[H]ard|OCP have posted their review of both of ASUS ROG STRIX RX O8G GAMING cards, one the Vega 56 the other the Vega 64. The Vega 56 model comes with an overclock of 1297MHz core, 1573MHz boost and the Vega 64 is similar with a core of 1298MHz and boost of 1590MHz; both cards memory clocks are at stock speed. [H]ard|OCP pushed them further with the Vega 64 set to 1700MHz and memory at 1050MHz in the driver while the Vega 56 managed 1720MHz and 850MHz memory, though it is worth noting neither card hit those speeds during use.
In the end the Vega 56 model happily surpassed the GTX 1070 Ti, even without the manual overclock and while the Vega 64 did struggle to match the GTX 1080 in some games, it was the winner in DX12 and Vulkan games. Check out [H]'s full review of the Vega 56 and Vega 64 to learn more about these two strong AMD cards..
"We have our first custom retail AMD Radeon RX Vega 56 GPU video card on the test bench in the form of the ASUS ROG STRIX RX Vega 56 O8G GAMING video card. ASUS has created a robust ROG STRIX version of RX Vega 56 with a factory overclock and overclocking ability. Let’s see what it can do. "
Here are some more Graphics Card articles from around the web:
- AMD Vega 8 Graphics Performance On Linux With The Ryzen 3 2200G @ Phoronix
- Radeon Linux OpenGL Driver Continues Giving Its Best Against Windows 10 @ Phoronix
- AMD Raven Ridge Graphics On Linux vs. Lower-End NVIDIA / AMD GPUs @ Phoronix
- External GPU Testing: GTX 1080 in a Box + Core i7-8550U Ultraportable @ Techgage
- Galax/KFA2 GeForce GTX 1050 Ti EXOC @ Guru of 3D
Now if the miners can not get
Now if the miners can not get their hands on these before the gamers that will be good.
The GTX1080(Based on GP104 silicon) was what the Vega 10 base die was, for gaming, engineered to compete with and not the GP102 based GTX1080Ti. So Vega 64(based on the Vega 10 die tapeout) competes farily well with the GTX1080 in DX11 titles and it’s even better for the Vega 64 in DX12/Vulkan against Nvidia’s GP104 based variants.
Nvidia having that GP102 tapeout is where the Vega 10 die tapeout can not compete. And it’s not for lack of Shaders or TMUs as the Vega 10 tapeout is heavy on the TMUs and Shader Cores(tons more Shaders than Nvidia’s consumer variants) but does not provide more than 64 ROPs. The GP102(96 total ROPs available) based GTX1080Ti’s 88 ROPs is where the Vega 10 tapeout, namely Vega 64, can not compete for raw pixel fill rates with Nvidia/GP102.
So the news for AMD Vega owners(in the future) is that the Games/Gaming Engine software is still not making use of all the new API(Vulkan/DX12) features. And there is still unused, mostly, in games/gaming engines the Vega hardware features(FP 16, Explicit Primitive Shaders, etc.) support to be added for games/gaming engines to target all of Vega’s new hardware features.
So Vega Fine Wine-2 is still something to look forward to once the Miners move on from buying up any and all discrete Vega SKUs they can get their hands on. I’m guessing that the next generation of gaming comsoles are in active development and will be Zen/Vega based next time around. So some console makers must be lookiing at Vega’s version of FP16 and Explicit Primitive Shaders as well as Vega’s HBCC/HBC(HBM2) IP to make use of and optimize for in consoles.
Hell I’ll bet that MS/Sony may even be looking at using Raven Ridge APUs as sorts of preliminary development platforms to test out some of the Vega FP16 and Explicit Primitive Shaders development while any Vega HBCC/HBC(HBM2) IP for gaming console development will have to wait for the actual next generation console development platforms that MS/Sony always create in-house along with their SOC/APU Partner.
Raven Ridge APU gaming will benefit from any development targeting any Zen/Vega based console SKUs as gaming developers look to get a head start on getting their games tweaked for the next generation of gaming consoles that will very likely be Zen/Vega based. And all that console game tweaking can benifit discrete Desktop GPU gaming also with all the extra optimization work that targets AMD’s graphics IP beacuse it’s used by both MS and Sony.
Hey, Best Buy(1) is now known
Hey, Best Buy(1) is now known as Best Spy and the Geek Squad has been renamed and now known as the Sneak Squad. They’ll sniff around all your files like a prize truffle hog!
