AMD on the road to recovery.

[smilingcrow]

On the one hand I hear people saying that AMD need a new GPU architecture and then they release new cards based on the older one and yet they still seem to take a long time to get the most out of them driver wise?
Seems strange!

 

[Gerard]

Then maybe AMD should make sure their drivers are already as good as they can be before releasing the card? Sure nvidia make driver improvements too but not to the extent AMD do.

Probably no such thing as a driver being as good as it can be as there's always room for improvement somewhere. I've always said day one drivers from amd are usually lacking and for a new card launch its a weird way to go about things. Just seems like the drivers have been scrambled together to meet the launch date.

 

[DarrenM343]

Probably no such thing as a driver being as good as it can be as there's always room for improvement somewhere. I've always said day one drivers from amd are usually lacking and for a new card launch its a weird way to go about things. Just seems like the drivers have been scrambled together to meet the launch date.

But you know GPU hardware development would have been in development for some time (years) so you'd expect driver development to be done alongside that.
Lots of different opinions out there of course, but one of mine is that I think AMD struggle to optimise the software to their hardware for whatever reasons, maybe an architecture that's difficult to optimise for, where as NV has pretty optimised drivers from day 1 hence why AMD get more out of theirs over the longer time while NV doesn't. Maybe NV give more thought to the software side of things when they develop the hardware design.
On the other hand, hardware cannot be changed so maybe AMD simply just spend more time/£ on that at first (due to lower R&D funding) and then once done, then switch over to software side of things which of course can be frequently changed.

Ultimately as customers we expect top performance from day 1. Performance doesn't come out of thin air so big optimisations further down the line means it was probably less than optimal to start with.

 

[nashathedog]

295x2 was a faster card than the titan z.

I was just looking at a review from back then and which was the faster card tended to depend on the game being tested, Even when there was a win the difference was often small & within margin of error. The 295 had the TZ beat hands down on price though so it would have been the smart one to buy of the two and today I'd put money on it being miles ahead in modern games even though it has less memory.

https://www.guru3d.com/articles_pages/nvidia_geforce_gtx_titan_z_review,1.html

 

[Gerard]

But you know GPU hardware development would have been in development for some time (years) so you'd expect driver development to be done alongside that.
Lots of different opinions out there of course, but one of mine is that I think AMD struggle to optimise the software to their hardware for whatever reasons, maybe an architecture that's difficult to optimise for, where as NV has pretty optimised drivers from day 1 hence why AMD get more out of theirs over the longer time while NV doesn't. Maybe NV give more thought to the software side of things when they develop the hardware design.

Again it could likely come down to money and available staff. There was an article a while ago saying a lot of staff working on Vega were taken to work on ryzen (or something along those lines). Could be people are pulling double duty and some things get put on the back burner until the last minute. We can only speculate though.

 

[EsaT]

Then maybe AMD should make sure their drivers are already as good as they can be before releasing the card? Sure nvidia make driver improvements too but not to the extent AMD do.

Time for you to go look into mirror:
Maybe AMD would have had more resources to make better drivers from the start, if more people had bought AMD cards whey they were good ones.

 

[gavinh87]

Time for you to go look into mirror:
Maybe AMD would have had more resources to make better drivers from the start, if more people had bought AMD cards whey they were good ones.

I for one did. And back them when AMD were enjoying success with GPUs they were struggling with drivers.
I'm not sure what they are like now but one thing I can see if that performance improves over time. Its too late by then because all major reviewers don't bother going back to retest months down the line.
They are actually hurting themselves.

 

[Gerard]

https://www.pcper.com/news/General-...om-AMD?utm_source=dlvr.it&utm_medium=facebook

More server room taking for epyc.

 

[Gerard]

Time for you to go look into mirror:
Maybe AMD would have had more resources to make better drivers from the start, if more people had bought AMD cards whey they were good ones.

One of AMD's big failures was taking that payoff from intel for the antitrust suit, they had intel bent over a barrel and instead of ploughing them raw they opted for the quick and easy approach. Had that case went the duration i could have seen AMD get a much larger payout than what they received.

 

[CAT-THE-FIFTH]

One of AMD's big failures was taking that payoff from intel for the antitrust suit, they had intel bent over a barrel and instead of ploughing them raw they opted for the quick and easy approach. Had that case went the duration i could have seen AMD get a much larger payout than what they received.

