nVidia Ampere missing one feature from Turing

[Haider]

Maybe it is that easy, but games aren't doing it is the underlying point.

So, again, this rant is only applicable outside of gaming, correct?

Here's a good article for you, Haider - https://www.tomshardware.com/news/nvidia-to-support-nvlink-only-for-rtx-3090

Until Rendition brought Verite, 3Dfx Voodoo, Videlogic/Imagination Technologies PowerVR PCX1 there were no 3D accelerated games. The hardware came and then the software followed. Pretty much the same story with hardware T&L, tile-based deferred rendering on PCX1, ray tracing... Now if you provide the features then PC can utilise the memory of two cards for very quality texture maps and models. Same problem with hydrogen cell cars, no hydrogen fuel stations

- - - Auto-Merged Double Post - - -

So this applies to to the 5 folks who still SLI? Thankfully I am not one of them..I never bothered with Sli...



Also no one should be buying even one 2080 Ti' at this point let alone two..Get one 3090 with 24 GB of memory.

I've seen two ROG 2080Ti OCs for £600 a piece...3090??? Remind me again isn't that the new Fermi? Heat issues I hear...

 

[EarthDog]

Until Rendition brought Verite, 3Dfx Voodoo, Videlogic/Imagination Technologies PowerVR PCX1 there were no 3D accelerated games. The hardware came and then the software followed. Pretty much the same story with hardware T&L, tile-based deferred rendering on PCX1, ray tracing... Now if you provide the features then PC can utilise the memory of two cards for very quality texture maps and models. Same problem with hydrogen cell cars, no hydrogen fuel stations

What's your point here? I'm struggling to find the relevance. That was over 20 years ago and games won't suddenly use it. So again, thanks for the warning but this isn't for remotely for gaming now or in the near future.

 

[Haider]

Only just saw this. The memory pooling over nvlink I took to be more useful for some creative uses. They're less demanding than gaming on bandwidth, but depending on what you're doing, could use the quantity.

Also while we're operating on rumours, AMD is supposedly going higher vram on some of their to be announced cards, and nvidia has another wave waiting in the wings to counter that. There may be some higher vram options in the not too distant future. Still not two cards, but if you get double a single card on a single card, would that suffice?

And in synopsis isn't that just Jensun, the greedy badger. Have higher spec cards waiting in the wings ready launch so he can mug people off first with lower spec cards then take them to the cleaners a second time. He must come from a long line of opium dealers...

- - - Auto-Merged Double Post - - -

What's your point here? I'm struggling to find the relevance. That was over 20 years ago and games won't suddenly use it. So again, thanks for the warning but this isn't for remotely for gaming now or in the near future.

The fact that they supported it (NVLink) for one generation then drop it the next means that it never had time to picked up and utilised by developers. Give it time and it will be adopted like ray-tracing...

 

[EarthDog]

The fact that they supported it (NVLink) for one generation then drop it the next means that it never had time to picked up and utilised by developers. Give it time and it will be adopted like ray-tracing...

I won't be holding my breath.

 

[dfonda]

Until Rendition brought Verite, 3Dfx Voodoo, Videlogic/Imagination Technologies PowerVR PCX1 there were no 3D accelerated games. The hardware came and then the software followed. Pretty much the same story with hardware T&L, tile-based deferred rendering on PCX1, ray tracing... Now if you provide the features then PC can utilise the memory of two cards for very quality texture maps and models. Same problem with hydrogen cell cars, no hydrogen fuel stations

- - - Auto-Merged Double Post - - -



I've seen two ROG 2080Ti OCs for £600 a piece...3090??? Remind me again isn't that the new Fermi? Heat issues I hear...

You mean used? That's not really Apples to Apples but OK. I thought you were talking about new. If its is new I am curious to see the link. I do not see new prices on old GPU's dropping at all.

1080 Ti's are selling for 1K+ on Newegg...Which is crazy for a 3 year old GPU...I bought mine for $699.

 

[mackerel]

And in synopsis isn't that just Jensun, the greedy badger. Have higher spec cards waiting in the wings ready launch so he can mug people off first with lower spec cards then take them to the cleaners a second time. He must come from a long line of opium dealers...

