News Intel GPUs - Intel launches A580

[bullzz]

Knowing Jensen, I wouldn't rule out the possibility that all their top developers have strict non-compete clauses preventing them from working for a competitor for a year or more.

Jensen cannot enforce non-compete in California

 

[Grabo]

July I guess and maybe December : ) Anyways, and sorry in advance if this was discussed vigorously on some past pages of this thread - has there been any indication of hybrid SLI or some other benefit of running an Arc on an Intel platform?

 

[IntelUser2000]

July I guess and maybe December : ) Anyways, and sorry in advance if this was discussed vigorously on some past pages of this thread - has there been any indication of hybrid SLI or some other benefit of running an Arc on an Intel platform?

The mobile platforms have Deep Link, which allows dynamic sharing of TDP and improving performance in media.

I don't expect anything to improve 3D performance. It seems the days of SLI are over.

 

[LightningZ71]

DX12 has provisions for mGPU setups, but the developers have to enable it. Ashes of the singularity was famous for supporting it. Traditional SLI is not really workable in the era of deffered rendering, so it has to be enabled deep in each game. Or, so I've read...

 

[LightningZ71]

Give them all an Arc GPU as a signing bonus- gives them a motivation to improve the drivers!

Replace their own GPUs with ARC ones and weld their cases shut. Force them to fix the drivers or suffer...

 

[ultimatebob]

Replace their own GPUs with ARC ones and weld their cases shut. Force them to fix the drivers or suffer...

Hey now, we want the good driver developers to stick around! I like my idea of bribing them with a Tesla better. I mean, this is California, Tesla's are like Honda Civics anywhere else.

 

[IntelUser2000]

Ashes of the singularity was famous for supporting it. Traditional SLI is not really workable in the era of deffered rendering, so it has to be enabled deep in each game. Or, so I've read...

Support isn't enough of course. GTA V scaled fantastically with core count but performance sucked anyway, making you wonder what they got in the first place. It's better to not scale and perform higher.

Multi-GPUs are pretty much dead. Multi-tile/die GPUs on the other hand are the future.

 

[KompuKare]

hmm, considering that the architecture is essentially the same as Xe in their iGPUs, which has been out well over a year, and still has issues in multiple games, I don't exactly have a lot of confidence that they'll get the drivers fixed in 60-90 days...

Intel has to PROVE to me that they won't suddenly decide "Hey, this is hard, lets walk away from it and pretend that it never happened" like they have with several video products over the last few years.

Would be very nasty for current paying customers, but we can hope that they have deliberately withholding Xe fixes to put all the good stuff into the ARC launch drivers?

A bit dodgy as then they'd get no proper beta testers, but still.

That's the most generous / hopeful interpretation. AMD used to try and do something special once a year for the autumn drivers, didn't they? Although that was more about new features or UI changes, less about game fixes.

 

[IntelUser2000]

Would be very nasty for current paying customers, but we can hope that they have deliberately withholding Xe fixes to put all the good stuff into the ARC launch drivers?

Better than nothing since the alternative is not improving at all in any product category.

I do expect similar though. You want some fanfare for your new product. For example with Xe launch they "rewrote" the DX11 portion for lower CPU utilization.

 

[igor_kavinski]

A bit dodgy as then they'd get no proper beta testers, but still.

They will promote the hell out of it, move substantial units and if enough land in the hands of actual gamers, well, THOSE would be the unwitting beta testers. And if they just accumulate in the basements of scalpers, tough luck for them. Their driver quality won't improve that much and that quickly then.

 

[Asterox]

It is not dead, hm ARC is just drunk.

https://twitter.com/x/status/1492179543590907914

 

[Fanatical Meat]

Still don’t get the hate. A new player is bringing new & more products to a category that desperately needs inventory and a third choice.
Why complain about drivers that have yet to be released.

 

[igor_kavinski]

Under current conditions, they can just put the cards out right now without releasing the drivers and they will still fly off the shelves. Gamers will be content dreaming about playing their favorite game with their expensive graphics card with the imaginary drivers.

 

[DrMrLordX]

Still don’t get the hate. A new player is bringing new & more products to a category that desperately needs inventory and a third choice.
Why complain about drivers that have yet to be released.

They are taking their sweet time, don't you think? Intel delays everything these days.

 

[Leeea]

Still don’t get the hate. A new player is bringing new & more products to a category that desperately needs inventory and a third choice.
Why complain about drivers that have yet to be released.

Well, it is like this.

They stirred up the hype pot and flipped the burner to high!

and then they just forgot about it.

So the hype pot is boiling over.

 

[Shivansps]

Ok i just tested today with the feb 7 driver (30.0.101.1340) on the DG1 and they fixed all the issues i saw on DX12/Vulkan. But its also slower than before.

This is SOTTR @1080P LOW

1121 DX11

https://imgur.com/BnRwKzQ

1340 DX11

1340 DX12

Maybe there is something weird going on with the 12th gen, as i didnt have the same components to test.

