AMD Vega MegaThread! FAQ and Resources

Page 25 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
AMD Vega 11 GPUs Entering Production, Vega 20 Coming On 7nm
By Khalid Moammer
8 hours ago
AMD-VEGA-20-specifications.jpg

Vega 11 is a mid-range part based on AMD’s new graphics architecture. It’s expected to replace Polaris 10/20 and the RX 580/570 graphics cards and is rumored to feature HBM2 technology rather than GDDR5. Digitimes’ report adds credence to this, as SPIL specializes in packaging HBM2 based GPUs.
Vega 20 is going to be the highest end part from AMD’s Vega family, at least that we know of right now. This GPU will go into future Radeon Instinct accelerators, feature 32GB of HBM2 memory and is believed to feature fully unlocked double precision compute. Vega 20 is aimed squarely at NVIDIA’s Volta V100 accelerator to address the growing AI and machine learning markets.
Click to expand...
FJ6RhSr.png


SPIL obtains packaging orders for AMD Vega 11 GPUs
Julian Ho, Taipei; Jessie Shen, DIGITIMES [Friday 8 September 2017]
AMD will continue to contract Globalfoundries to manufacture its next-generation Vega GPUs, and will have Siliconware Precision Industries (SPIL) as its major packager for the chips, according to industry sources.

The supply of AMD's Vega 10 GPUs has fallen short of demand, thanks to robust demand for artificial intelligence (AI), Bitcoin mining and gaming applications since 2017, the sources said.

Globalfoundries uses its 14nm FinFET process technology to fabricate AMD's Vega 10-series GPUs, and is expected to land orders for the next-generation Vega 11 series, the sources indicated.

Packaging specialist SPIL, which has already obtained orders for AMD's Vega 10-series chips, will continue to hold the majority of backend orders for the Vega 11 series, the sources noted.

The popularity of AMD's and Nvidia's high-end GPUs for applications such as AI has drawn much attention to the importance of packaging technologies for 2.5D/3D solutions, the sources identified. Heterogeneous integration for use in high-bandwidth memory (HBM) devices also requires such advanced packaging techniques.

Taiwan Semiconductor Manufacturing Company (TSMC) with its CoWoS (chip-on-wafer-on-substrate) technology has reportedly secured orders for AI chips from Nvidia and Google. TSMC has further enhanced its advanced packaging capability eyeing a bigger presence in the supercomputer field, the sources said.

AMD's rumored Vega 20 series will be applied to supercomputers, according to the sources. AMD is likely to switch orders to TSMC, which will use its 7nm FinFET process and CoWoS packaging to fabricate the Vega 20-series GPUs, the sources noted.
Click to expand...
 
So...

What is Polaris 20? I know the 580/480 and 570/470 is Polaris 10 and the 560/460 is Polaris 11.

If Vega 11 is replacing Polaris 10, that puzzles me. I would think that it would come in around between the 580 and Vega 56. This leads me to think they are saving it to compete with Volta, as they will need a new level of performance to stay in the game, and 2h 2018 is a long way off in GPU time.

Putting HBM2 in a mid-range card either signals that AMD believes there will be a shift in memory throughput in that segment, or Vega 11 was cheaper to produce without changing the memory architecture for the chip.


If Vega 20 is 2h 2018, what happened to Navi? According to their road-map, Navi was slated for 2018 with "scalability and next gen memory".

Also, if they are looking at 7nm next year, that is ahead of Intel, who is apparently having a heck of a time getting 10nm working for mass production. IIRC, this could be the first time AMD gets ahead in process tech.

I'm also surprised there is no mention of Samsung for production. I know AMD restructured their GF contract, which likely gives them options like TSMC, but I would think AMD would be talking to Samsung.
 


I don't understand ANY of it! They don't want to sell good consumer video cards anymore? They need the HBM2 too keep down temperature and design costs? I'm lost... They don't like us anymore is all I can think of.

I bet that's exactly it, it will be a refresh with HBM2 which will run slightly cooler/better to overclock.
 


