation after Navi and Volta will be MCM designs because there's no other option, but we haven't hit that wall quite yet.
shadow001 is offline   Reply With Quote
Old Jan 1, 2018, 12:34 AM   #41
badsykes
Radeon Arctic Islands
 
Join Date: May 2007
Location: Romania Romania, Constanta
Posts: 5,046
badsykes once held a door open for a complete strangerbadsykes once held a door open for a complete strangerbadsykes once held a door open for a complete stranger


Default

If the card is selling so well regarding of the market segment why raja quit amd after all.Lisa said in an interview that the crypto market is very little percentage in their sales.
__________________
"There is no beggining, and there is no end.There is no alpha, and there is no omega.You never began, and you will never end."
badsykes is offline   Reply With Quote
Old Jan 1, 2018, 02:31 AM   #42
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by badsykes View Post
If the card is selling so well regarding of the market segment why raja quit amd after all.Lisa said in an interview that the crypto market is very little percentage in their sales.

He did get one hell of a job offer at Intel though, along with the corresponding pay said position offers.....Intel chief architect, senior vice president of the newly formed Core and Visual Computing Group.


For someone that lead the team that designed the "failure" that is Vega, that sounds like a pretty good job promotion....
shadow001 is offline   Reply With Quote
Old Jan 1, 2018, 10:42 AM   #43
pax
Rage3D Spammer
 
Join Date: Apr 2001
Location: Canada Grand Falls, New Brunswick, Canada
Posts: 26,536
pax knows why the caged bird singspax knows why the caged bird singspax knows why the caged bird singspax knows why the caged bird singspax knows why the caged bird sings


Default

I dont see how Raja can do a gpu at intel without some significant licensed IP. Intel have tried many times over the years and not been able to do much and thats even with nvidia license. Its likely the amd license is more extensive and they need someone who knew it to make it work. It probably will mean more money for amd as well.

For intel to throw billions at this again might mean they have a chance this time to becoming a third major gpu producer albeit under license.
__________________
I talked to the tree. Thats why they put me away!..." Peter Sellers, The Goon Show
Only superficial people cant be superficial... Oscar Wilde

Piledriver Rig 2016: Gigabyte G1 gaming 990fx. FX 8350 cpu. Asus R290 DCUII Cats 19.9.3, SoundBlaster ZXR, 2 x 8 gig ddr3 1866 Kingston. 1 x 2tb Firecuda seagate with 8 gig mlc SSHD. Sharp 60" 4k 60 hz tv. Win 10 home.

Ryzen Rig 2017: Gigabyte X370 K7 F40 bios. Ryzen 1700x. 2 x 8 ddr4 3600 (@3200) Gskill. Sapphire Vega 64 Reference Cooler Cats 19.9.3. Soundblaster X Ae5 May 10th 2018 driver. 28" Upstar 4k ips 60hz panel, Intel 600p NVME 512GB. 4 TB HGST NAS HD. Win 10 pro.

Ignore List: Keystone... -My Baron he wishes to inform you that vendetta, as he puts it in the ancient tongue, the art of kanlee is still alive... He does not wish to meet or speak with you...-
"Either half my colleagues are enormously stupid, or else the science of darwinism is fully compatible with conventional religious beliefs and equally compatible with atheism." -Stephen Jay Gould, Rock of Ages.
"The Intelligibility of the Universe itself needs explanation. It is not the gaps of understanding of the world that points to God but rather the very comprehensibility of scientific and other forms of understanding that requires an explanation." -Richard Swinburne


www.realitysandwich.com

www.makepovertyhistory.org
pax is offline   Reply With Quote
Old Jan 1, 2018, 11:57 AM   #44
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by pax View Post
I dont see how Raja can do a gpu at intel without some significant licensed IP. Intel have tried many times over the years and not been able to do much and thats even with nvidia license. Its likely the amd license is more extensive and they need someone who knew it to make it work. It probably will mean more money for amd as well.