I suspect Intel could have lawyered up and waited for AMD to fold. That is what Creative did to one of its competitors even after they lost the case.

 

[Gerard]

I suspect Intel could have lawyered up and waited for AMD to fold. That is what Creative did to one of its competitors even after they lost the case.

Creative are smaller than amd i'm pretty sure, plus i think amd has some middle eastern investors as well. Had it went the distance it might have been a far heftier fine.

 

[Hagar]

Please try NOT to “Act-React-to”/be-Distracted-by this “commentary”.

Don't worry, I won't [DWIW].

 

[drunkenmaster]

Creative are smaller than amd i'm pretty sure, plus i think amd has some middle eastern investors as well. Had it went the distance it might have been a far heftier fine.

Creative was the far bigger company, it was Aureal's tech the stole and a lawsuit from Aureal, Creative just kept pushing for depositions and changes of venue, anything really to eat into Aureal's tiny cash reserves. Eventually from lack of sales and a dragged out court case they went bust and Creative bought out their patents at a fraction of their cost from the bankrupt company.

AMD was in significant debt, largely thanks to the things Intel did. This court case had already gone on for years and was costing in the 100s of millions in lawyer fees.

Intel lost a similar case to the EU but the EU has no reason to settle so the case went the distance, they got a slightly bigger fine (but no damages there, it's purely a fine for their behaviour). ALmost a decade later Intel are still appealing and still haven't paid.

In law the deeper pockets wins 99% of the time unless one of the parties is the government, even then the losing party can appeal and hold up payment for years, even decades. At some point I wouldn't be surprised if the EU settles with Intel for something like half the price just to get it done.

If AMD stayed to fight they would have won a judgement easily 5 times as big..... but that judgement would have still been some time away and then Intel would appeal it for a decade. The 1.25billion cash influx combined with savings on lawyer fees (which by the time appeals finished would easily been 100+ more million) is something that saved AMD and enabled them to make it to the point Zen released in reality. With a little less cash and a little more debt they very likely would have folded.

Then also got some pretty decent benefits from the settlement which iirc included not being limited to producing x86 at their own fabs, which allowed AMD to make the move to sell the fab side of their business, another massive step that enabled them to get to the point they are at today and enables them to make x86 cpus at an outside company(global and also now TSMC).

In a perfect world sure, they stay in, get a judgement in a few months and get 5-10billion and Intel have to pay lets say 75% of it upfront and appeals can't slow down payment. Unfortunately in the real world AMD would have gone bankrupt waiting for a big payout if they didn't settle.

 

[smilingcrow]

When the sole aim is to “Rob Peter In Order To Pay Paul” [RPTPP], then there is NO real difference between, for example, “AMD” or “Intel”. When the real aim is to Sell, Sell, & Sell, then Buy, Buy, & Buy is the avenue/”revenue”.

Hands up those who are unable to realise so other than to believe what they are “Told To Believe”. Be aware [but how?] that when The Blindness is self-certified as “20-20 Vision”, then curing/vision is not The Point but The Desire for Reality/”Purity” is.

Just because “TheBank of England’s Base Interest Rate Rise” [BoEIRR]
is counterable by, for example, “Easy-Coming”, does not mean that said BoEIRR will not affect The Vast Majority. It will because The Human Universe “Spins” on RPTPP.

And what does all the “Yo-Yo-ing” means? It means that without becoming-being Real, Reality is merely Relativity/Relativism. After all, without Being Real, there is NO such thing as Reality. Without The Desire for Reality, there is then only Relativity. The Moment when Another is, somehow but don’t know how, more Real/”Important” than Self/self. There is No such thing for when Another “donks” self “On The Head”, ONLY self feels said “donking”, regardless of The Empathy/Sympathy/Hypocrisy.

Since Absolute Knowledge existed before Relative/”Human” Knowledge, you know, when Absolute Knowledge was The One who preceded/”gave-rise-to” Mere Knowledge, Human Knowledge is not only Relative but is Blindly Accepted-Rejected as Absolute Knowledge. No wonder another name for “Evil”/”Dracula” is The-Joker. And how velly hee-hee, haw-haw of Him.

“The Establishment” might be “taxing pensions” time-and-again but “Pensioners” have NO means of minding so. More so when “Instituting Royal Parties” will easily “distract”/mesmerise The Afflicted/Addicted. You know, “Dumbing-Down” The Dumb, if such a thing is possible. You know, when “AMD vs Intel” is “Do-able”.