AMD's rumored Vega 20 series will be applied to supercomputers, according to the sources. AMD is likely to switch orders to TSMC, which will use its 7nm FinFET process and CoWoS packaging to fabricate the Vega 20-series GPUs, the sources noted.
Click to expand...
 


Vega 11 should be a smaller chip, like Polaris 11. But since Vega is designed to run HBM2, it is likely cheaper to just go with HBM2 in the mid-range instead or redesigning the chip for another memory. As for the keeping temps down, that's possible, but how much extra heat comes from GDDR5(x) as opposed to HBM2. Unless you meant TDP, which using HBM2 would help in that respect and no reason not to take advantage of it since the chip was designed with HBM2 in mind to begin with.

As for consumer vs. professional, there is a lot more money in cards for the fields they mention. It's the same thing nVidia is doing, making pro cards first, then make some tweaks to it for consumers. Basically, we get the imperfect chips.
 


[video="[video="https://www.youtube.com/watch?v=p9ih5vmcDEk&feature=youtu.be&ab_channel=GamersNexus"][/video]
Gamers Nexus talked about HBM2 By Steve Burke Published August 25, 2017 at 4:26 pm
Variations of “HBM2 is expensive” have floated the web since well before Vega’s launch – since Fiji, really, with the first wave of HBM – without many concrete numbers on that expression. AMD isn’t just using HBM2 because it’s “shiny” and sounds good in marketing, but because Vega architecture is bandwidth starved to a point of HBM being necessary. That’s an expensive necessity, unfortunately, and chews away at margins, but AMD really had no choice in the matter. The company’s standalone MSRP structure for Vega 56 positions it competitively with the GTX 1070, carrying comparable performance, memory capacity, and target retail price, assuming things calm down for the entire GPU market at some point. Given HBM2’s higher cost and Vega 56’s bigger die, that leaves little room for AMD to profit when compared to GDDR5 solutions. That’s what we’re exploring today, alongside why AMD had to use HBM2.

There are reasons that AMD went with HBM2, of course – we’ll talk about those later in the content. A lot of folks have asked why AMD can’t “just” use GDDR5 with Vega instead of HBM2, thinking that you just swap modules, but there are complications that make this impossible without a redesign of the memory controller. Vega is also bandwidth-starved to a point of complication, which we’ll walk through momentarily.

Let’s start with prices, then talk architectural requirements.

AMD’s Long-Term Play & Immediate Risk
AMD’s pricing structure for Vega uniquely leans on bundle packs to help improve the company’s value argument in a competitive market. MSRP is $400 on RX Vega 56, $500 on RX Vega 64, and an added $100 upcharge in exchange for two games and some instant discounts. AMD’s intention with this is to offer greater value to gamers, but clearly also will help the company increase margins and move more Ryzen parts, thereby recouping potentially low or negative margins on Vega. This is aided particularly with game bundles, where AIB partners pay AMD about $29 for the game codes, though that is often waived or offered in exchange for MDF. AMD also stated desire to stave off some mining purchases with increased bundle prices, as this would offset the value proposition of the card. Since the bundles are sold as standalone SKUs and can’t be broken by consumers (into parts), it seems that this is potentially an effective solution at keeping miners at bay.

This move is also mystified by complex economics surrounding potential long-term plays by AMD, with the company having already lost on a bet that HBM2 pricing would be cheaper by the time Vega rolled-out; in fact, price increased at least once within the past year, and Hynix ultimately failed to deliver on AMD’s demands in a timely fashion. In the meantime, AMD could have its outlook set on increased supply driving down cost to build Vega GPUs. This might mean taking a slight loss or running on slim margins for now, but hoping for a payoff down the line. Part of this hinges on Hynix potentially coming online with HBM2 at some point, which could help reduce the cost figures we’ll go over in this video. Ultimately, AMD also needs to reclaim gaming marketshare, and part of that reclamation process will be aided by gritting teeth through painfully slim margins or even losses at launch.
The Cost of HBM2
There are two major costs with a video card: The GPU die and the memory, with follow-up costs comprised of the VRM and, to a lesser extent, the cooler.