For intel to throw billions at this again might mean they have a chance this time to becoming a third major gpu producer albeit under license.


Hard to say whats intel's plan in the longer run, but the title of said job offer does give a clue " Core and visual computing ", meaning it's not just graphics cores mated to a traditional CPU in the same die, and it's no secret that Intel has been trying to kill the X-86 architecture for a long time.....If it wasn't for AMD introducing the 64 bit version of X-86, the original 32 bit X-86 instruction set would have been buried a long time ago.



With GPU's becoming ever more flexible in terms of the workloads they can handle, i don't think past IP and who owns what means much anymore, and if a newer version of any given feature that may have started with said IP, is now much faster and handles more workloads than the original was designed to handle to begin with.....It wouldn't be a copy and paste type deal that's easy to take to court basically.



Intel is lagging behind on the CPU front when it comes to raw math power when compared to GPU's for specific workloads, as it's 18 core desktop CPU just about hits 1 teraflop even in dual precision floating point, while a Volta GPU is 6x faster so for super computer workloads requiring that level of precision, and needing as much computing power as possible, still having backwards compatibility by supporting the X-86 instruction set isn't doing them any favors.....It's like the boat anchor they always have to drag along, no matter how good the hardware is, design wise....



We'll see what the future holds i guess.

Last edited by shadow001 : Jan 1, 2018 at 12:38 PM.
shadow001 is offline   Reply With Quote
Old Jan 1, 2018, 02:07 PM   #45
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Going back to Navi, it really needs to include double precision math support at half clock which the last AMD GPU to have it onboard was AMD's Hawaii GPU (R290x), and both Fiji and Vega support it but it's at a much lower performance level that similar to Nvidia's Pascal ( 1/32 clock ), and Volta does support it at half clock so it's a requirement for at least the version of Navi intended for the workstation and compute markets in order to have something competitive relative to Volta.



It is strange that the Volta version shown in reviews has no better single precision or half precision math performance than Vega as shown here in the chart posted in AnandTech's review:






Single precision math is at 13.7 teraflop used in graphics effects like shading and half precision is clocks in at 27.4 teraflop and can be used in AI and physics, and Vega is pretty much right up there on both those fronts at 13 and 26 teraflop respectively......Pascal doesn't support half precision.


Given that like it or hate it, Volta is still faster than Pascal in gaming despite the price tag, just how much of these shaders are really being used in gaming right now, even with the latest game releases, as Vega has similar shading power to Volta but is no where near close to that level of gaming performance to begin with?

Last edited by shadow001 : Jan 1, 2018 at 02:41 PM.
shadow001 is offline   Reply With Quote
Old Jan 1, 2018, 07:32 PM   #46
Treeckcold57
Good ol' ATI
 
Join Date: Sep 2004
Posts: 17,021
Treeckcold57 knows why the caged bird singsTreeckcold57 knows why the caged bird singsTreeckcold57 knows why the caged bird singsTreeckcold57 knows why the caged bird singsTreeckcold57 knows why the caged bird sings


Default

Quote:
Originally Posted by shadow001 View Post
Going back to Navi, it really needs to include double precision math support at half clock which the last AMD GPU to have it onboard was AMD's Hawaii GPU (R290x), and both Fiji and Vega support it but it's at a much lower performance level that similar to Nvidia's Pascal ( 1/32 clock ), and Volta does support it at half clock so it's a requirement for at least the version of Navi intended for the workstation and compute markets in order to have something competitive relative to Volta.



It is strange that the Volta version shown in reviews has no better single precision or half precision math performance than Vega as shown here in the chart posted in AnandTech's review:



Single precision math is at 13.7 teraflop used in graphics effects like shading and half precision is clocks in at 27.4 teraflop and can be used in AI and physics, and Vega is pretty much right up there on both those fronts at 13 and 26 teraflop respectively......Pascal doesn't support half precision.


