General Discussions

[kingfish]

MD unveiled its new Radeon RX 5700 line of graphics cards with 7nm chips at E3 last month, and with just days to go before they launch on July 7th, the company has announced new pricing. In the "spirit" of competition that it says is "heating up" in the graphics market -- specifically NVIDIA's "Super" new RTX cards -- all three versions of the graphics card will be cheaper than we thought.

The standard Radeon RX 5700 with 36 compute units and speeds of up to 1.7GHz was originally announced at $379, but will instead hit shelves at $349 -- the same price as NVIDIA's RTX 2060. The 5700 XT card that brings 40 compute units and up to 1.9GHz speed will be $50 cheaper than expected, launching at $399. The same goes for the 50th Anniversary with a slightly higher boost speed and stylish gold trim that will cost $449 instead of $499.

That's enough to keep them both cheaper than the $499 RTX 2070 Super -- we'll have to wait for the performance reviews to find out if it's enough to make sure they're still relevant.

AMD fires back at 'Super' NVIDIA with Radeon RX 5700 price cuts 

[pokester]

If they really were in touch with what their customers want they would have a place for you to add to their list of things they want you to choose from. I hate things like this that take no real interest in what the end user really wants and waists your time doing it. 

[colesdav]

I have asked before for a link to a brief Software Technical Specification summary of what they mean or are suggesting. 
Without that you have no idea what you are even voting for.

[black_zion]

They don't even need that. What has every tech company, besides AMD, been doing for the last decade or so? Copying the most popular features of a competitor. AMD may not be able to copy DLSS and nVidia Broadcast, at least not right now, but they can copy their minimalist control panel and GeForce Experience, dumping the rather very bloated Crimson 2020 for a more streamlined experience, and of course focus on driver features, and stability, and that's "easy" to fix, just pay skilled software engineers the cash, either find them on LinkedIn or poach them from another company Intel's doing it, and you can bet when Xe blasts onto the scene they won't be releasing some half broken and bloated driver set.

According to Indeed, nVidia pays $131,000 a year for a software developer and $231,000 a  year of a principle software engineer.

According to Indeed, AMD pays senior software developers $167,000 a year

If these figures are truly accurate, it would definitely explain why AMD has trouble attracting the top talent, and is something AMD, now flush with cash, needs to remedy quick, fast, and in a hurry.

[colesdav]

If you look at the Adrenalin 2020 20.9.1 installer picture above, there is a "Keep User Settings" option.

[pokester]

I have never actually used the 2020 interface as I could never get it to load on my RX 580 or earlier cards. My one attempt where it did load the computer restarted and never started again after that. 

[pokester]

I hope they all get replacement boards not a bios that cripples performance. 

[colesdav]

Looks like the MLCCs have better high frequency performance.
So removing a comple of POSCAPS and replacing them with a large number of smaller MLCC capacitors in parallel is maintaining high frequecy filtering effect.

EVGA Says Nvidia RTX 3080 Cap Issues Caused Crashes, Confirms Stability Issues | Tom's Hardware 

[hardcoregames_]

All I know is that eBay prices for RX 5700 XT cards are lower than ever

[black_zion]

The capacitors are not the issue because all 3080 cards, even the Founder Editions, are having instability issues, so they're just a mitigating factor. It's likely a combination of drivers, clock speeds, the underlying architecture, and, of course, the manufacturing process. A new BIOS or softmod reducing performance isn't going to buy them more than a lawsuit...There may be a way for them to fix the issue with a combination of a new VBIOS AND driver softmods if the problem can be patched around, in a sense, if it can be addressed with modification of other variables not normally user accessible, which is a possibility given that none of the reputable review sites encountered any issues when testing...

Look at the bright side though, this should leave all those Aholes who used bots to preorder and grab up every 3080 and 3090 card as fast as possible with a mountain of cards they can't sell.

[colesdav]

Sure - filtering capacitors are there to filter out ripple on VRM output and also high frequency noise from GPU clock and to much less extent, VRAM clock.
Capacitor equivalent circuit = resistance, capacitor and inductance in series.
Above the self resonant frequency of the eqiuivalent circuit,  it becomes an inductor, which is bad news for PSU supply rail noise. 
Lowering the GPU clock frequency is bound to help reduce power supply noise.
There are definitely many factors that may cause the noise, and yes, the capacitors are there to filter the noise as much as possible.