Let’s start with HBM2 and interposer pricing, as that’s what we’re most confident in. Speaking with David Kanter of Real World Tech, the analyst who broke news on Maxwell’s tile-based rasterization and who previously worked at Microprocessor Report, we received the following estimate: “The HBM2 memory is probably around $150, and the interposer and packaging should be $25.” We later compared this estimate with early rumors of HBM2 pricing and word from four vendors who spoke with GamersNexus independently, all of which were within $5-$10 of each other and Kanter’s estimate. This gave us high confidence in the numbers. Taking his $175 combined HBM2 + interposer figure, we’re nearly half-way to the MSRP of the Vega 56 card, with the rest of costs comprised of the VRM, GPU, and dime-a-dozen electrical components. It’d cost a “normal person,” for instance, about $45 to build the VRM on Vega – that’d include the $2.70 per-phase cost of the IRF6894s and IRF6811 hi- and lo-side DirectFETs, about $8.80 for all six of the IR3598 drivers, and roughly $4 on the IR35217 (from public sellers and datasheets). AMD is a large company and would receive volume discounts. Even as individuals, we could order 10,000 of these parts and drive that cost down, so these numbers are strictly to give an idea of what it’d cost you to build the VRM.

We’re not sure how many AMD ordered and aren’t going to speculate on what the company’s discount would be, but those numbers give an idea for what someone might pay if not a major corporation. This primarily helps explain why AMD opted for the same PCB and VRM on Vega: FE, Vega 64, and Vega 56, especially given the BIOS and power lock on V56. Although Vega can certainly benefit from the advanced VRM, the necessity of it lessens as we get to V56. The increased volume from V56 orders could offset cost across the entire product stack, to a point of it being cheaper to overbuild Vega 56 than to order two or three completely different sets of VRM components and PCBs.

Regardless, we’re at about $150 on HBM2 and $25 on the interposer, putting us around $175 cost for the memory system.

We did speak with numerous folks, including Kanter, on estimated GPU die pricing, but the estimates had a massive range and were ultimately just educated guesses. Without something more concrete to work with, we’re going to just stick to HBM2 and interposer pricing, as that’s the figure we know. GPU cost and yield cost are only really known by AMD and GlobalFoundries, at this point, so no point in working with total speculation.
The next question is what GDDR5 costs. A recent DigiTimes report pegs GDDR5 at about $6.50 for an 8Gb module, though also shows pricing for August onward at $8.50 per module. With old pricing, that’s around $52 cost for an 8GB card, or $68 with new pricing. We do not presently know GDDR5X cost. This puts us at around 3x the cost for HBM2 which, even without factoring in yields or the large GPU die, shows why AMD’s margins are so thin on Vega. We also know that AMD is passing along its HBM2 cost to partners at roughly a 1:1 rate – they’re not upcharging it, which is what typically happens with GDDR. There’s no room to upcharge the HBM2 with Vega’s price target.

Ignoring GPU cost and cost of less significant components, like the VRM and cooler, we’re at $100-$130 more than 8GB of GDDR5 cost to build. This is also ignoring other costs, like incalculable R&D or packaging costs. Again: We’re just focusing on memory today.

Why AMD Had to Use HBM2
Click to expand...
hynix-hbm2.png

Now that we know how much HBM2 costs, we need to talk about why AMD decided to use it. Like most of AMD’s hardware, the company is partly trying to make a long-term technological play in the market. This started with Fiji and has progressed through Vega.

There’s more to it, though. HBM2 critically allows AMD to run lower power consumption than GDDR5 would enable, given the Vega architecture.

Speaking with Buildzoid, we know that Vega: Frontier Edition’s 16GB HBM2 pulls 20W max, using a DMM to determine this consumption. This ignores the voltage controller’s 3.3v draw, but we’re still at 20W memory, and no more than an additional 10W for the controller – that’s less than 30W for the entire memory system on Vega: Frontier Edition.