And thats the other thing i wanted to tell, it seems that the DG1 firmware is present on all Asus motherboards for 11 and 12th gen, for example, a B460M-A R2.0 dosent work with factory bios (its way too old) but the newerest bios on the website supports the DG1. The Asus H610M-E D4 factory bios already supports the DG1 as well.

Im going to coment this on Steve video test of the DG1, maybe he can test again with this new info. But i dont think anyone is going to read that.

EDIT: its the GPU clock, this card only maintains about 900mhz at 25W... the other card was able to maintain 1250...

WTH

Maybe, just maybe, the h610 dosent actually have the dg1 firmware and is booting with the firmware meant for alder lake igp?

 

[IntelUser2000]

Maybe there is something weird going on with the 12th gen, as i didnt have the same components to test.

Interesting. Why didn't you just update the drivers for the 10100F system?

900MHz clock reminds me of when Ivy Bridge iGPU had a bug where it would sometimes get stuck at 900MHz GPU clocks rather than go up to 1.15GHz.

The visual artifacts being gone is good news though.

 

[Dayman1225]

New video from RedGamingTech on Alchemist claiming major delays and performance issues in games

 

[DrMrLordX]

@Dayman1225

Hope that isn't true, but . . . it sounds believable.

 

[IntelUser2000]

New video from RedGamingTech on Alchemist claiming major delays and performance issues in games

Man I want to emphasize that Intel should really stick to the dGPU venture until they find success because it's not looking good for Alchemist if that's true. If the GPU prices drop, Eth goes PoS, and supply improves because lockdowns finally end(it should have 2 years ago) then their incentive to continue will face real roadblocks.

Being slow in DX11 is weird since @Shivansps test is showing DG1 doing better on DX11.

"80% hardware, 20% software" also supports my suspicion that Intel gets blamed too much for bad drivers when the culprit was often the hardware itself.

 

[VirtualLarry]

If, in it's current state, it is doing well in "synthetics", might not it also do well for mining? C'mon, Intel. Release "ARC Alchemist, mining edition". Especially if the first batch of chips off the line has hardware defects that affect gaming, but hopefully not mining. (Just change the label on the shroud, Intel, and release them for the miners!)

 

[igor_kavinski]

"80% hardware, 20% software" also supports my suspicion that Intel gets blamed too much for bad drivers when the culprit was often the hardware itself.

That lying Raja. 80% is his fault. He said that software is really hard. Oh, yeah it's gonna be hard when your hardware still has bugs!

 

[NTMBK]

If, in it's current state, it is doing well in "synthetics", might not it also do well for mining? C'mon, Intel. Release "ARC Alchemist, mining edition". Especially if the first batch of chips off the line has hardware defects that affect gaming, but hopefully not mining. (Just change the label on the shroud, Intel, and release them for the miners!)

*pointedly taps the thread title*

 

[IntelUser2000]

If, in it's current state, it is doing well in "synthetics", might not it also do well for mining?

I'm not against miners. I used to be one. But I hope Intel can succeed based on the merits of the product, which is 3D graphics.

Also: https://www.tomshardware.com/news/fanless-ethereum-mining-box-promises-65-mhps-at-30w

1TB/s bandwidth at 65MH/s. Which is pretty low since a real 1TB/s bandwidth should allow a theoretical 120MH/s performance and even at 90% efficiency that's almost 110MH/s.

So it doesn't mean that it'll necessarily do well because it does well in synthetics. Although in the particular case of the fanless box the chip itself might not provide enough performance to fully take advantage of the memory bandwidth, making a memory controller that fully utilizes memory is hard, so maybe the practical bandwidth is something like 0.6TB/s.

The theoretical maximum bandwidth for the DG2-512 using 18GT/s memory is 576GB/s, which puts theoretical Ethereum performance at 72MH/s. The actual performance might easily end up being 40-45MH/s though.

 

[Shivansps]

Being slow in DX11 is weird since @Shivansps test is showing DG1 doing better on DX11.

yeah that is weird, that is not what im seeing. And it is not what i would expect, DX11 driver perf depend almost enterely on Intel and they already been working on it for years, DX12/Vulkan on the other hand, may need game side optimizations that no one ever cared to do for intel igps that couldnt run the game anyway.

Also it is not believable that DX11 performance can hold a new GPU launch in 2022.

and keep in mid that DX11, specially W3 perf is way higher than i would expect for this gpu.

Interesting. Why didn't you just update the drivers for the 10100F system?

900MHz clock reminds me of when Ivy Bridge iGPU had a bug where it would sometimes get stuck at 900MHz GPU clocks rather than go up to 1.15GHz.

The visual artifacts being gone is good news though.

I did just found out that DG1 work with new website bios on some Asus motherboards and i put a DG1 on a H610 just to see what would happen, and that PC already had the SOTTR/RDR2/W3 installed because i was testing the RX6500XT. I didnt have the time to test it again on a 10100F.