Given that like it or hate it, Volta is still faster than Pascal in gaming despite the price tag, just how much of these shaders are really being used in gaming right now, even with the latest game releases, as Vega has similar shading power to Volta but is no where near close to that level of gaming performance to begin with?
Probably it has somewhere bottleneck else in Vega's new architecture...for instance, the ROP? But other than that. It's still excellent for compute side performance. Vega was designed to be a GPGPU card...well...it's all in round- Workstation, HPC, and gaming. While Geforce is gaming while gimped on compute side but Titan and Quadro.
__________________

AMD Phenom II X2 555 @ stock clock
Xigamtek Knight cooler
ASUS M4A79XTD EVO
G.Skill 8GB DDR3 1333 (4x4GB)
Intel 530 240GB SSD
XFX ATI Radeon 4870 1GB
Antec Truepower 750W
NZXT Source 210
Windows 7 x64



AMD FX-8350 @ stock clock
Gigabyte GA-990FX-UD5 R5
G.Skill Sniper 16GB (8x2) DDR3 1866
Arctic Freezer 7 Pro 7 rev. 2
Gigabyte Windforce 7950 3GB Ghz Edition
Samsung 840 Pro 128GB SSD
EVGA SuperNova 650W
NZXT Source 210 w/ two Noctua F-12 fans
Ubuntu MATE 64-bit
Intel i5 3570K @ stock clock | G.Skill 16GB (8GBx2) DDR3 1866 | Silicon Power 60GB SSD | Win 10 Pro x64 | NZXT Source 210
Treeckcold57 is offline   Reply With Quote
Old Jan 1, 2018, 11:35 PM   #47
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by Treeckcold57 View Post
Probably it has somewhere bottleneck else in Vega's new architecture...for instance, the ROP? But other than that. It's still excellent for compute side performance. Vega was designed to be a GPGPU card...well...it's all in round- Workstation, HPC, and gaming. While Geforce is gaming while gimped on compute side but Titan and Quadro.

Could be i suppose, but one thing is for sure the shaders aren't being used in gaming to the point that it shows performance wise, since Nvidia's GP 104 found in the regular GTX1080 has some 9 teraflop worth of shading power, yet still runs neck and neck with Vega, which packs 13 teraflop across 4096 shaders.


It'll be fun to watch in a couple of years time, even if most have moved on to faster hardware by then, what the situation looks like once a lot more shader intensive games are out by then......
shadow001 is offline   Reply With Quote
Old Jan 2, 2018, 12:21 AM   #48
pax
Rage3D Spammer
 
Join Date: Apr 2001
Location: Canada Grand Falls, New Brunswick, Canada
Posts: 26,536
pax knows why the caged bird singspax knows why the caged bird singspax knows why the caged bird singspax knows why the caged bird singspax knows why the caged bird sings


Default

Quote:
Originally Posted by shadow001 View Post
Could be i suppose, but one thing is for sure the shaders aren't being used in gaming to the point that it shows performance wise, since Nvidia's GP 104 found in the regular GTX1080 has some 9 teraflop worth of shading power, yet still runs neck and neck with Vega, which packs 13 teraflop across 4096 shaders.


It'll be fun to watch in a couple of years time, even if most have moved on to faster hardware by then, what the situation looks like once a lot more shader intensive games are out by then......
GN said Vega was memory bw limited and that GCN is maxing out at 4096 shaders...

Dunno but seems no driver can improve the perf much anymore so the last thing to look forward to is FP16 perf in games like Farcry 5.
__________________
I talked to the tree. Thats why they put me away!..." Peter Sellers, The Goon Show
Only superficial people cant be superficial... Oscar Wilde

Piledriver Rig 2016: Gigabyte G1 gaming 990fx. FX 8350 cpu. Asus R290 DCUII Cats 19.9.3, SoundBlaster ZXR, 2 x 8 gig ddr3 1866 Kingston. 1 x 2tb Firecuda seagate with 8 gig mlc SSHD. Sharp 60" 4k 60 hz tv. Win 10 home.

