A confusing market
We test the Vega 64, Vega 64 Liquid, and Vega 56.
I feel like I have been writing about AMD non-stop in 2017. Starting with the release of Ryzen 7 and following through last week’s review of the HEDT Threadripper processor, AMD has gone from a nearly-dormant state in 2015-2016 to a wildly active and successful organization with more than a dozen new product launches under its belt. Today we will reveal the company's first consumer products based on the new Vega GPU architecture, thrusting the Radeon brand back into the fight at the $400+ price segments.
At this point, with architecture teases, product unboxings, professional card reviews, and pricing and availability reveals, we almost know everything we need to know about the new Radeon RX Vega 64 and RX Vega 56 products. Almost. Today we can show you the performance.
I want to be honest with our readers: AMD gave me so little time with these cards that I am going to gloss over some of the more interesting technological and architectural changes that Vega brings to market. I will come back to them at a later time, but I feel it is most important for us to talk about the performance and power characteristics of these cards as consumers finally get the chance to spend their hard-earned money on them.
If you already know about the specifications and pricing peculiarities of Vega 64 and Vega 56 and instead want direct access to performance results, I encourage you to skip ahead. If you want a refresher those details, check out the summary below.
Interesting statistics from the creation of this review in a VERY short window:
- 175 graphs
- 8 cards, 8 games, 2 resolutions, 3 runs = 384 test runs
- >9.6 TB of raw captured video (average ~25 GB/min)
Radeon RX Vega 64 and Vega 56 Specifications
Much of the below is sourced from our Vega 64/56 announcement story last month.
Though the leaks have been frequent and getting closer to reality, as it turns out AMD was in fact holding back quite a bit of information about the positioning of RX Vega for today. Radeon will launch the Vega 64 and Vega 56 today, with three different versions of the Vega 64 on the docket. Vega 64 uses the full Vega 10 chip with 64 CUs and 4096 stream processors. Vega 56 will come with 56 CUs enabled (get it?) and 3584 stream processors.
Pictures of the various product designs have already made it out to the field including the Limited Edition with the brushed anodized aluminum shroud, the liquid cooled card with a similar industrial design, and the more standard black shroud version that looks very similar to the previous reference cards from AMD.
| RX Vega 64 Liquid | RX Vega 64 Air | RX Vega 56 | Vega Frontier Edition | GTX 1080 Ti | GTX 1080 | TITAN X | GTX 980 | R9 Fury X | |
|---|---|---|---|---|---|---|---|---|---|
| GPU | Vega 10 | Vega 10 | Vega 10 | Vega 10 | GP102 | GP104 | GM200 | GM204 | Fiji XT |
| GPU Cores | 4096 | 4096 | 3584 | 4096 | 3584 | 2560 | 3072 | 2048 | 4096 |
| Base Clock | 1406 MHz | 1247 MHz | 1156 MHz | 1382 MHz | 1480 MHz | 1607 MHz | 1000 MHz | 1126 MHz | 1050 MHz |
| Boost Clock | 1677 MHz | 1546 MHz | 1471 MHz | 1600 MHz | 1582 MHz | 1733 MHz | 1089 MHz | 1216 MHz | – |
| Texture Units | 256 | 256 | 224 | 256 | 224 | 160 | 192 | 128 | 256 |
| ROP Units | 64 | 64 | 64 | 64 | 88 | 64 | 96 | 64 | 64 |
| Memory | 8GB | 8GB | 8GB | 16GB | 11GB | 8GB | 12GB | 4GB | 4GB |
| Memory Clock | 1890 MHz | 1890 MHz | 1600 MHz | 1890 MHz | 11000 MHz | 10000 MHz | 7000 MHz | 7000 MHz | 1000 MHz |
| Memory Interface | 2048-bit HBM2 | 2048-bit HBM2 | 2048-bit HBM2 | 2048-bit HBM2 | 352-bit G5X | 256-bit G5X | 384-bit | 256-bit | 4096-bit (HBM) |
| Memory Bandwidth | 484 GB/s | 484 GB/s | 410 GB/s | 484 GB/s | 484 GB/s | 320 GB/s | 336 GB/s | 224 GB/s | 512 GB/s |
| TDP | 345 watts | 295 watts | 210 watts | 300 watts | 250 watts | 180 watts | 250 watts | 165 watts | 275 watts |
| Peak Compute | 13.7 TFLOPS | 12.6 TFLOPS | 10.5 TFLOPS | 13.1 TFLOPS | 10.6 TFLOPS | 8.2 TFLOPS | 6.14 TFLOPS | 4.61 TFLOPS | 8.60 TFLOPS |
| Transistor Count | 12.5B | 12.5B | 12.5B | 12.5B | 12.0B | 7.2B | 8.0B | 5.2B | 8.9B |
| Process Tech | 14nm | 14nm | 14nm | 14nm | 16nm | 16nm | 28nm | 28nm | 28nm |
| MSRP (current) | $699 | $499 | $399 | $999 | $699 | $599 | $999 | $499 | $649 |
If you are a frequent reader of PC Perspective, you have already seen our reviews of the Vega Frontier Edition air cooled and liquid cards, so some of this is going to look very familiar. Looking at the Vega 64 first, we need to define the biggest change to the performance ratings of RX and FE versions of the Vega architecture. When we listed the “boost clock” of the Vega FE cards, and really any Radeon cards previous to RX Vega, we were referring the maximum clock speed of the card in its out of box state. This was counter to the method that NVIDIA used for its “boost clock” rating that pointed towards a “typical” clock speed that the card would run at in a gaming workload. Essentially, the NVIDIA method was giving consumers a more realistic look at how fast the card would be running while AMD was marketing the theoretical peak with perfect thermals, perfect workloads. This, to be clear, never happened.
With the RX Vega cards and their specifications, the “boost clock” is now a typical clock rate. AMD has told me that this is what they estimate the average clock speed of the card will be during a typical gaming workload with a typical thermal and system design. This is great news! It means that gamers will have a more realistic indication of performance, both theoretical and expected, and the listings on the retailers and partner sites will be accurate. It also means that just looking at the spec table above will give you an impression that the performance gap between Vega FE and RX Vega is smaller than it will be in testing. (This is, of course, if AMD’s claims are true; I haven’t tested it myself yet.)