(1)
“Fresh docs detail 10-year link between Geek Squad informers and Feds”
https://www.theregister.co.uk/2018/03/07/new_docs_detail_the_tenyear_relationship_between_geek_squad_informers_and_feds/
Since these cards are usually
Since these cards are usually unavailable and/or overpriced I don’t see the point
in the news story. Reviewing something u cannot buy is pointless.
Indeed this is the longest period of time I have not purchased a new video card, ever.
Its quite obvious that production has not been lifted to meet demand. We are nearly eight months into the Current Card Crysis Catastrophe.
That’s because a big block of
That’s because a big block of the Top Binned Vega 10 dies are destined to be used in the Radeon Pro WX 9100s/Radeon SSGs(the 9100 with on GPU PCIe card SSD) and Radeon Instinct compute and AI SKUs. So the professional compute/AI markets are taking loads of Vega 10 dies. Just look at that Project 47 petaflop(SP FP) supercomputer in a single cabinet that uses 20, 32 core Epyc CPUs/SP3 MBs and 80 Vega 10 dies in the form of 80 Radeon Instinct MI25s per cabinet.
There is where many of the Vega 10 dies are going even bofore the remaining stock of lower binned Vega 10 dies that do not make the grade for pro usage are available for the AIB partners and those lower binned dies are even fruther binned for the AIB’s various SKUs. So there is not much remaining for the miners and gamers to go after and AMD did get burned the last time around overproducing for the a coin market that was not stable and AMD had a take a big writedown on unsold GPU inventory that AMD swore it would never allow to happen again.
So the gamers way back at that time when AMD’s GPU prices where lower than Marianas Trench where damn happy but AMD was far from happy with that outcome. So for gamers what goes around comes around and AMD’s Vega 10 GPU Dies with all that excess of Compute(In the Form of Shader cores) are more poular for compute workloads(mining, AI, GPGPU) than they are for gaming and AMD is selling all the Vega 10 dies that they can produce.
So then is then and now is now and all those Extra Vega 10 shader cores are earning AMD plenty of revenues with no stock remainging on the inventory sheets for AMD to have to write down as a loss. You can thank JHH over at Nvidia for creating that compute demand for GPUs in the professional markets that AMD is also profiting from in tandem along with the Epyc professional CPU sales. In fact Epyc paired with Radeon Pro WX 9100s and Epyc paired with Radeon Instinct MI25s/others make for some very popular parings for the Compute and that Mad New AI market that’s going to usher in a new multi-billion dollar revenue stream for those CPU and GPU makers ready. And AMD has both a CPU and GPU foot in the door to that new revenue pot-o-gold.
Blockchain is more than just coin-hashes and there are other uses for blcokchain as well as AI/Compute Accelerated on GPUs. AMD is still going to get more on the Graphics market back in the form of integrated graphics market share taken from Intel(has that majority integrated graphics market currently) and both AMD and Nvidia are selling plenty of GPUs for all the markets.
Both AMD and Nvidia really do not need a brand new GPU micro-arch. Nvidia still has its GP102 tapeout with its 96 available ROPs with the GP102 based GTX1080Ti only using 88 out of the GP102’s full complement of 96 ROPs and AMD only really needs to just tapeout a new Vega GPU Micro-Arch based die with at least 96 ROPs to compete with GP102.
Really there is no great need to please gamers for all things as both Nvidia and AMD are selling all the dies that they cane make on their higher end offerings. AMD’s Epyc sales alone will begin to surpass AMD’s GPU revenues and the Vega 10 dies used for the professional market actually get a great markups for both AMD, the same for Nvidia on Nvidia’s pro SKUs. And That’s way over the amounts that any consumer market can demand. Lisa Su stated that AMD will no longer be the low price GPU maker and that’s one thing that she made true in focusing on making Vega 10 a compute powerhouse and Vega 20 at 7nm is not for the consumer gaming market in the first place.
AMD’s gaming market share will grow in spite of most of AMD’s flagship(Even Mainstream) diecrete GPUs going to miners as AMD will take back the APU/Integrated GPU market with Raven Ridge mobile, and now desktop Raven Ridge also. Billions of folks will game on AMD RR APUs around the world and that APU/Integrated Graphics market will pull in the games/gaming engine developers just as the console games developers tweak their games for the consoles. The Next generation of Comsoles will be Zen/Vega also so there probably is a few developers targeting Raven Ridge’s FP16 and Explicit primitive shaders in advance of getting their hand on the next generation console development platforms from Sony/MS.