Ryzen Rig 2017: Gigabyte X370 K7 F40 bios. Ryzen 1700x. 2 x 8 ddr4 3600 (@3200) Gskill. Sapphire Vega 64 Reference Cooler Cats 19.9.3. Soundblaster X Ae5 May 10th 2018 driver. 28" Upstar 4k ips 60hz panel, Intel 600p NVME 512GB. 4 TB HGST NAS HD. Win 10 pro.

Ignore List: Keystone... -My Baron he wishes to inform you that vendetta, as he puts it in the ancient tongue, the art of kanlee is still alive... He does not wish to meet or speak with you...-
"Either half my colleagues are enormously stupid, or else the science of darwinism is fully compatible with conventional religious beliefs and equally compatible with atheism." -Stephen Jay Gould, Rock of Ages.
"The Intelligibility of the Universe itself needs explanation. It is not the gaps of understanding of the world that points to God but rather the very comprehensibility of scientific and other forms of understanding that requires an explanation." -Richard Swinburne


www.realitysandwich.com

www.makepovertyhistory.org
pax is offline   Reply With Quote
Old Jan 2, 2018, 10:40 AM   #49
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by pax View Post
GN said Vega was memory bw limited and that GCN is maxing out at 4096 shaders...

Dunno but seems no driver can improve the perf much anymore so the last thing to look forward to is FP16 perf in games like Farcry 5.

I suppose it's possible, though shader effects in general aren't known to be memory bandwidth intensive to begin with, and we have the GTX 1080 (GP 104 ) with less memory bandwidth and still matching Vega's performance, while the TI version is pretty close to the same bandwidth as Vega and is still a fair amount faster overall.


The funny thing is that I sometimes run these little experiments such as using higher levels of AA, even though i'm already at 4K so I don't need to use that much to begin with, and even when I push it quite a bit higher than normal, the Fps doesn't even budge at all and AA is known to be memory bandwidth intensive to begin with.


I really do get the impression that the cards can handle a lot more work than the majority of current games throw at them, and that the bottleneck is somewhere else and that is the challenge presented to most GPU hardware engineers these days......Design the architecture in such a way to show it's a lot better even with current games that were never designed with said hardware in mind, or design it to have longer term endurance with a much higher workload in games that won't see the light of day for another couple of years....They're still very much in development.



That's going to be the problem even on Nvidia's end as they get a gaming volta out....It's obviously going to be more capable than Pascal in every possible way, but it still needs to show it in curent games that weren't designed with it in mind, in order to convince buyers to part with their money.....In a funny way, it's the software holding the hardware back, design wise.
shadow001 is offline   Reply With Quote
Old Jan 2, 2018, 06:52 PM   #50
NWR_Midnight
Radeon Volcanic Islands
 
Join Date: Jul 2001
Location: United States Under the Sun
Posts: 3,595
NWR_Midnight once won a refrigerator on 'The Price is Right'NWR_Midnight once won a refrigerator on 'The Price is Right'NWR_Midnight once won a refrigerator on 'The Price is Right'NWR_Midnight once won a refrigerator on 'The Price is Right'NWR_Midnight once won a refrigerator on 'The Price is Right'NWR_Midnight once won a refrigerator on 'The Price is Right'


Default

Quote:
Originally Posted by shadow001 View Post
I suppose it's possible, though shader effects in general aren't known to be memory bandwidth intensive to begin with, and we have the GTX 1080 (GP 104 ) with less memory bandwidth and still matching Vega's performance, while the TI version is pretty close to the same bandwidth as Vega and is still a fair amount faster overall.


The funny thing is that I sometimes run these little experiments such as using higher levels of AA, even though i'm already at 4K so I don't need to use that much to begin with, and even when I push it quite a bit higher than normal, the Fps doesn't even budge at all and AA is known to be memory bandwidth intensive to begin with.