The air-cooled Vega 64, in both Limited Edition and standard versions, will have a base clock of 1247 MHz and a Boost clock of 1546 MHz. The base clock is more than 100 MHz lower than the Vega FE air cooled card, which is troubling perhaps, but the Boost clock looks like it is 54 MHz lower than the Frontier Edition. However, if AMD’s move to a typical/average clock rating for RX Vega is true, that will be HIGHER than the 1440 MHz average clock rate that we observed with the Vega FE card last month. That ~100 MHz could give the RX Vega a performance advantage of 7-8% over the Vega Frontier Edition.
- Radeon RX Vega 64 – $499 – Amazon.com
- Radeon RX Vega 64 Limited Edition – Amazon.com
- Radeon RX Vega 64 Liquid Edition – Amazon.com
- Radeon RX Vega 56 – $399
The liquid cooled RX Vega, which will not be a limited edition according to AMD, has a base clock of 1406 MHz (much better than the air-cooled card, as expected) but a boost clock of 1677 MHz. If that lives up to the claims, it is significant, as it would mean it would be 16-17% higher than the performance of our first air-cooled Vega FE testing. That would give it enough of a performance boost to push up past the GeForce GTX 1080 in nearly all comparisons we performed.
Memory speed and bandwidth for the HBM2 implementation are the same at 1.89 GHz and 484 GB/s of total bandwidth. The RX Vega product family will have 8GB of memory implemented, half of the Vega Frontier Edition. NVIDIA’s GeForce GTX 1080 has 8GB of memory though the 1080 Ti has 11GB. Clearly, the high cost of HBM2 memory went into this decision and though AMD will be touting the benefits of the HBCC and its capability to address other memory more easily, that feature will need work with developers to be properly implemented and to impact gaming performance.
The power draw of the Vega FE cards was telling for the results we see on RX Vega. The Vega 64 liquid cooled will hit 345 watts, the air-cooled version hits 295 watts, and the Vega 56 comes in at a cool 210 watts. These are power hungry cards and getting these clocks necessary to take on the GTX 1080 have pushed the Vega design outside its most efficient area.
The Vega 56 might be the most interesting product on this slide though – it was a surprise release and we don’t have any relatable experience from the likes of a professional card to lean on. With a ~15% drop in shader count and lower clock speed, this product will be at a noticeably lower performance level than the Vega 64. The HBM2 memory is running at 1.6 GHz rather than 1.89 GHz, bringing memory bandwidth down to 410 GB/s. With a price of $399, it will be taking on the incredibly popular GeForce GTX 1070.
The clock speeds and power rating of the RX Vega 56 indicate that it will be a massive overclocker. You should not be surprised if gamers can push the RX Vega up to the same clock speeds, or higher, as the RX Vega 64, getting a massive performance boost. Obviously, that will increase your power consumption up to near Vega 64 levels, but if the cooler can keep up with it, we should have no issues.
Pricing is going to be an interesting discussion. The base price of the RX Vega 64 is $499 and the Vega 56 is $399. That positions the Vega 64 against the GTX 1080 (in a world where prices are back to normal in the GPU market) and the Vega 56 against the GTX 1070. The Limited-Edition Vega 64 will be $599 and the liquid-cooled Vega 64 will be $699 but come with the option of Radeon Packs. What are those you ask?
A Pricing Pickle – Radeon Packs
One of the big things that AMD has been trying to address, much at the behest of the gaming community, is how to get graphics cards in the hands of gamers rather than miners that drive up prices. Though a sale is a sale to both AMD and NVIDIA, there is a danger to the long-term vitality of the PC gaming market if these trends continue. I wrote about these risks on MarketWatch back in June, and the risks still remain real. AMD would also like to get users more invested in the other hardware ecosystems that promoted the Radeon brand including FreeSync displays and even the Ryzen processor and motherboard platform. AMD built the Radeon Packs idea as a way to hit both angles.
The simplest way to understand the packs is that gamers that buy the RX Vega 64 in a Radeon Pack will receive two free games (that differ by region but are Wolfenstein II and Prey in the US), a $200 discount on a 3440×1440 100 Hz FreeSync display and a $100 discount on a Ryzen CPU + motherboard bundle. The purchase of the display and the CPU+MB are optional, and are not required to complete the sale, but the discounts are only offered at the time of purchase. The games will come with the card purchase, regardless.
The RX Vega 64 will have two packs available: Black and Aqua. The Aqua Pack is for the liquid-cooled version of the RX Vega 64 and prices the card itself at $699. The Black Pack is for the RX Vega 64 air-cooled card and comes with a $599 price tag and covers both the Limited-Edition card and the standard black shroud card. Finally, the Red Pack is for the RX Vega 56 air-cooled card and is $499.
You’re probably wondering about the supposed price increase on the RX Vega 56 and the RX Vega 64 standard card; they are both priced $100 higher than the stand-alone units. First, this is the only way to get the Limited-Edition RX Vega 64 card and it will not be sold at $499 individually. Once the limited-edition cards are sold out, the standard card will replace it in the Black Pack. The liquid-cooled card is also only going to be available for $699 and in the Radeon Aqua Pack.
For that added $100 you are getting two free games and $300 in potential discounts on other hardware. If you already have a platform you are happy with, or a monitor you like, and don’t want to upgrade them, you do not have to take advantage of those discounts. Instead, you will be paying $100 more for the fancier card and the two bundled PC titles.
AMD’s intent is two-fold. First, they want to get more users to upgrade to FreeSync displays (and Ryzen systems of course) as it helps them for the relative comparisons to NVIDIA hardware today and incentivizes gamers to stay in the Radeon ecosystem. AMD also hopes that this helps to deter the miners from buying these cards because of their higher price and bundle complications. I personally don’t feel that miners will be deterred by any price change that wouldn’t also scare away all PC gamers and leave AMD in a nearly impossible spot. The only thing that COULD have hindered the purchase of cards by cryptocurrency miners is if AMD required these packs to be purchased with the discounted hardware, but that would be unfair to the gamers that might have already invested in the AMD ecosystem.