We also know that an RX 480 uses 40-50W for its 8GB, which is already a significant increase in power consumption per-GB over Vega: FE. The RX 480 also has a memory bandwidth of 256GB/s with 8GB GDDR5, versus Vega 64’s 484GB/s. The result is increased bandwidth, the same capacity, and lower power consumption, but at higher cost to build. In order for an RX 480 to hypothetically reach similar bandwidth, power consumption would increase significantly. Buildzoid calculates that a hypothetical 384-bit GDDR5 bus on Polaris architecture would push 60-75W, and an imaginary 512-bit bus would do 80-100W. For this reason alone, HBM2 saves AMD from high power budget that would otherwise be spent solely on memory. This comes down to architectural decisions made years ago by AMD, which are most readily solved for with HBM2, as HBM2 provides greater bandwidth per watt than GDDR5. HBM is effectively a necessity to make Vega at least somewhat power efficient while keeping the higher memory bandwidth. Imagine Vega 56, 64, or FE drawing an additional 70-100W – the world wouldn’t have it, and it’d be among the hottest cards since the GTX 480 or R9 290X.

The Vega architecture is clearly starved by memory bandwidth, too: Overclocking HBM2 alone shows this, as its gains are greater than just core clock increases. AMD didn’t have another choice but to go with HBM2, even though costs would be roughly one-third on the memory. GDDR5 might be possible, but not without blowing power consumption through the roof or losing on performance by limiting bandwidth.

AMD provided GN with a statement pertaining to choices revolving around HBM2, which reads as follows:

“AMD chose HBM2 memory for Vega because this advanced memory technology has clear benefits on multiple fronts. HBM2 is a second-generation product that offers nearly twice the bandwidth per pin of first-generation HBM thanks to various refinements.

“As we noted in the Vega whitepaper, HBM2 offers over 3x the bandwidth per watt compared to GDDR5. Each stack of HBM2 has a wide, dedicated 1024-bit interface, allowing the memory devices to run at relatively low clock speeds while delivering tremendous bandwidth. Also, thanks to die stacking and the use of an interposer, Vega with HBM2 achieves a 75% smaller physical footprint for the GPU die plus memories versus a comparable GDDR5 solution.

“The combination of high bandwidth, excellent power efficiency, and a compact physical footprint made HBM2 a clear choice for Vega. We have no plans to step back to GDDR5.”

AMD ended up opting for two stacks of HBM2 on the current Vega cards, which limits its bandwidth to Fury X bandwidth (2x 1024-bit Vega vs. 4x 1024-bit Fury X), ultimately, but AMD does benefit in the bandwidth-per-watt category. That’s the crux of this decision.
Battling for Marketshare & Margin
As for cost, knowing that the memory system gets us up to nearly $200 as a starting point, it is inarguable that AMD has lower margins on Vega products than could be had with GDDR5 – but the company also didn’t have a choice but to use HBM2. NVidia forced AMD’s hand by dropping the 1080 Ti in March, followed by 1070 and 1080 MSRP reductions. That’s ignoring the current insane GPU pricing (which inflates the 1070s & V64s into overpriced oblivion) and just looking at MSRP, as that’s ultimately where the two companies battle under normal conditions. AMD and nVidia also do not see a dollar of the upcharge by distributors and retailers, so margins for either company are theoretically unimpacted by the inflated consumer-side pricing. That’s why, ultimately, we’re looking at MSRP – AMD and nVidia sell their product to the AIB partners for a cost which should theoretically be largely unchanging. AMD is able to make back some of these margins with bundle packs, where a pair of games can be sold to AIBs for ~$29, then to consumers for $100, or where Ryzen and motherboard parts help recoup margins. Each motherboard sold is another chipset moved, and Ryzen sales go completely to AMD. Either way, AMD has to increase its GPU marketshare, and fighting through early losses or slim margins is part of that. The long-term play is clearly hoping that increased demand and supply will lower cost to build, so it remains to be seen how that’ll play-out.