[colesdav]

The power draw and size of that RTX3090 is ridiculous for an Nvidia GPU.
Big Chungus: The RTX 3090 is here... Here we go again... - YouTube 

[black_zion]

I was looking across nVidia's new driver release today, and two things stood out. It appears nVidia is countering AMD's XBOX 1 Year sub with a 1 year sub to their own cloud gaming service, and they are now including an automatic overclocking feature which will run for up to 30 minutes to find out how well your card can overclock, and that's far more comprehensive than AMD Overdrive.

[ajlueke]

With the new cards overclocking headroom at about 5%, that probably isn't a difficult algorithm to write.

[black_zion]

I tried that overclocking algorithm. Even on voltage and power set to full boost, it only gave me 88mhz boost on my core, which is a far sight shorter than manually overclocking...

[pokester]

Yep I messed with it too last night. Mine gave me an extra 78. I have been super stable with an extra 300 on core clock. So this algorithim isn't very aggresive. 

Although as I have found what works for shader cores does not necessarily work on rt cores. For instance the game Control is very sensitive and does not work with overclocks that worked in other games. 

[pokester]

I'm shocked, low stock, who could have seen that coming? LOL

[ajlueke]

"Some Etailer sites went down today because of the number of people trying to purchase an RTX3080"

Do scalper bots count as people?

[colesdav]

Pretty sure there is a massive demand for those Nvidia RTX3080 cards.

[samscot]

Why Nvidia card always damages the board inside the laptop?

[ajlueke]

Looks like NVidia is indeed going to launch higher VRAM versions of the RTX 3080/3070.

https://videocardz.com/newz/nvidia-geforce-rtx-3080-20gb-to-launch-in-december 

Also, the rumor indicates that the RTX 3070 Ti has been canceled.  Which, could just be difficultly getting fully functional GA104 GPUs as the article speculates.

It looks mostly like NVidia's effort to take away a marketing advantage of the RX 6000 series.  Which, if the RTX 3080 and 3070 were already significantly faster than the RX 6000 competition without the VRAM, NVidia likely wouldn't bother. 

[pokester]

I am glad to hear they will have the 16gb 3070. I was not going to consider an 8gb version. 

[black_zion]

Haven't seen anything about the 3070 Ti being canceled, though I would suspect it's just delayed both to bolster 3070/3080 stocks, and to see how the RX 6000 series actually performs so they know how to set the bin levels, though I suspect nVidia doesn't really think RDNA2 is going to be a real competitor, and really nobody knows until they get on the market if they are going to be a serious competitor or not, or just a throw away product like Navi that's priced way too close to nVidia with too many drawbacks to even consider for current nVidia owners to switch.

[pokester]

I did see a video the other night that said the 3070 ti is cancelled but only speculated as to why. Saying they thought it was an issue with getting enough of the chip to be available. Apparently it was going to use the same die as the 3080 not the lower 3070?

[ajlueke]

Well, I was able to snag my first NVidia GPU since my 8800 GTX from Best Buy in the Founders Edition dump.  RTX 3090 arrived in the mail today.  Now the question becomes, do I actually want to keep it?

[pokester]

So you ordered it online or in store and they delivered it?

[ajlueke]

Got it on their online store when NVidia's stock moved over.  Apparently Best Buy cancelled any orders that tried to obtain more than one card.

[colesdav]

Why didn't you at least wait for "Big Navi" or "Bigger Navi" launch?

[ajlueke]

I am intrigued by the smallish PCB.  I am hoping that when the water blocks launch for the FE edition it would be possible to do some powerful small form factor builds.  If not, they still sell for $2,400 or so on eBay.

[colesdav]

OK ... you never know, there migh be an HBM2e version of "Bigger Navi" which should have a R9 Nano / Powercolor RX Vega 56 Red Dragon sized PCB.

However - going for an Nvidia RTX3080 might be the better option until they end up using TSMC 7nm or other better process next year...
I think that is what will happen next, the RTX3000 series on Samsung Process may just be a temporary release.

[ajlueke]

Also, the RTX 3080 with 20GB of VRAM will largely make the 3090 superfluous.  Do you have any info on bigger Navi?  Not seeing anything online about it.

[black_zion]

Remember reports earlier this year point to Ampere only lasting one year with "Hopper" being made on TSMC's 5nm node in 2021, of course if RDNA2 really turns out to be junk they could delay it until 2022 and use the pre-booked space for Ampere Refresh. Given the issues non-professional software has with Ampere currently, it'd make little sense to jump to Hopper next year, so an Ampere Refresh "Super" line would be quite feasible.

https://community.amd.com/thread/252565 

[pokester]

I am pretty positive I read recently on one of the tech sites that said Nvidia told them they would switch Ampere in the coming year to TSMC 7nm. Beyond that no elaboration on exactly what the product would be, but they did go on to say it would be pretty much impossible to be the same as current cards so it would be some sort a gen2 Ampere at that point. I wish I could remember what site to link to. 