I really do get the impression that the cards can handle a lot more work than the majority of current games throw at them, and that the bottleneck is somewhere else and that is the challenge presented to most GPU hardware engineers these days......Design the architecture in such a way to show it's a lot better even with current games that were never designed with said hardware in mind, or design it to have longer term endurance with a much higher workload in games that won't see the light of day for another couple of years....They're still very much in development.



That's going to be the problem even on Nvidia's end as they get a gaming volta out....It's obviously going to be more capable than Pascal in every possible way, but it still needs to show it in curent games that weren't designed with it in mind, in order to convince buyers to part with their money.....In a funny way, it's the software holding the hardware back, design wise.
Maybe it is a coding issue within the games, where the majority of them are coded and optimized specifically for Nvidia hardware. That is why you get a few selected titles that the Vega does well in, coming close,matching, and even rarely beating the 1080ti.. But when that happens, people say stop cherry picking titles, or the Oh 1 or 2 titles, but what about all the rest.. well.. all the rest are a disadvantage for AMD because of the selective coding. It wouldn't be difficult to have code that hampers performance. Remember in the past when changing the name of the EXE gave completely different results in some games. It wouldn't take much for developers, to send AMD calls down an unoptimized, or even deliberate path to slow it down. Specially if Nvidia is giving them kickbacks of some sort. I am not saying they are, but it would not surprise me, since we live in a world of manipulation and lies now days. Just a thought.
__________________
I speak my mind! if you can't handle that, you might want to leave, because **** is going to get real!!

~I had the right to remain silent, I just didn't have the ability. ~ Ron White
~You can't fix Stupid! ~ Ron White
~There's not a pill you can take; there's not a class you can go to. - ~Stupid is forever. ~ Ron White
~Life is a hard teacher, it gives you the test before it teaches you the lesson.
~It's never to late to have a good childhood! The older you are, the better the toys! ~ My Dad
~Live everyday as though it is your last, it can all end at any moment!

Last edited by NWR_Midnight : Jan 2, 2018 at 06:59 PM.
NWR_Midnight is offline   Reply With Quote
Old Jan 2, 2018, 08:29 PM   #51
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Well one has to take games development like any other business, and that's one intended to make money selling as many copies as possible and at the end of the day, Nvidia does have the majority of the market to themselves, so the odds of said games running on Nvidia hardware are inherently higher than on AMD hardware.......


They try to be as hardware agnostic as possible when developing games for sure, but the extra mile may go for what's out there in higher numbers and cater to their strong points as well.


One of the things AMD did with Vega is seriously beef up it's geometry handling capabilities, which has long been an Nvidia strong point going back an extremely long time, same for going bonkers on texture handling and fill rate, while AMD put stronger emphasis on shading power and more memory bandwidth for a long time now......It's signature tendencies for both companies, and not a question of who's right or wrong, but rather where the emphasis on the artwork within a given game is put.
shadow001 is offline   Reply With Quote
Old Jan 8, 2018, 09:50 PM   #52
bill dennison
Radeon Arctic Islands
 
Join Date: Jan 2007
Location: United States phoenix
Posts: 19,251
bill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Diesel


Default

Quote:
AMD All But Relinquishes High-End Desktop Graphics In 2018
https://wccftech.com/amd-relinquishe...graphics-2018/

bill dennison is online now   Reply With Quote
Old Jan 9, 2018, 12:15 AM   #53
badsykes
Radeon Arctic Islands
 
Join Date: May 2007
Location: Romania Romania, Constanta
Posts: 5,046
badsykes once held a door open for a complete strangerbadsykes once held a door open for a complete strangerbadsykes once held a door open for a complete stranger