AMD’s investing a lot of effort to try and recoup some of the margins: Bundle packs are a major boon, either through direct cost of games sold or through accompanying Ryzen product sales, and reusing the same PCB & VRM further helps slim margins. This is particularly true with a low volume part like Vega: FE, as using the same board will help meet MOQ thresholds for discounts, aided by higher volume V64 and V56 parts. Without immediate production runs of significant quantities on each SKU, it makes more sense for AMD to reuse the VRM and board than to design a cheaper V56 board, as cost across all three SKUs is lowered with higher quantities. This particular move has the upshot of benefitting V56 consumers (though the benefit would be more relevant with unlocked BIOS), as you end up with a seriously overbuilt VRM for what’s needed. The VRM can easily handle 360W through the card and is more than the V56 will ever draw stock or with a 50% offset. We’ve even tested up to 406W successfully (with a 120% offset) through V56, though it’s probably inadvisable to do that for long-term use.

But that gives us a starting point, and helps to better contextualize what people mean when they say “HBM2 is expensive.” It is – about 3x more than GDDR5, give or take $20 and assuming high yield – and that’s rough for AMD’s profitability. That said, the company really couldn’t have reasonably gotten GDDR5 onto Vega’s present design without severe drawbacks elsewhere. It wouldn’t compete, and we’d be looking at another Polaris mid-range GPU or much higher power consumption. At less than 30W for 16GB of HBM2, GDDR5 just can’t compete with that power consumption under command of the Vega architecture. It’d be cheaper to make, but would require significantly higher power consumption or a smaller bus, neither of which is palatable. AMD isn’t using HBM2 to try and be the “good guy” by pushing technology; the company was in a tough spot, and had to call riskier shots out of necessity. Although it’d be nice if every GPU used HBM2, as it is objectively superior in bandwidth-per-watt, both AMD’s architecture and market position pressure the company into HBM adoption. HBM2 would benefit a 1070 by way of lower power consumption, but the 1070 doesn’t need HBM2 to get the performance that it does – the architecture is less bandwidth-hungry, and ultimately, nVidia isn’t in the same market position as AMD. In a 30-to-70 market, AMD has to make these more expensive plays in attempt to claw back marketshare, with hopes of enjoying better margins further down the generation.
Click to expand...

40% less power consumption with HBM2 vs GDDR5.
 
HW News: GDDR6, HBM3, Zen CCX Architecture, PCIe Gen4
Gamers Nexus
Published on Aug 26, 2016
[video="https://www.youtube.com/watch?v=Ym_KtnG_sgQ&ab_channel=GamersNexus"][/video]

Edit:
http://www.gamersnexus.net/news-pc/2577-hw-news-samsung-gddr6-hbm3-pcie-gen4-power
Transcript
Memory manufacturer Samsung is developing GDDR6 as a successor to Micron's brand new GDDR5X, presently only found in the GTX 1080 and Titan XP cards. GDDR6 may feel like a more meaningful successor to GDDR5, though, which has been in production use since 2008.

In its present, fully matured form, GDDR5 operates at 8Gbps maximally, including on the RX 480 and GTX 10 series GPUs. Micron demonstrated GDDR5X as capable of approaching 12-13Gbps with proper time to mature the architecture, but is presently shipping the memory in 10Gbps speeds for the nVidia devices.

Samsung indicates an operating range of approximately 14Gbps to 16Gbps on GDDR6 at 1.35V, coupled with lower voltages than even GDDR5X by using LP4X. Samsung indicates a power reduction upwards of 20% with post-LP4 memory technology.

Samsung is looking toward 2018 for production of GDDR6, giving GDDR5X some breathing room yet. As for HBM, SK Hynix is already looking toward HBM3, with HBM2 only presently available in the GP100 Accelerator cards. HBM3 will theoretically run a 4096-bit interface with upwards of 2TB/s throughput, at 512GB/s per stack. We'll talk about this tech more in the semi-distant future.