Another side benefit that might be occurring with this bundle system is that it helps resellers like Newegg refocus on the gaming customer. Selling a graphics card is typically a low margin sale and resellers care very little if the card ends up in the hands of a dedicated PC gamer itching to play Prey or a miner putting it on a shelf to earn Ethereum 24 hours a day. But if you can tie other, higher margin products to the sale of the RX Vega, that raises ASPs (average selling prices) and could give Newegg/Amazon/etc. a reason to target and hold product back for these types of sales.
Some Sexy New Hardware
For our review today, we are going to be looking at all three of the new RX Vega derivatives: the RX Vega 64 air-cooled (which I will just call the Vega 64 going forward), the RX Vega 64 liquid-cooled card (which I will refer to as the Vega 64 Liquid), and the RX Vega 56.
The RX Vega 64 is a blower-style two-slot graphics card that shares a lot of design characteristics with previous Radeon graphics solution with a red/black shroud and red-blazoned Radeon logo. There is a limited edition of this card that will be available today as well, and though it looks much more stylish with the brushed aluminum design, it performs and functions identically to card we are looking at.
The reference card has the same PCB as the Vega Frontier Edition products we reviewed previously, as do all the Vega offerings that have seen thus far.
A back plate protects the rear of the PCB though it does not make any thermal contact with the unit to aid cooling as far as we can tell.
The Radeon logo along the top of the card illuminates when powered on.
Both the Vega 64 and the Vega 56 will require a pair of 8-pin power connections to work and based on our power testing you’ll find on the next page, these guys will utilize them both.
Behind the ATX power connections is the GPU Tach, a set of LEDs that light up as the GPU load increases, like a car tachometer. In my experience, this either sat with 1 light lit at idle or nearly all lights on during gaming sessions. I’m not sure what a “half way” GPU Tach workload would be, but it adds to the style nicely when the rest of the card will be facing down and away from most case windows. The dip switches on the card allow you to disable them or change been red and blue LEDs. (Odd, no green?)
Display outputs consist of three full-size DisplayPorts and a full-size HDMI, matching other recent graphics card releases.
Note that the Vega 64 and the Vega 56 share the exact same physical design so there is no need to show it to you individually.
The Radeon RX Vega 64 Liquid uses the same PCB and GPU design as the Vega 64, but replaces the blower style air cooler with the same liquid cooler used on the Vega Frontier Edition liquid cooled card. Instead of the gold+blue design of it, the RX Vega 64 Liquid uses a brushed silver shroud with red light accents. The design is incredibly high quality with fully extruded aluminum encasing.
Removing the shroud reveals the exact same cooler design that we found on the Vega Frontier Edition. This is a diaphragm-style pump with an expansion tank.
Along the back you’ll find the same dual 8-pin power connections as the air-cooled Vega 64 but the design adds the red ‘R’ voxel/pixel that lights up. It has no other function than looks, but I must admit to being smitten with it.
The Radeon logo along the top of the card also lights up red though the Vega ‘V’ logo does not.
AMD has included the same 120mm radiator and fan integration we saw on the Vega Frontier Edition. With extended use and just a bit overclocking, this radiator will warm to the touch.
The back plate on this card is also just for looks, but it does look damn good, with the same brushed aluminum style that wraps the rest of the card.
Again, the same GPU tech exists on the RX Vega 64 Liquid as does on the standard card, with the same capabilities.
Display outputs also remain the same: three DisplayPort and one HDMI.
Overall, the family of Radeon RX Vega cards is a combination of old and new. While I do wish that AMD had committed to higher quantities of the limited-edition air-cooled card with the same aluminum body, they will instead depend on partners to release custom cooled models in the September time frame to make up for it.
So what is the mining
So what is the mining performance?
It’s not good as of today.
It's not good as of today. ~35 MH/s.
did you manage to find a new
did you manage to find a new optimized mining code, given mining code it typically adjusted for each GPU and given there are new instruction sets just for hashing on this card using a code compiled for RX580 is not a good benchmark?
we used Claymore Dual 9.8
we used Claymore Dual 9.8 which is optimized for Vega, as well as ethminer
That is really good news..
That is really good news.. hope it stays this low. giving us gamers and blender(GPGPU) users a chance to actually get one …
This card is a dog. More than
This card is a dog. More than a year late. A power guzzling POS. Only an AMD fankid would buy this trash.
Or someone like me who has a
Or someone like me who has a freesync monitor.
Yeah AMD zealots will still
Yeah AMD zealots will still defend it saying incomplete drivers and etc. They had over a year longer to come out with product that is over clocked near maximum (to just match 1080 more or less). Definitely way beyond the efficiency of the architecture. It needed HBM2 or consumption would have been worse or maybe couldn’t have matched 1080 even.
Could have delayed things a little longer and polished it but crap (GCN is outdated) doesn’t shine up very well. LOL
You are a rather Daft one not
You are a rather Daft one not to see that even Nvidia’s SM/shader based sorts of designs have not changed much generation to generation like AMD’s GCN keeps getting it’s share of improvements. What AMD lacks currently is the funding to come up with a mobile first sort of base GPU micro-arch design like Nvidia bagan with Maxwell’s mobile first design that is different from Nvidia’s compute/professional designs. AND lacked the funding for 2 seperate base designs and nobody can not change that fact.
So Nvidia’s basic SM/shader designs are similar to AMD’s GCN/NCU sort of shader designs in that tweaks/featres are added each “generation”. Just because Nvidia’s marketing uses an entirely different famous scientist’s name does not imply any radically new GPU micro-arch simply with using a completely different name for each “new” generation.
And the GTX 1080Ti’s 88 ROPs are directly responsible for those higher FPS metrics compared to AMD compute hevy Vega 64 design. So looking at Vega 56’s 210 Watts
and 64 ROPs(same number as the GTX 1080’s 64 ROPs/and Vega 64’s 64 ROPs) should tell you that if Vega 56 had just a little more compute stripped out that maybe it could be clocked higher and get damn close to the 1080’s performnace metrics. Vega 56 with some AIO water cooling may do just that and get Vega 56 damn near the GTX 1080 in performance.
GCN/shader design is not Outdated, it’s just a name, same for Nvidia and its Shader designs. And all AMD really needs to do with Vega is keep up with removing any excess compute out to have a “Mobile First” design, because that is all Nvidia really did with Maxwell.