Default

Quote:
Originally Posted by bill dennison View Post
Again the market objectively speaking doesn't need new cards except exotic 4k and stuff like this...This news is purely bad news because the competition has a roadmap...Nvidia and Intel are similar in a way.They have big budgets and can afford releasing a GPU or CPU every year or so even if it's needed or not.
I am not saying AMD didn't make some mistakes but in my eyes is not so big drama.The biggest drama is actually the stock problem and the AIB partners have low quantities and high prices.It's not like too many afford the 1080ti or Titans.I think the media is sucumbed in this competition war and forgot the end user.
__________________
"There is no beggining, and there is no end.There is no alpha, and there is no omega.You never began, and you will never end."
badsykes is offline   Reply With Quote
Old Jan 9, 2018, 08:34 AM   #54
Exposed
ESB Sports Bookie
 
Join Date: Jan 2002
Location: United States Orlando, FL
Posts: 5,406
Exposed knows why the caged bird singsExposed knows why the caged bird singsExposed knows why the caged bird singsExposed knows why the caged bird singsExposed knows why the caged bird sings


Default

Quote:
Originally Posted by badsykes View Post
Again the market objectively speaking doesn't need new cards except exotic 4k and stuff like this...This news is purely bad news because the competition has a roadmap...Nvidia and Intel are similar in a way.They have big budgets and can afford releasing a GPU or CPU every year or so even if it's needed or not.
I am not saying AMD didn't make some mistakes but in my eyes is not so big drama.The biggest drama is actually the stock problem and the AIB partners have low quantities and high prices.It's not like too many afford the 1080ti or Titans.I think the media is sucumbed in this competition war and forgot the end user.
Call it what it is....AMD is giving up on gamers. 2019 for just a Vega refresh?

They aren't interested in the midrange and high end graphics sector anymore, they're content with low end. Yes, low end. 1440p you will see your framerates drop further and further as newer games come out. You will be buying 1080 level cards from AMD for the next 2 years just for 1080p.

At this point it should have been AMD with the 1080 branding since that's exactly what you'll get from here on out.
__________________
Quote:
Originally Posted by Jay20016 View Post
But hey, let us not forget that while you bemoan about how Hollywood is bending over backwards for that market, you're also asking them to simultaneously consider the 'delicate' sensibilities of those state side who need to see a flag plastered all over the place like some Michael Bay movie barfed into another equally worse Michael Bay movie that then **** into the mouth of the team who designs the Call of Duty campaigns with all of their red, white and blue awesomeness with such an overly large boner for America that one does not need to actually 'fly' to the moon, they simply could tight rope on top if it. And if you're concerned about whether or not that penis is strong enough to be load bearing for such a journey, you forget that what fills all of those engorged crevices is freedom.
Exposed is offline   Reply With Quote
Old Jan 9, 2018, 10:09 AM   #55
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by bill dennison View Post


They should still transition gaming Vega to the 12nm process, if nothing else to make them cheaper by being a smaller die and use less power, assuming no other changes happen to the architecture itself.


That said, i always doubted Navi would be a reality, with it's release scheduled the following year after Vega....It's not in AMD's habit to release hardware that fast.
shadow001 is offline   Reply With Quote
Old Jan 9, 2018, 11:39 AM   #56
badsykes
Radeon Arctic Islands
 
Join Date: May 2007
Location: Romania Romania, Constanta
Posts: 5,046
badsykes once held a door open for a complete strangerbadsykes once held a door open for a complete strangerbadsykes once held a door open for a complete stranger


Default

Quote:
Originally Posted by Exposed View Post
Call it what it is....AMD is giving up on gamers. 2019 for just a Vega refresh?

They aren't interested in the midrange and high end graphics sector anymore, they're content with low end. Yes, low end. 1440p you will see your framerates drop further and further as newer games come out. You will be buying 1080 level cards from AMD for the next 2 years just for 1080p.

At this point it should have been AMD with the 1080 branding since that's exactly what you'll get from here on out.

