So over much time, I have been thinking about Intel graphics cards since before Intel themselves took it seriously..
In short.. Intel need to build a consumer-ship from scratch, and on every side they have the potential to be squeezed out of the market.
So with that in mind, I have some.. recommendations for the Intel Graphics Department..
1. you need a catchy name for your cards that has nothing to do with your other products.. you already have 'Optaine'.. perhaps 'nitro' ? 'Intel Nitro Graphics Cards'.. it makes it stand out as its own thing, if you just called it Optaine, then it's likely to get drowned when people search for info and pricing when shopping for one..
2. Sad but true.. it'll have to be a mid range graphics card.. just throwing out some lowend thing is not very smart, or trying to compete with the ultra high end is equally not as smart..
a.on the low end the market is full of both new and second hand cards that are more than capable
b. on the high end your drivers and driver Devs need time to mature and handle games better, just throwing a load of power at something doesn't translate well to 'good gaming' just ask AMD who have had endless woes with their drivers.
What imho Intel needs to do is put out something.. around 2000 shader cores.. seeing its Intel.. 2048 shader cores because its nerdy :)) 120 TMU 64 ROUs with 12GB of GDDR6 .. but with a pleasant caveat I'll come to in a moment.
(update 31/May/2022 : 6144 shader cores, 384 TMU, 194 ROUs, 48 RTC with 16GB GDDR6 for mid level, £/$200.. top level, double the above: 12288 shader codes, 768 TMU, 184 ROUs, 92 RTC with 24 to 32 GB GDDR6 £/$400 )
Why have the media/game data in a separate device? Why not throw an integrated 1TB-4TB stick of PCIE Optaine as part of the graphics card? Direct on the PCB, with Optaine chips specially designed not to be Graphics RAM, but as fast as Graphics RAM.. how? .. I would suggest that the card has its own PCIE bus lanes and mini CPU.. 'system on card' but with the entire purpose of creating graphics. Some easier parts could be sent to the mini-CPU (2c/4t Low Power cores recommended) and anything that required heavy lifting could be sent to the mainboard CPU.
Imagine those with weak CPUs.. they could buy your card without the worry of getting the worst possible performance if their CPU is pants.. as some of what is done will go to the on-card mini-CPU.
All those people who want an easy life.. plug in and play, if they have an old desktop with 2 cores but its all they have, they could buy your card and job done. Dad or Moms work PC? this card = job done just by plugging it in.
Even perhaps cards that have no need for VRam.. but read and process the data directly, from storage from the 'system on card'.
3. It is has to be CHEAP.. unfortunately for you, you can't go the expensive route.. because you don't have driver confidence, and imho that will take 5 to 10 years of successful products including the drivers themselves.
NV is an excellent example of what a company can do if they pay attentions to gamers needs and put much effort in the features of their interface (NVCP). Not being able to make custom res's is a deal breaker for me.. something small like that even can make life just that bit nicer.
But NV isn't perfect.. oh they are the best in the business.. but still they have driver problems with each release.
And so with that explained.. Intel is new, doesn't matter your rep in the CPU market, this is GPUs and here its cut throat.. and its been mostly NV doing the cutting of the throats.
Being new means.. you need a competent product that is not irrelevant due to other companies offerings. You NEED to offer much more value for money than either competitor.. or AMD will swallow your marketshare whole.
So I suggest putting out the above card for around $250-300/£250-300. (update May 2022:mid level £/$200)
Yes you will make a loss.. but you are buying customers loyalty in the hopes they try>enjoy>recommend.
Don't worry NV has done this before to when they hit a slump or messed up so badly everyone started talking about jumping ship to AMD..
What place is there in the market for Intel?
To begin with.. not much, however if you can beat AMD on price vs performance AND have competent drivers with decent features.. releasing a new card every two years (four years for new refresh).. ohw maybe need to explain this..
Every two years GPU manufacturers release 'new' cards.. they take the higher cards from their previous line and rebrand them as mid level cards while they get ready for totally new cards in two years.. so for Intel you would be starting with a blank slate and potentially only one or two cards so you couldn't do it this way until you had 6-8 years under your belt..but you could do an initial release and then 'beef up' the cards specs.. all the time in the interim tightening the drivers with feedback from customers..
Flip! no one will tell you this, but customer feedback is an essential component of GPU production.. because IF you listen to their feedback.. however obnoxious or nasty.. they may have a grain of truth in what they are trying to say.
Look at AMD.. ignores its customer feedback, drivers are usually awful. So they end up just throwing power at it and hope it works. That's not a good way to keep your fans.
For CPUs I was an AMD guy for many years, I used NV for Graphics cards.. I tried the AMD + AMD thing.. welp doesn't work if your drivers are nonsense. I bought a HP Workstation after a water incident.. with my twin Xeon X5650s (at the time of writing*) and am happy with them. I also rock a NV 2080ti.. but I'm not your target customer base just yet.. people who can afford this stuff will buy NV. They aren't coming over to be the test rabbits for your experiment..
Aim too high and no one will care. Aim too low and no one will care. Harsh but I imagine no one else will tell you this stuff.. and by the amount I have written.. I think you can see I have hopes for you in this marketspace, Intel.
I digress..