AMD really need to get its own version of mini-PC similar to Intel’s NUC and get some of that eSports and HTPC market revenues. Laptops and Mini-PC/NUC sorts of s revenues are larger than any Top end gaming rig revenues. And that’s going to be where AMD can get the majority of its gaming market share the quickest with Raven Ridge APUs. Disctrete Mobile Vega in on the way also with at least 4GB of HBM2 so that’s going to be popular on laptops alongside Raven Ridge APUs with their Vega integrated graphics.
Vulkan 1.1 has that multi-GPU working for GPUs of the same make/model/micro-arch so that will supplant the deprecated CrossFire in the drivers IP. Khronos is working on getting Vulkan working for multi-GPU configurations of different makers GPUs to reach feature parity with DX12, but at least for Raven Ridge APUs Vulkan 1.1 will also for integrated Vega/GCN Graphics paired with discrete mobile Vega/GCN graphics for the HTPC/Steam Box market on AMD’s CPU/GPU Zen/Vega/GCN IP.
2000MHz vs 1600MHz (NVidia
2000MHz vs 1600MHz (NVidia Pascal vs AMD Vega):
25% difference!!
With this Asus Strix card managing solid 1600MHz and similar NVidia cards hitting regular solid 2000MHz I find it exciting that AMD could get a big boost to performance simply by OPTIMIZING its fab process and chip layout.
Add some other improvements on to this for hardware AND software (especially DX12/Vulkan, game engine) and we could get a big jump forward for AMD products.
(Not that NVidia will sit still but it’s worth noting where the current differences are)
The way to OC Vega is to
The way to OC Vega is to undervolt it.1000mV S6 and 1050mV S7, to reduce heat generation. Vega is ultimately thermally limited.
That extra power usage is
That extra power usage is because of those thousands more shader cores on the Vega 10 base die based Vega 56 and 64 GPUs that where really designed Raja/Team(Lesa told Raja to make a compute/AI oriented Vega 10 die tapeout) more for compute/AI ueage. The coin miners love that for sure, So that’s where Vega is! It’s In that compute/AI Radeon WX 9100s and MI25s and also down in the hashing mines hashing coins. Poor gamers can’t get their hands on Vega 56/64 but there is Raven Ridge Ryzen G series desktop APUs there for gaming until the discrete desktop GPU pricing comes down.
Vega is actually a very efficient GPU micro-arch and that larger power usage metric from the Vega 10 base die tapeout is the direct result of those extra shaders and is not the fault of the Vega GPU micro-arch. Miners love them some shaders and could care less about the Nvidia GTX 1080Ti’s 88 ROPs that are only useful to fling out the high FPS metrics. The Vega 10 base die tapeout, the only one tapeout that AMD could afford to do at that time, had to do double duty as a professional GPU Compute/AI oriented tapeout and a gaming oriented tapeout to compete with the GP104 based GTX 1080(64 ROPs max). So it’s the GTX 1080 against the Vega 10 based Vega 64(64 ROPs max).
Vega is a great competitor against the GTX 1080 in gamimg and that Vega 10 base die design was frozen long before Nvidia used the GP102 base die tapeout for the GTX 1080Ti(88 ROPs out of the GP102’s 96 available ROPs).
AMD probably has the funds to do a new Vega high end base die tapeout with more than 64 ROPs max, but AMD is very busy getting that Vega 20 professional compute/AI base die tapeout for that high margin professional market where the real money is made. AMD’s selling all the Vega silicon that it can make curremtly so that good for AMD.
Discrete Mobile Vega is what’s coming next and that’s going to have 4GB of HBM2 for Vega’s HBCC to use as HBC and mobile games with large amounts of textures that can be larger than 4GB in size with Vega’s HBCC/HBC(HBM2) making that IP easy to take advantage of. Discrete mobile Vega also has Rapid Packed math and Explicit primitive shaders IP to make use of for laptop/mobile gaming. So games developers will be tweaking their games for that with some extra future motivation in knwoing that in the process of tweaking games for Discrete Vega and Zen/Vega that they will also be ahead of the game for tweaking their games to run on the next gemeration of consoles which will also be using Zen with Vega graphics in those gaming console systems.