AMD’s Vega GPUs are power hungry because Shaders/cores are power hungry animals, same for Nvidia’s shader cores. And to make a shader core able to run with those higher clocks requires deeper ALU/INT/FP piplines so boht AMD and Nvidia have to use more transistors in their high clock speed GPU designs.
GCN has nothing to do with shader counts or shader counts relative to ROPs/TMUs and that’s where Nvidia gets its lead by stripping out shader cores and keeping the ROP/TMU to total shader cores ratios in check by providing just enough shaders to serve the ROP/TMU flow and no more than that.
Vega is not a gaming only GPU design it’s a compute design also. And some of that extra compute helps with DX12/Vulkan games that are designed to make use of the extra shaders resources on Vega. It helps with coin mining also and that’s a GPU sold for whatever reason towards AMD’s increasing revenues to match those Zen/Epyc revenues that are about to dwarf AMD’s Gaming Only focused market/revenues is size!
I’m all for GPU compute usage putting gaming only usage for GPUs on the back burner and with DX12/Vulkan that back burner may still net increased gaming performance with that compute becoming necessary for gaming also along side those ROPs/TMUs!
Or someone like me wants open
Or someone like me wants open source driver.
Who cares Nvidia with Nvidia
Who cares Nvidia with Nvidia Founders Tax and Gsync monitor prces still a better value/performance.
Especially since the news is
Especially since the news is now if you want a non package Vega black it will now be $100 more. The original price was limited to launch only and prices increased to new MSRP after an hour when initial supplies ran out.
Not a good value at all now. Fanboys and miners get your wallets out.
As of now with new mining driver from AMD Vega 56 can get up to 41MH in Ether
Vega 56 is the only one worth
Vega 56 is the only one worth buying and only at $399, so it wins for now but no one can actually buy one. 1070s are still well over $400 right?
You should also 100% undervolt Vega to improve performance. Hopefully AIB cards are actually available as well in September, it’s going to make Vega look far better, not sure why AMD still bothers to release terrible blower cards that make their cards look really bad.
No matter the GPU Video
No matter the GPU Video memory size requirements the ethereum mining folks may require for the hashes/hash tables in the future, that Vega HBCC/HBC using the HBM2 as a last level cache IP and 512TB of Virtual memory addressing in Vega will be hard trip up with any hash table sizes larger than the HBM2’s actual VRAM size on any Vega SKUs! Vega can spill over it’s HBM2 cache into/out of main system memory via its HBCC/HBC technology to fill up system DRAM space, and do so at a relatively little degradation in hashing performance. The mining code will love that HBCC/HBC-HBM2 IP in short order, so get Vega while you can stocks will be in short supply.
Who started the rumor that it
Who started the rumor that it was 70-100mh? Probably someone who had AMD stock and wanted to unload it before it tanked. Do we have to wait two years for this to materialize as well? Driver magic. LOL
Not really true, Vega only
Not really true, Vega only needs to get it M-Hashes/Watt metric a little better and undervolting/underclocking will fix that power usage metric PDQ!
They have already been tweaking on Vega FE those Minier 49ers with the software chops to take advantage of Vega’s new IP, and that software development process takes a little time! But, in short order, those mining hash browns will be as delicious as gamer tears!
How do you think the HBCC
How do you think the HBCC will improve performance with new drivers?
Latest drivers 17.10.3
Latest drivers 17.10.3 claymore miner I got average of 46Mh/s on Vega64 water cooled with 145W power draw!
No you did not!
Put on a
No you did not!
Put on a power meter and see 400 watt.
This entire article is a good
This entire article is a good read on Vega’s compute performance, and it shows Vega 64 beating Quadro in some interesting ways and consumer Vega offers (1/16 DP FP compute ratio of its FP32 rate) vs Pascal’s (1/32 DP FP compute ratio of its FP32 rate) compute. And That NVIDIA Quadro P6000 SKU has 24GB of V-memory.
[See the LuxMark chart]
“Well, well, well. Would you look at this? AMD’s Radeon RX Vega 64 falls just behind the Radeon Pro WX 7100 in Cinebench, but it soars past the same GPU in both LuxMark and V-Ray. In fact, I should stress that your eyes don’t deceive you: Vega 64 really does beat the entire lineup in LuxMark. That includes beating out the $5,000 Quadro P6000.”
[See page 2 titled:”Rendering: Autodesk 3ds Max, Cinebench, LuxMark & V-Ray Benchmark”](1)
At the Techgage website concerning mining potential the author states:
“Hot damn. I feel like these kinds of gains are those that AMD should promote out the wazoo. So many reviews posted today are likely to paint a rough picture of this card’s gaming performance, but on the other side of the fence, compute performance on Vega quite simply kicks ass. The results here may be able to give an impression of Vega 64’s future mining performance. Mining benchmarks you’ll see around the web in other launch reviews will show an edge over a top-end NVIDIA card. I would not be surprised if AMD optimizes its driver sometime in the future to vastly improve mining performance. Vega should technically be better than the 30MH/s you’ll see reported today, based on all of the compute performance seen here.”(1)
[Go to the page titled:”Sandra: Cryptography, Science, Finance & Bandwidth that is on page 5 of the article}
(1)
“A Look At AMD’s Radeon RX Vega 64 Workstation & Compute Performance”
“by Rob Williams on August 14, 2017 in Graphics & Displays”
https://techgage.com/article/a-look-at-amds-radeon-rx-vega-64-workstation-compute-performance/5/
All that hype and waiting was
All that hype and waiting was for THIS? Gamers have had this performance (or faster) for more than a year already. I can’t wait to see how long it will take the AMD fanboyz to start up with “Just wait until Navi!”
Well this is not good for the
Well this is not good for the GPU market, it means Nvidia don’t have to cut prices or release new cards.
+1
I think this is what
+1
I think this is what frustrates me the most. Nvidia won’t be forced to compete because AMD failed.
I guess expecting AMD to be competitive with both Intel and Nvidia is too much to ask…
Go look at Vega’s compute
Go look at Vega’s compute benchmarking against even a Titan XP, with the Titan XP even having more VRAM, and you can see where your worried assesment is wrong. Gaming is not going to be AND’s most profitable usage for the Vega GPU Micro-Arch, and AMD’s Epyc sales/revenues will eclipse AMD’s gaming only sales in short order. [see the Techgage website refrences in some of the posts on this page for some interesting Vega 64 compute benchmarking results against even the Quadro P6000-24GB SKU]
AMD’s does not even need to outright beat Nvidia in the top end DX11 gaming metrics as AMD’s mainstream Polaris and now high end Vega SKUs are really showing their form in the DX12/Vulkan optimized titles. But gaming markets are not really as interesting as the CPU/GPU compute and AI markets where no matter HBM2’s current costs those SKUs will command the highest markups and that professional market will get the Cherry Picked Vega Dies with the best thermal metrics.