PCIe

Tom's Hardware this week reported on the new PCI Express 4.0 specification, primarily detailing a push toward a minimum spec of 300W power transfer through the slot, but could be upwards of 500W. Without even talking about the bandwidth promises – moving to nearly 2GB/s for a single lane – the increase of power budget will mean that the industry could begin a shift away from PCI-e cables. The power would obviously still come form the power supply, but would be delivered through pins in the PCI-e slots rather than through an extra cable.

This same setup is what allows cards like a 750 Ti to function only off the PCI-e slot, because the existing spec allows for 75W to push through the PCIe bus. PCI-e 4.0 should be ratified by the end of 2016 by the PCI-SIG team, but we don't yet know the roll-out plans for consumer platforms.
Click to expand...

Note: this was 1 year ago!
 
@martel Polaris 20 is the "improved" polaris 10 that currently being used in RX580/570. Vega 20 will be AMD successor for Hawaii. More specifically it was AMD new FP64 accelerator replacing Firepro S9150. though they said it will be built uaing 7nm node. So i'm not sure if they wil be ready in 2018 time frame. 7nm for SoC type of chip might be ready in 2018 but for bigger die size it might ot be ready until 2019 time frame.

Also when it comes to process it might be useless to compare the process nide from different fab sincr each of them uaing their own definition. 14nm for example from what i heard only intel have have "true" 14nm node. Others like TSMC for example their 16nm is based on earlier 20nm node. That's why some people said TSMC 16nm is just their 20nm but with FinFet instead of planar. Though when it comes to this fab company i think TSMC is still one of the best.
 
By the time Vega 11 (and 20) are ready, I would imagine SK Hynix and Samsung would have solved their HBM2 capacity issues? That should drive the bulk price down to a point where it might make it feasible to aim at a mid-range product stack with them. Plus, like Martell pointed out, reworking the IMC for GDDR5X or 6 would change the PCB layout too much for Vega, so it could even be more expensive to go that route.

Additionally, they realized too late Vega wasn't going to cut it, so this time I would imagine they'll try to revamp the whole product stack to compete on efficiency terms with nVidia. That is to say, get a RX-580 replacement that sits as close as possible to the 1080 with less consumption. I actually like that.

Cheers!
 
Information on node scaling taken from https://www.semiwiki.com
AXdmPFR.png

Metal_Track_Scaling.jpg

Transistor performance
17230d1462285489-speed-improvment-graph-min-2-jpg

20113d1499886241-standard-node-trend-2-jpg


14nm 16nm 10nm and 7nm - What we know now
by Scotten Jones
Published on 04-07-2017 05:00 AM
Conclusion
Intel's 14nm process is significantly denser than the competing processes from GF/SS and TSMC, >1.5x. It has taken roughly 3 years for SS and TSMC to introduce 10nm processes that are only slightly denser than Intel's 14nm process.

Later this year when Intel introduces their 10nm process they will again take the process density lead, ~1.7x denser until TSMC's releases their 7nm process with slightly better density.

When TSMC introduces their 7nm process they will have a 1.13x density advantage over Intel's 10nm process. In late 2018 Samsung is later expected to introduce their 7nm process with 1.23x the density of Intel's 10nm process, also later in 2018 TSMC will introduce 7+ with a further density improvement.

Intel's 10nm is more similar to TSMC and GF/SS 7nm processes than to the competing 10nm processes. Even though Intel has a significant density advantage at each node the forthcoming 7nm foundry process will likely pass Intel for process density and maintain that yield for at least a few years.
Click to expand...

Standard Node Trend
by Scotten Jones
Published on 07-15-2017 05:00 PM
Conclusion
By either the old or the new standard node values Intel has lost their multiyear density lead over the foundries. Based on the new more accurate standard node value the average node value for Intel's 10nm and the foundry 7nm processes is 8.05nm and all four companies are within a 0.5nm standard node value of each other.
Click to expand...