[ajlueke]

Yeah, sounds like that is definitely a thing.

https://www.guru3d.com/news-story/nvidia-allegedly-moving-ampere-to-7-nm-tsmc-in-2021.html

[black_zion]

Something the Hexus article on the TSMC move pointed out was that Ampere was first manufactured at TSMC at 7nm, the A100 GPU, and in the report it mentions that the RTX 3000 series will be joining the A100 at 7nm, so there may be so little difference between TSMC 7nm and Samsung 8nm that there won't be any difference between this first batch of Samsung cards and the new TSMC cards, so there may not even be a new SKU.

https://hexus.net/tech/news/graphics/142573-nvidia-ceo-announces-ampere-architecture-a100-gpu/

[pokester]

I would bet that architecturally there is no difference. I would think that the node change however could change power draw and possible a little IPC. 

[black_zion]

It will give them a chance to refine some things, but a minor revision at most, to fix the voltage and frequency issues which are now being addressed in drivers.

[hardcoregames_]

ajlueke wrote:

Yeah, sounds like that is definitely a thing.

https://www.guru3d.com/news-story/nvidia-allegedly-moving-ampere-to-7-nm-tsmc-in-2021.html

I saw a lot of posts on Asian forums. Samsung has not been able to keep up with the market demands which is why TSMC has cut wafer costs on their N7 line which is what AMD was using earlier. Apple was using the N5 line.

[black_zion]

Check this out: XFX has sold their entire Navi stock (allegedly) to cryptominers in China. Think it's a good reason to axe XFX off the list of brands to consider. Though AMD likely doesn't care, they should be pressed to axe them as an AIB.

According to MyDrivers, XFX sold pretty much all of their allotment of chips of the RX 5800, RX 5700 XT, RX 5600 XT and even the RX 590 to GPU miners during the month of September 2020. This isn't a random month by the way, cryptocurrencies like Etherium have been a surge in interest and rallied almost 100% during the last few months. As the price of these cryptocurrencies goes up, so does the interest in GPU mining. Cryptocurrencies are forecasted to go even higher in the coming year having finally recovered from the bear run that began with the 2018 crash so if things get even more lucrative for miners - the strain is going to start extending to other AIBs and even NVIDIA GPUs.

https://wccftech.com/xfx-allegedly-selling-almost-all-amd-radeon-rx-5000-series-gpus-to-mining-farms/

[hardcoregames_]

black_zion wrote:

Check this out: XFX has sold their entire Navi stock (allegedly) to cryptominers in China. Think it's a good reason to axe XFX off the list of brands to consider. Though AMD likely doesn't care, they should be pressed to axe them as an AIB.

According to MyDrivers, XFX sold pretty much all of their allotment of chips of the RX 5800, RX 5700 XT, RX 5600 XT and even the RX 590 to GPU miners during the month of September 2020. This isn't a random month by the way, cryptocurrencies like Etherium have been a surge in interest and rallied almost 100% during the last few months. As the price of these cryptocurrencies goes up, so does the interest in GPU mining. Cryptocurrencies are forecasted to go even higher in the coming year having finally recovered from the bear run that began with the 2018 crash so if things get even more lucrative for miners - the strain is going to start extending to other AIBs and even NVIDIA GPUs.

https://wccftech.com/xfx-allegedly-selling-almost-all-amd-radeon-rx-5000-series-gpus-to-mining-farms...

Riding on the coattails of the digital coins is how I am able to afford to have a large collection of video cards.

My RX 480 was an ex mining card, I got it dirt cheap and the card still works fine. It seems that Sapphire is invincible.

I also got a dirt cheap price for a GTX 1060 but that card needed some work on it as the thermal material fried.

Power supplies are cheap, I picked up a Corsair RM650 for US $95 shipped brand new. This is the new white label model.

[black_zion]

Here's some interesting results from TomsHardware's test of Marvel's Avengers DLSS 2.1. As DLSS becomes more common, the extra boost is going to put it quite a bit above RDNA2 offerings, and as we know DLSS Quality mode is quite good looking.

AMD needs to factor this into their pricing decisions, because if with one toggle the competition can increase performance by 20% (or more), that's going to make their "6900XT is faster than the 3080!" statement quite hollow unless it is priced at least 20% less than the 3080, and given their price structure on Navi as well as with the upcoming Ryzen 5000 series, I don't see them doing it.

https://www.tomshardware.com/news/marvels-avengers-dlss-21-support-performance-tested