You should also note that Those Cherry Picked Vega 10 dies will also be clocked lower in the professional GPU SKUs and that Vega 10 will do very nicely on that compute performance/watt metric what with Vega 10’s massive amounts of Shaders providing plenty of higher DP FP and SP FP MFlops/GFlops, and for Vega 16/8 FP GFlops also! Pascal’s DP FP rate in only 1/32 its SP FP rate while Vega 10’s DP FP rate is 1/16 its SP FP rate!
AMD’s management told Raja that they needed a Vega Compute focused GPU micro-arch to complement AMD’s Zen/Epyc CPU micro-arch and that’s waht Raja produced, and Vega games pretty well also. AMD’s still stands to gain performance on its Vega SKUs now that The Vulkan/DX12 graphics APIs are being used for more and more gaming titles. So look for Nvidia to be left, like Intel, scrambling. Only in Nvidia’s case, it will be to get that compute ability back into its consumer GPU SKUs. That AMD/Nvidia Async-Compute debate will be raging again with each and every new fully DX12/Vulkan optimized title that becomes available.
The only market where AMD needs to maybe create a mobile first type of GPU/Maxwell like GPU micro-arch is for mobile/laptop market where battery power needs conserving and thermal headroom is lacking. But AMD’s Epyc revnues will give AMD enough by themselves to fund AMD/RTG’s creation of a more enegry efficient Vega/mobile GPU micro-arch variant.
Nvida’s JHH is about to be able to raise its prices as AMD appears to be wanting to set a range of wholesale prices its carges its Retaile Channel partners. This is so AMD can benifit more from the spot demand pricing that the retailers have been mostly profiting from at AMD’s expense! So Nvidia’s prices are going up as are AMD’s prices until the miners have their needs met! And AMD has R&D and other costs to amortize as quickly as possible so AMD will have more pricing latitude later when GPU demand pricing relaxes and GPU prices fall.
to the contrary, it’s very
to the contrary, it’s very good for the market because it will be the 1st time in a very long time that GPUs will have a larger life span, and that’s very good for the consumer considering the price premiums we had to pay for Pascal
also, nVidia expressed no intention to launch consumer Volta this year – taken from the recent earnings call
Question for people who were
Question for people who were around during this time: Was it really better in the late 90’s when the new GPU that you bought would be outdated and obsolete by the years end, or now, when a three year old GPU is still regarded as high end and quite capable.
Certainly as a reviewer, having new stuff out every 6 months that dwarfs everything out previously is great and a lot of fun, but I’d imagine that to a gamer, it’s frustrating.
It’s much better now for the
It’s much better now for the consumer when you buy something and it lasts a long time.
I got a 7970 back at the early part of 2012(reference design card) and ran it over 4 years for 1080p gaming until I got the funds to get a 1440p, 144hz VRR monitor and compatible GPU. Otherwise I’d still be running it, still a great card for 1080p gaming.
Exactly this, now NVIDIA
Exactly this, now NVIDIA themselves are saying they probably wont be releasing Volta for gamers until late 2018 (unless AMD drivers can drastically improve Vega of course). If we all thought Vega was late, lol just wait for Volta, NVIDIA will rewrite the book on being late with their next architecture. No sense at all dumping millions of dollars into a new GTX architecture that you don’t even need. Thanks to AMD’s lack of extreme performance and high cost power consumption NVIDIA can keep on asking multi thousands of dollars for their Pro Volta cards and keep the aging 10XX series around for much much longer. Not only that think about this, what if NVIDIA decides to just drop the process node down on their current Pascal, say down to 12nm or 10nm and keep the same prices? LOL, NVIDIA knows how to win and knows how to keep the market anti-consumer as long as they want. AMD is being pushed into being anti-consumer and not only that it looks to me like AMD is pushing compute performance over anything else (good for them actually). If it weren’t for the fact we need to limit anti-consumer practices I would be purchasing the 1080, but I can’t do that knowing AMD is the one who needs the money more in order to try and keep competition strong. It’s all about having competition in the market space. Without competition between companies, we are all going to be BIG time losers. That absolutely can not be allowed, PERIOD.
Here is the old Volta road map (before it was changed) which says 2016, lol. Now its going to be late 2018, especially for us gamers. All thanks to AMD not having enough money to pull off a decent Vega architecture for gaming. http://cdn.wccftech.com/wp-content/uploads/2017/04/05791082-photo-nvidia-gpu-roadmap-2013-740×493.jpg
Not long at all look at a
Not long at all look at a post below. Wait for Navi. LOL
Vega leaks showed it to be an undeperformer compared to Pascal. No big surprise here.
There is no disappointment in all of AMD land for there are plenty of fanboys that will still buy one of these.
Hurry up and buy RX Vega now at today’s prices (as much as Nvidia cards) but less performance per watt(over 33%). That fine wine drivers might have you there in 2 years but by then the card will be 50% off of retail.
If you got to have Vega wait a few months until they rebate it some.
Most people don’t buy the
Most people don’t buy the 1080 Ti anyway. So, guess what? You have a situation where gamers can now choose from competitive products in all ranges below the high-end 1080 Ti.
They did not have that choice before today.
Prices will probably go down as a result of competition.
Competition is almost *never* a bad thing.
Gold Award? Really? Is that
Gold Award? Really? Is that the PC Perspective participation trophy?
Newegg is already charging $100 more than MSRP.
Complete effing fail.
I can’t believe I actually considered replacing my 1080.
The Vega 56 got the award,
The Vega 56 got the award, and nothing has been IN stock yet!
https://www.newegg.com/Produc
https://www.newegg.com/Product/Product.aspx?Item=N82E16814202300&cm_re=rx_vega-_-14-202-300-_-Product
64’s in stock as of posting…. $599.99 (Read in Josh doing the voice thing)
Will be interesting to see how much markup is on 56’s when they release later this month.