You very clearly have your very particular viewpoint and needs, but let's be realistic here. The currently announced Ampere cards should cope with any currently existing game comparable or better than Turing cards. AMD are expected to do what AMD does, try to shift the industry perception in their favour. The pushed core counts with Ryzen. Now while not officially announced yet, they intend to push VRAM quantity as a selling point. It will cost more to put it on cards and for most people will not give any short term benefit. However, nvidia do not want to be seen lagging in that area hence the cards on standby if AMD follow through with the rumour indications.

The fact that they supported it (NVLink) for one generation then drop it the next means that it never had time to picked up and utilised by developers. Give it time and it will be adopted like ray-tracing...

nvlink is more a pro level feature. It's presence on gaming cards is really not significant apart from enabling SLI for the three people out there who still do it. If you really need nvlink, get the cards that support it. It is irrelevant to the main market.

 

[don256us]

Hi,

Just a heads up to everyone nVidia is back up to it's old tricks again; they really are a disgrace. It's true what they say, you can have all the money in the world but you can't buy class; Jensun is a true spiv. Ampere is worse than Turning. nVlink on RTX 2070 Super, 2080, 2080TI allows you pool the memory on two cards together to get 16GB or 22GB for applications. Not so now on the RTX 3080. They push you to buy the Turing (RTX) Titan replacement RTX 3090...Before the nVidia fanboys start foaming at the mouth, I currently have an Inno 3D iChill X3 GTX 1070...


Thanks
Haider

Good info that I didn't know about and good info from the ensuing conversation. Thank you.

 

[Haider]

Good info that I didn't know about and good info from the ensuing conversation. Thank you.

As they say on the other side of the pond, it's all good in the hood

 

[PanteraGSTK]

You very clearly have your very particular viewpoint and needs, but let's be realistic here. The currently announced Ampere cards should cope with any currently existing game comparable or better than Turing cards. AMD are expected to do what AMD does, try to shift the industry perception in their favour. The pushed core counts with Ryzen. Now while not officially announced yet, they intend to push VRAM quantity as a selling point. It will cost more to put it on cards and for most people will not give any short term benefit. However, nvidia do not want to be seen lagging in that area hence the cards on standby if AMD follow through with the rumour indications.



nvlink is more a pro level feature. It's presence on gaming cards is really not significant apart from enabling SLI for the three people out there who still do it. If you really need nvlink, get the cards that support it. It is irrelevant to the main market.

Exactly. SLI was a nice feature when I got my 8800GT back in the day and a second one went on sale. It was neat, but there were quite a few games that didn't support SLI at all, not to mention the fact that scaling was never 100%. I MAYBE got an average 25% boost across the board with my games, but some did VERY well. Those were too few to worry about, so while I got a good performance boost, it isn't something I'd ever do again.

Hell, I had a 970 as a physx card up until I got my Ryzen system because in some VERY heavy physx games I got a decent 10-20fps boost with the 970 paired with my 2070.

The 3080 I plan to grab in the morning (hopefully) will provide more than enough power for any game I can throw at it. It'll be the first 80 series GPU I've purchased in well over a decade. I think my 3440x1440 monitor will do just fine for many years with that card. Just like my 8800GTs did me way back when.

nvlink isn't for "regular" gamers at all. The cards that support it now are good for 99% of people using a single card. Using more than one is just bragging rights and benching. The people that use it upgrade a lot more often than the rest of us so long term isn't their goal like it was with me years ago.

 

[Woomack]

Somehow I find this thread pointless, not to mention that the title is highly confusing for all who seen test results in actual games

 

[Automata]

Somehow I find this thread pointless, not to mention that the title is highly confusing for all who seen test results in actual games

I updated the thread title to make it less sensationalist.

 

[RJARRRPCGP]

1080 Ti's are selling for 1K+ on Newegg...Which is crazy for a 3 year old GPU...I bought mine for $699.

Yikes! Like it's 2018, especially if you got yours for a lot less during the year of launch or a year after!

Looks like the best I will have for a while, will be the GTX 1660 Super from Zotac, on my other Ryzen rig, which is a Ryzen 5 2600.