Please try NOT to “Act-React-to”/be-Distracted-by this “commentary”.

Sounds as if you might need to consider upgrading your VIA chip!

 

[Vince]

Sounds as if you might need to consider upgrading your VIA chip!

I'd be surprised if anybody got through the first couple of sentences. I tried and gave up.

 

[smilingcrow]

I'd be surprised if anybody got through the first couple of sentences. I tried and gave up.

Maybe I was wrong with my VIA reference and this is the latest Intel security flaw!
I do have a sense of where they are coming from and it's not good.

 

[Vince]

Maybe I was wrong with my VIA reference and this is the latest Intel security flaw!
I do have a sense of where they are coming from and it's not good.

If we say there is some sort of conspiracy going on and it's the AI / machine learning that people are talking about, then judging by that it's got a long way to go

I would be fuming if I spent all that money and that was the output.

 

[Panos]

But you know GPU hardware development would have been in development for some time (years) so you'd expect driver development to be done alongside that.
Lots of different opinions out there of course, but one of mine is that I think AMD struggle to optimise the software to their hardware for whatever reasons, maybe an architecture that's difficult to optimise for, where as NV has pretty optimised drivers from day 1 hence why AMD get more out of theirs over the longer time while NV doesn't. Maybe NV give more thought to the software side of things when they develop the hardware design.
On the other hand, hardware cannot be changed so maybe AMD simply just spend more time/£ on that at first (due to lower R&D funding) and then once done, then switch over to software side of things which of course can be frequently changed.

Ultimately as customers we expect top performance from day 1. Performance doesn't come out of thin air so big optimisations further down the line means it was probably less than optimal to start with.

AMD the last 5 years had four different architectures. Hawaii, Fiji, Polaris, Vega. Each bringing a lot of changes and fancy things.
True the driver team had always to keep up, but that made some products like the 290X to survive even today sitting in upper mid range (with full DX12 support), while it's direct competitor the 780Ti (and the Kepler Titans) bitten the dust in performance years now. (even the 1050ti beats it).


Nvidia has "better drivers" from the start, because is stuck at same architecture since Kepler. There are tiny changes and just node shrinking ever since. No fancy stuff going on the hardware. No hardware async compute etc. And is the reason Nvidia is trying tooth and nail to keep the market (aka games) to DX11.
The moment games diverge to DX12 and/or HDR the performance drop of Nvidia GPUs is ridiculous. There is a long discussion in the graphics forum about HDR and DX12 comparisons. There are there games were the GTX1080 loses almost half it's performance by switching to DX12 and HDR compared to DX11 & SDR. In comparison Vega 64 loses handful of FPS at most if nothing at all, and with the latest drivers those benchmarks are run (May / June 2018), is already ahead in perf in DX11 & SDR compared to GTX1080.

When Nvidia was truly changing architectures (Fermi etc) fast, their driver team couldn't keep up also. Leaving AMD dominating the performance and price.

 

[Panos]

https://www.pcper.com/news/General-...om-AMD?utm_source=dlvr.it&utm_medium=facebook

More server room taking for epyc.

Yes. Especially now AMD getting all the classifications required for the server market, their share will go up.
Already there are new server lineups of companies using AMD CPUs, and yesterdays announcement that the new 2019 Xeons will be at 14nm, means that they can carve bigger piece of the pie, because the product is cheaper to the end consumer, cheaper to make (aka higher profit), and more powerful compared to the competition, which is stuck on monolithic chips hitting their performance limits due to size and architecture.
Intel will struggle to pull a monolithic 32 core server chip next year or a 48 core for 2020, yet AMD already is sampling 64 core CPUs to it's partners for mass availability early next year. And after that there is Zen 3, Zen 4, while 5nm will bring Zen 5 (Glofo schedule is 2022-3).

We might see 96 core chip before the end of the decade from AMD. While Intel will be struggling to make one half that size.

 

[humbug]

AMD +1.4% on Goldman upgrade for Intel competition

https://seekingalpha.com/news/33816...-goldman-upgrade-intel-competition#email_link

Goldman Sachs upgrades AMD (NASDAQ:AMD) from Sell to Neutral and raises its price target to $21, a 10% upside to yesterday’s close.