Those $599 are the Packs (see
Those $599 are the Packs (see the free games listed)
Its $499 without the Packs Here
Those Packs will not matter
Those Packs will not matter to those MAD Hatter Miners, so look on eBay for those very Packs’ parts not needed for mining to go up for bid. There are even Nvidia fans looking to get a Vega 64 SKU to sell that at enough markup to the Mad Hatter Miners and afford the hardened Nvidia Fan a GTX 1080Ti on that little bit of wheeling and dealing action on some Vega Mad markup individual Vega sales to the Mad Hatter Miners.
It’s a Mad Mad Mining dash in that MAD MAD GPU coin mining world that may even net Nvidia Itself some extra indirect sales via some Vega back room markup deals from some Nvidia fans with their Vega purchasing reservations already in place in order to make enough monty off of Vaga to get their GTX 1080Ti fix filled for less money in that sort Mad Hatter Mining market dealings.
Newegg is showing Out of
Newegg is showing Out of Stock for all but one part. Did they already sell out, or they just haven’t received yet?
They sold out. Probably
They sold out. Probably within a minute, but definitely after 5 minutes.
What are your thoughts on a
What are your thoughts on a dual RX Vega 56 on one PCIe card variant with maybe not as high a core clock, and the two RX Vega dies communicationg via the Infinity Fabric protocol rather than and PCIe protocol method. That Infinity Fabric on the Zen/CPU cores appears to support more cache coherency capability among the Zen/CPU processor cores across dies/CCX units. So the Infinity Fabric and 2 Vega 56 dies on the same PCIe card or on a Dual GPU MCM type of arrangement and the GPUs having a more unified HBC/HBM2 cache sharing arrangement managed by the GPU die’s HBCC controllers able to link up logically via the Infinity Fabric IP.
So, Vega 56 gets the
So, Vega 56 gets the gold…
What does Vega 64 get? Bronze? Did it even make it to the medal rounds?
I call bullshit on your post.
I call bullshit on your post. What were you going to replace your 1080 with, (stop it with making Freud seem correct)? There is no serious reason to consider replacing a card that good.
Vega 56 is complete trash.
Vega 56 is complete trash. 300W power consumption and terrible performance. Wonder how much AMD had to pay to get that award.
Vega 56 used around 200w.
Vega 56 used around 200w. Might want to get your eyes checked.. I suppose the 1070 also has terrible performace then..
Is the Vega 56 overclock
Is the Vega 56 overclock being artificially held back? I don’t understand why it can’t at least match the average clocks of the Vega 64 (air). It seems, if anything, it would clock a bit higher than the 64. Was AMD worried about their $400 part being within 3% or so of their $500 part when overclocked?
Ryan, if you have the time, please do more Vega 56 overclock testing, especially once partner boards are available!
Answering my own question
Answering my own question after reading the answer elsewhere. Apparently the 56 is hard limited in the bios to 300W draw. If that limit could be removed (hopefully in partner boards) we could be looking at 1080 performance for $400.
At more than 300W 🙂
At more than 300W 🙂
Ugh, so much for newegg and
Ugh, so much for newegg and amazon selling 64 air at 499.
literally no reason to buy at their 599 price.
So Newegg Canada has Vega RX
So Newegg Canada has Vega RX 64 priced $200-300 CAD above AIB GTX 1080. AMD likes to talk a good game about prioritizing gamers above miners but they would sell effectively 0 cards (in Canada at least) when they are priced like 1080Ti if it weren’t for miners.
The 6 in the Witcher 3
The 6 in the Witcher 3 results page at the end of that page says Vega is 6% faster than 1070 doesn’t have the percentage sign.
Your review will now 0.000000000001% better once it’s fixed! 😛
LOL will fix!
LOL will fix!
https://media.giphy.com/media
https://media.giphy.com/media/26n6Ocpiegpl3wXCg/giphy.gif
The comparison table on the
The comparison table on the first page has a the wrong memory bandwidth figure listed for Vega 56.
Fixed! 409.6 GB/s it is.
Fixed! 409.6 GB/s it is.
Price performance leadership
Price performance leadership are you serious? The street price for Vega 56 will be much higher and if we go by MRSP then the GTX 1070 wins since it is $349 vs $399. Add the higher energy costs to that and you get a mediocre price performance ratio for gaming.
Provide a link to a single
Provide a link to a single $350 GTX 1070 available and in stock on Amazon.
“Provide a link to a single
“Provide a link to a single $399 Vega 56 available and in stock on Amazon.”
FIFY
Show me a link to buy an
Show me a link to buy an Amazonian Loch Ness monster for tree fiddy.
Mining has inflated 1070
Mining has inflated 1070 prices. However you can get a 1080 cheaper than a 1070 and Vega 64 now. $540
https://www.amazon.com/Gigabyte-GeForce-Graphic-Computer-Graphics/dp/B0748KG9CK/ref=sr_1_4?s=electronics&ie=UTF8&qid=1502739021&sr=1-4&keywords=1080
Amazon is disappointing, but
Amazon is disappointing, but I did find one on ebay for $3.50 at time of writing this comment:
https://www.ebay.com/p/Water-Horse-Crusoe-Loch-Ness-Monster-Toy-Plush-6-Beverly-Hills-Teddy-Bear-2007/1040566377?iid=122041109914
it’s probably the worst GPU
it’s probably the worst GPU release from AMD in a long time.
how can a 480mm² die get a Golden Award for barely matching a competitor with a die 170mm² smaller, a year later.
raja needs to put down the bottle and the cigare, vega was 100% his design afterall, i wonder if he will survive this, because imo the board should fire him.
I seriously doubt it’s
I seriously doubt it’s anything but a redesign of Fiji and that’s a redesign of Hawaii – GCN finally showing its limits
Look at the anandtech review,
Look at the anandtech review, a lot of work was put into getting higher clock speeds (hence the much higher clocks than Polaris)
this was not a simple die
this was not a simple die shrink of fiji, they changed alot of stuff, important stuff that should yield significantly more performance than fiji, but they ended up with worst perf.
the power draw i get it, from polaris and ryzen, it was clear that glofo’s node isn’t up to the task for high frequency, where it just sky rockets when it reachs a certain point.
but what i do not get is the GPU core efficiency, it doesn’t scale as itn should, it’s like FIJI all over again, like if AMD doesn’t know how to optimize workloads passed 2.5k stream processors.