 
Looking at verkkokauppa.com, one of Finland's premier retailers, looks like the waiting time for RX Vega 64 is just 1-2 weeks. Only 9 units have been pre-ordered and the price is 760eur.

The card might be a disappointment to miners due to the high power consumption (early reports for Vega I saw indicated the performance is OK) and for gamers because of the price. 2017 is a dark year for system builders when combined with the high memory prices.

Or then nobody just has faith that the retailer can actually get their hands on the cards in time. Availability of Radeon and Geforce cards has improved somewhat from the most difficult periods during the summer, but prices remain on the high side.

Vega Frontier Edition is in stock with 16 units - a whopping saldo in these days considering it's a high-end professional card!
 
Hot Chips 2017: We'll See PCIe 4.0 This Year, PCIe 5.0 In 2019
by Paul Alcorn August 29, 2017 at 6:30 AM
aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9DLzEvNzA2NzUzL29yaWdpbmFsLzEwMC5QTkc=

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9DL0EvNzA2NzYyL29yaWdpbmFsLzEwNS5QTkc=

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9DLzIvNzA2NzU0L29yaWdpbmFsLzEwMS5QTkc=

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9DLzMvNzA2NzU1L29yaWdpbmFsLzEwMi5QTkc=

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9DLzcvNzA2NzU5L29yaWdpbmFsLzEwNy5QTkc=

aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9DLzYvNzA2NzU4L29yaWdpbmFsLzEwNC5QTkc=

We caught up with PCI-SIG, the organization that defines the PCI Express bus and form factor specifications, at the yearly Hot Chips conference. PCIe is the ubiquitous engine that pulls a big part of the computing locomotive down the track—it touches nearly every device in your computer. As such, it is the linchpin for the development of many other technologies, such as storage, networking, GPUs, chipsets, and many other devices.
Considering its importance, it isn't surprising to find the PCI-SIG with 750 members worldwide. Unfortunately, large organizations tend to move slowly, and PCIe 4.0 is undoubtedly late to market. PCIe 3.0 debuted in 2010 within the normal four-year cadence, but PCIe 4.0 isn't projected to land in significant quantities until the end of 2017—a seven-year gap.
PCI-SIG representatives attributed part of the delay to industry stagnation. The PCIe 3.0 interface was sufficient for storage, networking, graphics cards, and other devices, for the first several years after its introduction. Over the last two years, a sudden wellspring of innovation exposed PCIe 3.0's throughput deficiencies. Artificial intelligence craves increased GPU throughput, storage devices are migrating to the PCIe bus with the NVMe protocol, and as a result, networking suddenly has an insatiable appetite for more bandwidth.
The industry needs PCIe 4.0 to land soon, and PCI-SIG assures us it will ratify the new specification by the end of 2017. The sluggish ratification process hasn't hampered adoption entirely, though. Several IP vendors already offer 16GT/s controllers, and many vendors have already implemented PCIe 4.0 PHYs into their next-generation products. These companies are plowing ahead with the 0.9 revision of the specification, whereas the final ratified spec debuts at 1.0.
PCI-SIG says it is accelerating the development and feedback processes, along with simplifying early specification revisions, in a bid to reduce time to market for future specifications.
PCI-SIG indicates that PCIe 4.0 will be a short-lived specification because the organization has fast-tracked PCIe 5.0 for final release in 2019.
Click to expand...
aHR0cDovL21lZGlhLmJlc3RvZm1pY3JvLmNvbS9DL0UvNzA2NzY2L29yaWdpbmFsLzIwNC5QTkc=