Why have the media/game data in a separate device? Why not throw an integrated 1TB-4TB stick of PCIE Optaine as part of the graphics card? Direct on the PCB, with Optaine chips specially designed not to be Graphics RAM, but as fast as Graphics RAM.. how? .. I would suggest that the card has its own PCIE bus lanes and mini CPU.. 'system on card' but with the entire purpose of creating graphics. Some easier parts could be sent to the mini-CPU (2c/4t Low Power cores recommended) and anything that required heavy lifting could be sent to the mainboard CPU.
Imagine those with weak CPUs.. they could buy your card without the worry of getting the worst possible performance if their CPU is pants.. as some of what is done will go to the on-card mini-CPU.
All those people who want an easy life.. plug in and play, if they have an old desktop with 2 cores but its all they have, they could buy your card and job done. Dad or Moms work PC? this card = job done just by plugging it in.
Even perhaps cards that have no need for VRam.. but read and process the data directly, from storage from the 'system on card'.
3. It is has to be CHEAP.. unfortunately for you, you can't go the expensive route.. because you don't have driver confidence, and imho that will take 5 to 10 years of successful products including the drivers themselves.
NV is an excellent example of what a company can do if they pay attentions to gamers needs and put much effort in the features of their interface (NVCP). Not being able to make custom res's is a deal breaker for me.. something small like that even can make life just that bit nicer.
But NV isn't perfect.. oh they are the best in the business.. but still they have driver problems with each release.
And so with that explained.. Intel is new, doesn't matter your rep in the CPU market, this is GPUs and here its cut throat.. and its been mostly NV doing the cutting of the throats.
Being new means.. you need a competent product that is not irrelevant due to other companies offerings. You NEED to offer much more value for money than either competitor.. or AMD will swallow your marketshare whole.
So I suggest putting out the above card for around $250-300/£250-300. (update May 2022:mid level £/$200)
Yes you will make a loss.. but you are buying customers loyalty in the hopes they try>enjoy>recommend.
Don't worry NV has done this before to when they hit a slump or messed up so badly everyone started talking about jumping ship to AMD..
What place is there in the market for Intel?
To begin with.. not much, however if you can beat AMD on price vs performance AND have competent drivers with decent features.. releasing a new card every two years (four years for new refresh).. ohw maybe need to explain this..
Every two years GPU manufacturers release 'new' cards.. they take the higher cards from their previous line and rebrand them as mid level cards while they get ready for totally new cards in two years.. so for Intel you would be starting with a blank slate and potentially only one or two cards so you couldn't do it this way until you had 6-8 years under your belt..but you could do an initial release and then 'beef up' the cards specs.. all the time in the interim tightening the drivers with feedback from customers..
Flip! no one will tell you this, but customer feedback is an essential component of GPU production.. because IF you listen to their feedback.. however obnoxious or nasty.. they may have a grain of truth in what they are trying to say.
Look at AMD.. ignores its customer feedback, drivers are usually awful. So they end up just throwing power at it and hope it works. That's not a good way to keep your fans.
For CPUs I was an AMD guy for many years, I used NV for Graphics cards.. I tried the AMD + AMD thing.. welp doesn't work if your drivers are nonsense. I bought a HP Workstation after a water incident.. with my twin Xeon X5650s (at the time of writing*) and am happy with them. I also rock a NV 2080ti.. but I'm not your target customer base just yet.. people who can afford this stuff will buy NV. They aren't coming over to be the test rabbits for your experiment..
Aim too high and no one will care. Aim too low and no one will care. Harsh but I imagine no one else will tell you this stuff.. and by the amount I have written.. I think you can see I have hopes for you in this marketspace, Intel.
I digress..
What place is there in the market for Intel?
To begin with.. not much, however if you can beat AMD on price vs performance AND have competent drivers with decent features.. releasing a new card every two years (four years for new refresh)..increasing your performance 10-20% every four years for the same amount of money for the same price point.
This is where Intel could snuggly fit into the market.. getting enough press about the gaming performance itself to salivate consumer appetites, all free press garnering curiosity and interest.. but something deeper.. prestige that people can tell other people they own one and have a conversation about owning one.. maybe I'm over stating it..yes I am but its illustrative of something Intel has NEVER needed to do.
Create a userbase and brand recognition from scratch.
Dava
To begin with.. not much, however if you can beat AMD on price vs performance AND have competent drivers with decent features.. releasing a new card every two years (four years for new refresh)..increasing your performance 10-20% every four years for the same amount of money for the same price point.
This is where Intel could snuggly fit into the market.. getting enough press about the gaming performance itself to salivate consumer appetites, all free press garnering curiosity and interest.. but something deeper.. prestige that people can tell other people they own one and have a conversation about owning one.. maybe I'm over stating it..yes I am but its illustrative of something Intel has NEVER needed to do.
Create a userbase and brand recognition from scratch.
Dava
------------
* Now I have a HP Z840 twin Xeon E5-2660v3's 20c/40t, DDR4-2133 32GB, 2080Ti, load of PCIE and SSD and couple of HDD, totalling idk 15TB? 4K TV and a small HP 900p monitor.
Old set up was.. twin Xeon X5650, 48GB of DDR3 1066, 2018ti/1070 and drives/monitors as above.