GPUs are very sensitive to
GPUs are very sensitive to software optimization. If we had DX12 based games (designed from the ground up, not hacked in) with proper optimization for Vega, I suspect Vega cards would outperform the Nvidia competition by a significant margin. Unfortunately, optimization will be targeted at the current market leader, helping them maintain a near monopoly position. Nvidia has always seemed to have a policy of there will be another GPU in 6 months, so they are not very forward looking and they don’t care how their cards perform in a year or two. In fact, it is better if their cards perform worse in a year or two so that loyal Nvidia customers will buy a new Nvidia card to replace the obsolete one. Given the stagnation in process tech, I suspect that people will be keeping their cards a lot longer than previous generations. The move back to 16-bit could be a big thing that Nvidia cards don’t have yet (great for planned obsolescence). This doesn’t help AMD sell cards now unfortunately. It looks like the trolls are out more in force for the GPU release than they are for the cpu release though. Interesting. Maybe they need to kill it quick before anyone optimizes software for it. I personally would probably buy a Vega 56 if it wasn’t for the inflated prices.
Please, take into account the
Please, take into account the area of memory dies.
it is 484mm² for the gpu die
it is 484mm² for the gpu die alone not including the hbm2 stack, that is why it’s so weird to see this kind of performance.
If I may be constructively
If I may be constructively critical, the average FPS performance summary tables are all very confusing. Alternating between V64 Air, V64 Water, V56, and the cards compared against within each table change.
I suggest consolidating these summary tables into one giant table on its own page as an overall performance summary.
Sorry for my english level, I
Sorry for my english level, I am novice level.
Ryan Shrout and Brian Maltavino please answer my query, if this information is already aware, Thank you forward:
1. I want to know, does vega CARDS have Draw Stream Rasterizer enabled on this driver version?
2. Has Primitive Shader features are enabled?
3. I want to estimate potential improvement of card, thank you for all your hard working on review and in depth analysis on technology.
I will purchase via link when card gets stock.
***edited by Jeremy ***
Start
***edited by Jeremy ***
Start acting like you are old enough to have been potty trained, the bunch of you.
“the bunch of
“the bunch of you.”
Seriously?
Yes, the thread I killed was
Yes, the thread I killed was worse than the usual crap spewed out in the comments.
It looks like they will be
It looks like they will be competing on price until drivers improve unless someone is interested in 4K gaming in which case the Vega64 is the clear leader. If the Vega 56 is actually available for purchase in September and not sold out to miners then that will be the one I go with to replace a GTX 970.
Vega 64 does not lead in 4k
Vega 64 does not lead in 4k gaming. That crown is held by Nvidia’s 1080ti. 1080 is not for 4k gaming really so by extension neither is Vega 64. These cards are best at 1440 resolution. It’s why the subjective tests they ran while showing off Vega 64 against a 1080 ran in 1440. Look up a few reviews that show 4k results along with frame times. Good luck with your Vega 56. Still not as good as a 1070 in performance per watt but much better than Vega 64 is.
With 88 ROPs on the 1080Ti
With 88 ROPs on the 1080Ti over Vega 64’s, only 64 ROPs that 4K gaming lead should belong to the GTX 1080Ti. Now with more DX12/Vulkan gaming and maybe more gaming features depending on mre Vega’s extra compute things may age differently.
And no one should focus on DX11 in the future more than say currently they are, to an excess, as DX12 and Vulkan are in the process of taking over. So will the games developers be able to begin to make more use of any extra compute for gaming effects over the next year and longer before the Navi and Volta SKUs arrive.
There is also the question of FP16 and more AAA games taking advantage of FP16 in order to do more at 4K.
No resonable person expects the RX vega 64 to beat the GTX 1080 TI with that 88 ROP lead over any of the current Vega 64 and 56 offerings, but Nvidia lacks as much available compute compared to AMD, and currently the miners may like that more!
But the games makers are Eyeing GPU compute and they are also Eyeing Vega’s HBCC/HBC IP and that IP’s ability to make available 521TB’s of addressable texture/mesh data space available via Vega’s HBCC/HBC using the HBM2 as a last level cache to leverage all of the PC system’s regular DRAM and NVM memory paging pools for an effective VRAM size of 512TB of effectve VRAM addressing space for Textures/Mesh Data on any Vega Micro-Arch based discrete GPU SKU.
This Vega HBCC/HBC and HBM2 IP and the FP16 bit IP(rapid packed math) and the New DX12/Vulkan API’s able to do more gaming effects with a GPUs available compute. And even the Vega GPU Micro-Arch’s support for the Infinity Fabric and the potential for Maybe a Dual Vega 56 on one PCI variant that can use the Infinity Fabric IP to make those 2 Vega 56 dies act as one bigger die logically, similar to the way the Infinity Fabric does that for the Zen/CCX unit/Zeppelin dies on TR/Ryzen and Epyc SKUs.
Do we know if AMD has enabled
Do we know if AMD has enabled tile-based rendering, IIRC rumors said it was disabled in early testing/drivers.
I am relatively certain it is
I am relatively certain it is enabled.
“The AMD Radeon RX Vega
“The AMD Radeon RX Vega Review: Vega 64, Vega 64 Liquid, Vega 56 Tested”
Really the title should read: The AMD Radeon RX Vega Preliminary Review(with lots of DX11): Vega 64, Vega 64 Liquid, Vega 56 Tested.
And yes the power metrics on the Vega 64 is going to be up there with the RX Vega 64’s more shader/cores!
RX Vega 64’s:
Shading Units: 4096
TMUs: 256
ROPs: 64
Compute Units: 64
Pixel Rate: 98.94 GPixel/s
Texture Rate:395.8 GTexel/s
Floating-point performance:12,665 GFLOPS
And the RX Vega 56’s:
Shading Units: 3584
TMUs: 224
ROPs: 64
Compute Units: 56
Pixel Rate: 94.14 GPixel/s
Texture Rate: 329.5 GTexel/s
Floating-point performance: 10,544 GFLOPS
and look at the RX Vega 64’s TMU and ROP counts relative to the RX Vega 56’s. If only AMD would design its GPU micro-archs in such a way as to be able to strip out only the excess compute rather than any ROP/TMU units they could match Nvidia’s power usage matrics more closely. That said Vega 56’s remaining TMU/ROP capability is rather close to the Vega 64’s TMU/ROP capability.