PCIe 4.0 will bring us 64GBps of throughput, while PCIe 5.0 will double that to 128GBps. Both revisions still leverage the standard 128b/130b encoding scheme that debuted with PCIe 3.0. PCI-SIG representatives said they are satisfied with the 20% reduction in overhead facilitated by the 128b/130b encoding, and further encoding refinements to reduce the current 1.5% overhead are subject to a diminishing point of returns.
Although we have yet to see PCIe 4.0 on the market, the fast-tracked PCIe 5.0 is already on revision 0.3, and the group predicts 0.5 will be available by the 4Q17.
The PCI-SIG defines the specification, but it has no control over when the end devices make it to market. Intel and AMD are the key enablers for the broad desktop market; we certainly won't see PCIe 4.0 GPUs and SSDs without a slot to plug them into. AMD has slated PCIe 4.0 for 2020. We imagine Intel is also chomping at the bit to deploy PCIe 4.0 3D XPoint and NVMe SSDs, but the company remains silent on its timeline.
Click to expand...
 
well there is some wild speculation about what this going to mean for Raja and RTG everywhere. right now Lisa Su will helm RTG in the absence of Raja. when i heard this news i suddenly remember about the talk Raja being intel mole to get RTG away from AMD and to be acquired by intel haha.
 


Should add that Verkkokauppa has many loyal customers and seems they've been exploiting the situation on the market to the full with high prices (and yet surprisingly slim supply) on GPUs, compared to local competitors.

At Mindfactory.de they seem to have stock and already hundreds of Vega cards sold. RX 56 models start at 509eur. Being a reference design, the Vega cards are not that stellar yet but at least this is something.
 


Are these the standalone cards, or the bundle deals?
 


Standalone.
 
AMD To Change Suppliers for Vega 20 GPUs on 7nm, HBM2 Packaging for Vega 11
by Raevenlord Friday, September 8th 2017 15:15
AMD's RX Vega supply has seen exceedingly limited quantities available since launch. This has been due to a number of reasons, though the two foremost that have been reported are: increased demand from cryptocurrency miners, who are looking towards maximizing their single node hashrate density through Vega's promising mining capabilities; and yield issues with AMD's Vega 10 HBM2 packaging partner, Advanced Semiconductor Engineering (ASE). It's expected that chip yield for Vega 10 is also lower per se, due to it having a 484 mm² die, which is more prone to defects than a smaller one, thus reducing the amount of fully-enabled GPUs.

AMD's production partner, GlobalFoundries, has historically been at the center of considerations on AMD's yield problems. That GlobalFoundries is seemingly doing a good job with Ryzen may not be much to say: those chips have incredibly small die sizes (192 mm²) for their number of cores. It seems that Global Foundries only hits problems with increased die sizes and complexity (which is, unfortunately for AMD, where it matters most).
Due to these factors, it seems that AMD is looking to change manufacturers for both their chip yield issues, and packaging yield problems. ASE, which has seen a 10% revenue increase for the month of August (not coincidentally, the month that has seen AMD's RX Vega release) is reportedly being put in charge of a much smaller number of packaging orders, with Siliconware Precision Industries (SPIL), who has already taken on some Vega 10 packaging orders of its own, being the one to receive the bulk of Vega 11 orders. Vega 11 is expected to be the mainstream version of the Vega architecture, replacing Polaris' RX 500 series. Reports peg Vega 11 as also including HBM2 memory in their design instead of GDDR5 memory. Considering AMD's HBM memory history with both the original Fury and and now RX Vega, as well as the much increased cost of HBM2's implementation versus a more conventional GDDR memory subsystem, this editor reserves itself the right to be extremely skeptical that this is true. If it's indeed true, and Vega 11 indeed does introduce HBM2 memory to the mainstream GPU market, then... We'll talk when (if) we get there.

As to its die yield issues, AMD is reported to be changing their main supplier for their 7 nm AI-geared Vega 20 from GlobalFoundries to Taiwan Semiconductor Manufacturing Company (TSMC), who has already secured orders for AI chips from NVIDIA and Google. TSMC's 7nm and CoWoS (chip-on-wafer-on-substrate) capabilities have apparently proven themselves enough for AMD to change manufacturers. How this will affect AMD and GlobalFoundries' Wafer Agreement remains to be seen, but we expect AMD will be letting go of some additional payments GlobalFoundries' way.
Click to expand...
Link to Gamers Nexus News talking about this.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom