I'm pretty sure Ampere is one of the fastest growing nvidia generations on steam's hardware survey in recent history, despite everyone complaining about stock. If there were low yields they must have improved it very quickly.
I'm feeling somewhat vindicated after arguing with people in the comments section of the 12th gen coverage lol. I bet the real deal ADL mobile will be even better. Hopefully they offer LPDDR5 too to save power on memory. Imagine what this would be like on 5nm. Intel still has it in them to win...
Taiwan is not a part of China. In fact ryzen isn't even fabricated by any east Asian company. They're manufactured by GF, an American company
They also arent in the lead for AI. The US still dominates the field in patents. China only has a lead in one narrow subtopic...
Wouldn't the cost of the 8114Y be prohibitive for a $400 tablet? The surface pro with the most basic options has ultra low-performance cores but starts at $800 and it's not fantastic imo
I highly doubt they regressed backwards to low single digit %. They started at low double digit and grew to >15% and expected to hit 20%. Even with half gen consoles I highly doubt AMD would accept a lower margin than what they grew...
Hmm not bad. But apparently there's a J4105 that's quad-core with the same TDP and price point. When will we see that in a NUC? In any case, the value is hurt badly by RAM prices right now. 8GB of DDR4 is $90 now, 3/4 the cost of that NUC on newegg!
Interesting to see a high profile processor have such a mundane problem. Is it fixable in software? And why do android phone SoCs still rely on big.LITTLE instead of just conventional voltage and frequency scaling. Is there an actual measurable advantage?
Intel's is just buying actual physical GPUs. They're not licensing anything from AMD AFAIK. GCN ISA is a thing though, Intel just doesn't have the rights to use it.
DSBR works, they just disabled it because worst case it lowers performance and best case there's negligible gain. It's mostly for saving bandwidth and for deferred rendering as shade once, seemingly, can only happen in a deferred rendering context with full information of depth.
The 12nm process promises +10% over leading 16nm and 14nm so we should see about those speeds for the 1800x equivalent. Maybe more from high level optimizations.
Technically they claim it is with the new NGG fast path but last we got ANY information about it it's still not active though it's detailed and measured in the whitepaper.
It's probably not possible to get the "max" out of Vega at the moment with the front-end it has. You can see GCN's performance per Teraflop change as you go down the models, with the RX 460 having among the highest of the recent GCN cards. In comparison, Vega's perf/TF is like half that of the...
No crashes yet, it's just failing HCI memtest and google's stressapptest.
Edit: seems to pass in Linux mint. I thought it was my windows installation being bad so I reinstalled it fresh but it's still failing and the only thing I have on it is basically Firefox and the AMD drivers. And in mint...
Why is bandwidth the culprit here? Wouldn't geometry be the more obvious problem considering GCN has historically had a problem with geometry rich scenes? Tessellation was a huge weakness and still a sore spot, but a measly 4 geometry engines seems archaic
Anyone have memory instability after installing Vega? I put a new Vega 56 into my system a few days ago and now all of my RAM is testing unstable. It's a ryzen build with a 1700, crosshair vi hero, and Samsung b-die yet it's unstable down to 2400, and even on a 2nd pair of ram that was stable...
Vega 11 should be out within 6 months considering it supposedly taped out months ago. It should be the new 680. I imagine we won't see a high-end refresh till 2018 though.
I get that, I'm just wondering why there has been 0 talk about primitive shaders, which will supposed be one of the biggest changes to GCN in its entire lifespan. Primitive binning refers to the fact that it's not "fully" tiled rasterization. Only the primitive geometry (vertices, tris/polys)...
Have the linux driver guys announced any efforts on primitive shaders?
Edit: Reading the white paper again it seems like the binning is only useful for deferred rendering
TPU shows it faster by 20% at 1080p, but the gap widens as resolution goes up. The 1080 is probably bottlenecked at the backend at higher resolutions.
Edit: dumb autocorrect
Perhaps forza 7 is significantly more shader intensive than it is geometry intensive. These performance results are a bit closer to what you might expect just by looking at compute throughput, minus some performance loss from a minor bottleneck (some efficiency loss from scaling shaders up)...
Unfortunately I don't think you can just look at the geometry complexity of a game and determine how it affects performance. Pretty sure overwatch isn't heavy on geometry, and neither is Hellblade probably (it's by a small time studio iirc). There are other things to consider like if the game...
Actually, after reading that it sounds like improved geometry throughput will help. Sounds like GCN has a lot of troubles.
The gist seems to be that unseen geometry is a huge waste because it takes up CU time to render nothing (since it's never shaded) and forces other CUs to idle because the...
Front-end is more than just the geometry shaders... Front end also includes the work distributor, which ngg purports to improve. Higher geometry throughput is just another benefit, and even if you improvement throughput if you don't fill the wavefronts you won't be able to fully utilize the...
I doubt the geometry throughout is the bottleneck, especially since games historically do their own culling before the gpu has to do anything. The packing of object instances sounds more likely to give a boost in performance. Iirc wavefronts (and warps) of x size have to have all x threads work...
The async compute claims aren't so ridiculous, it's just a little out of context. The first time Asynchronous Computation was used was by PS4 devs, who, afaik implemented it themselves since the PS4 uses some sort of very low-level openGL derived API designed by Sony. The same concept was...
Very few partners seem to be mentioning the card manufacturer partners. I don't think it's impossible that they're jacking up prices to get a cut of the mining craze. Pretty sure AMD still doesn't produce any of the PCBs or do any of the card assembly themselves.
At least the horse power is all there. And we've seen in at least one geometry benchmark that Vega is strong in geometry throughput, though apparently not quite as strong in geometry culling. So where's the weakness? Is it that pesky non-functional tiled rasterizer? Maybe it's even the memory...
Based on the content creation program benchmarks for Vega FE it seems to match big Pascal in those programs just fine at a similar level of efficiency. So HBM is just fine for those purposes apparently, the only problem is price. Until we know more about why games perform so much worse I...
Not sure how anyone can question HBM. It's just expensive for now because it's new but even HBM2 can slam dunk over GDDR6's projected bandwidth with a 4-stack at full speed (1000MHz) IIRC. Something like 1024GB/s with a 4-stack @1000MHz and at lower than GDDR5X power consumption. Price is the...
Honestly I'd wait for RX to come out to judge it. It seems odd that AMD would claim higher IPC over the last generation as recently as like a month ago only to have Vega have a severe IPC regression compared to two generations ago.
It's also possible they decided to put their best guys on Navi, which was probably started a year or so after Vega was started, maybe even at a similar time.
Fiji was probably already in development in 2015 when that happened. It's more likely that RTG China's team is just really bad. A lead can only do so much with poor talent.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.