The GTX 1080’s:
Shading Units: 2560
TMUs: 160
ROPs: 64
SM Count: 20
Pixel Rate: 110.9 GPixel/s
Texture Rate:277.3 GTexel/s
Floating-point performance: 8,873 GFLOPS
The GTX 1080 Is closer to the RX Vega’s 56’s feature set but Nvidia has stripped out of more compute and clocked its SKUs higher. So if AMD could strip out more of Vega 56’s compute and get the clocks up there then the RX Vega 56 could probably get closer. If AMD could do the same for It’s Vega 64 SKUs in stripping out the compute only then those clocks could be high enough to maybe get the Vega 64 closer into the range of the GTX 1080TI.
The GTX 1080Ti”s:
Shading Units:3584
TMUs: 224
ROPs: 88
SM Count: 28
Pixel Rate: 139.2 GPixel/s
Texture Rate: 354.4 GTexel/s
Floating-point performance: 11,340 GFLOPS
And looking at the GTX 1080Ti’s ROP(88 ROPs) count lead over RX Vega 64(64 ROPs) and 56(64 ROPs), we can see why Nvidia can flip those FPS’s out there. I think that if the RX Vega 56 can be over clocked higher that it may just get closer to the GTX 1080 than then RX Vega 64 will be able to get to the GTX 1080 TI, with the Ti’s 88 ROP lead over RX Vega’s 64’s only 64 ROPs.
And we can see why Nvidia can remain ahead of AMD in the DX11 gaming only benchmarks and it’s those ROPs and less compute getting in the way of the higher clocks for Nvidia and, the jury is still out on Vega’s DX 12 gaming perfotmance across a larger sample of DX12 fully optimized titles.
AMD/Games makers have the usual gaming/driver tweaking to do but more DX12 optimized titles are available this time aroud. AMD with Navi needs to totally decouple it’s compute from its ROP/TMU functionality is such a way that AMD can come up with a base GPU mciro-arch where AMD can fuse off/power gate only compute while leaving in just enough ROP/TMU resources to flip out the frames and win the FPS contest.
That said I’d rather have the extra compute over any FPS overkill above 60-140 FPS low/high end metric for a 4K gaming optimal gaming target with settings turned down. There needs to be more FPS variability gaming benchmarks available also.
The Dual Vega GPUs on a single PCIe card SKUs rumors abound also so maybe a dual RX Vega 56 with less compute and more ROP/TMU capability sounds interesting. AND with AMD’s Infinity Fabric support on is Vega GPU Micor-Arch based products maybe the 2 Vega dies on one PCIe card communicating over the Infinity Fabric can be made to act more like a single monolithic GPU core design to the Graphics API/driver software. That prospect has got me excited!
All those words and basically
All those words and basically said nothing; sad. Someone needs to start a referral service, like “Ask Gary”, but for AMD victims. Professional help may be required before they can move past this over-hyped disappointment, and get on with their lives.
What is an Ask Gary?
What is an Ask Gary?
And yet you are so personally
And yet you are so personally offended by someone showing some actual ROP/TMU figures and doing a little bit of simple inductive and deductive reasoning in attempt at getting at some of Nvidia ROP advantages and Nvidia using less compute/more ROPs to get those Frames flung out there at all costs to win the FPS/power uasge metrics in mostly DX 11 titles. All while there are now some interesting DX12/Vulkan titles results at some other websites and Vega doing good(In DX12/Vulkan) in spite of some ROP disadvantages or clock speed disatvantages to Nvidia’s GTX 1080/1070 SKUs.
And you can only attack the numbers of words or the TechPowerUp data base GPU figures That I used in my comparsion and contrasting the ROP/TMU/Compute resources differences and how that affects the ability to fling the frames out the for those frames flung out for that FPS metric contest.
I personally think AMD did better and will do better focusing on the Professional CPU/GPU markets for the mad revenues that those markets will provide and that AMD’s Vega micro-arch based professional GPU Accelerator/AI Accelerator usage performance is the more necessary for AMD’s growth in revenues over any gaming only performance metrics. And AMD’s Vega SKUs are great in competing with Nvidia’s GTX 1080 and GTX 1070 SKUs for DX12/Vulkan Gaming and not so bad for the DX11 gaming that is on it’s way out.
So that Sad applies more to your taking offense to a little closer look at things for gaming’s future rather that that DX11 only past. And what is really sad is that you can not offer any constructive rebuttal other than expressing your “Sadness”. Now that just leaves the question of Your true Sadness and that is more to do with your lack of abilities for offering any cogent rebuttal to what was posted.
About time dx11 is gone and
About time dx11 is gone and dx12 takes over Vega 64 will be in a dirt nap too. Meaningless to compare future to what is today. Dx11 just works and dx12 requires lazy and overworked programmers to code everything video card related into it. That costs more time and money.
Compute overkill isn’t needed in games I’m sorry to say.
If AMD starts fuzing two Vega cores onto one die I’m going to start buying stock in electric companies and become a millionaire.
Vega has a lot of compute and
Vega has a lot of compute and who the hellz cares about gaming only metrics, the real Epyc money is in the Professional CPU/GPU accelerator/AI markets for AMD, and compute matters more there than any FPS metric!
Gamers can get over themselves they do not matter at all in the larger scheme of things for both AMD’s and Nvidia’s future revenue growth potential!
so which version are we
so which version are we getting for the vega 64 air cooled? the ones with black plastic shroud or the brushed metal limited edition ones?
limited edition is limited,
limited edition is limited, once they are gone then only the black shrouds will be left, that is until the AIB’s come out with their custom coolers, then at that point who cares about the reference designs.
Now someone needs to do 4X
Now someone needs to do 4X Crossfire testing with Vega on Threadripper. I know more than 2X is diminishing returns, or even negative scaling in some cases, but it’s one of the things AMD has that Nvidia doesn’t as of the Pascal generation.
Hi Ryan – any intelligence on
Hi Ryan – any intelligence on availability, looks like the card is only meant for reviewers.. goin by today’s launch 🙂