The cards are compute specialized at heart.The gaming dedication was last thing on AMD list for this card.
__________________
"There is no beggining, and there is no end.There is no alpha, and there is no omega.You never began, and you will never end."
badsykes is offline   Reply With Quote
Old Jan 9, 2018, 01:26 PM   #57
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by badsykes View Post
The cards are compute specialized at heart.The gaming dedication was last thing on AMD list for this card.


Kinda funny when it's all said and done, and the almighty, all conquering GTX 1080 TI only does between 10~15 FPS better than a Vega can and does at 4k resolution.....Not bad for a 1080p card, where 1080p is 4x less demanding on the same game and settings with the only difference being the resolution itself.


Being that it comes from the usual *** of these boards, by trying to instill fear for future titles at 1440p and 1080p for those using Vega cards, I'm not surprised though.....
shadow001 is offline   Reply With Quote
Old Jan 9, 2018, 02:13 PM   #58
bill dennison
Radeon Arctic Islands
 
Join Date: Jan 2007
Location: United States phoenix
Posts: 19,251
bill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Dieselbill dennison exchanges holiday cards with Vin Diesel


Default

Quote:
Originally Posted by shadow001 View Post
Kinda funny when it's all said and done, and the almighty, all conquering GTX 1080 TI only does between 10~15 FPS better than a Vega can and does at 4k resolution.....Not bad for a 1080p card, where 1080p is 4x less demanding on the same game and settings with the only difference being the resolution itself.


Being that it comes from the usual *** of these boards, by trying to instill fear for future titles at 1440p and 1080p for those using Vega cards, I'm not surprised though.....
but kind of a big deal when that single 1080 ti is only just doing 4k gaming at 40 to 55 fps with aa off in new games

take 15 fps off that and you need crossfire and for it to work well, and the price of two vega 64's is nuts over a single 1080 ti



vega is the first ati/amd card I have skipped in years and I may have to go all nv if they keep this up

Last edited by bill dennison : Jan 9, 2018 at 02:19 PM.
bill dennison is online now   Reply With Quote
Old Jan 9, 2018, 02:26 PM   #59
badsykes
Radeon Arctic Islands
 
Join Date: May 2007
Location: Romania Romania, Constanta
Posts: 5,046
badsykes once held a door open for a complete strangerbadsykes once held a door open for a complete strangerbadsykes once held a door open for a complete stranger


Default

Hot Vega: Gigabyte Radeon RX Vega 56 Gaming OC 8G Review



http://www.tomshardware.com/reviews/...ew,5413-7.html

Intersting conclusion
__________________
"There is no beggining, and there is no end.There is no alpha, and there is no omega.You never began, and you will never end."
badsykes is offline   Reply With Quote
Old Jan 9, 2018, 05:11 PM   #60
shadow001
Captain thread derail....
 
Join Date: Jul 2003
Posts: 9,219
shadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete strangershadow001 once held a door open for a complete stranger


Default

Quote:
Originally Posted by bill dennison View Post
but kind of a big deal when that single 1080 ti is only just doing 4k gaming at 40 to 55 fps with aa off in new games

take 15 fps off that and you need crossfire and for it to work well, and the price of two vega 64's is nuts over a single 1080 ti



vega is the first ati/amd card I have skipped in years and I may have to go all nv if they keep this up


If that is the case in real world conditions where a single GTX1080 TI can't quite handle 4K, said user that already has one will have to sell it and hope that a Volta high end card finally pulls it off....still means spending more money even though high end Pascal is at most a year old if users bought one as soon as it was announced in January of last year.


Only other option is getting another TI, and go multi GPU as well.....Same boat as Vega in real world terms.
shadow001 is offline   Reply With Quote
Reply


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -5. The time now is 05:23 PM.



Powered by vBulletin® Version 3.6.5
Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
All trademarks used are properties of their respective owners. Copyright ©1998-2011 Rage3D.com
Links monetized by VigLink