They are 100 % putting pressure on the GPU market. If they were not here, how do you think it would look like? How do you think Nvidia would price their GPUs? How do you think they would configure their chips? Do you really think they would push their chips out of the efficiency sweet spot like they are doing currently? They are not making huge power hogs for fun.
Market share is pretty irrelevant here, it does not counter the argument that AMD is putting pressure on the market. They have a good arcitecture that offers good bang for the buck and is pretty efficient too. The RX 7000 will progress forward in a manner that Ryzen progressed forward. Nvidia can not do what they want, they need to moderate how they price their products to not be overly ridiculous and make it too obvious they are overpriced. AMD holds them back in this regard.
Market share also does not necessarily tell you what the best products are. I am not talking sales numbers here, that is for narrow-minded share holders, I am talking choices that people have available. We would not have the current choices if AMD was not here, it is very naive to think so.
You will have to take it up with consumers why they are making uninformed decisions and enabling Nvidia to set ridiculous prices for mid-range chips. It takes people several years to get detached from their reputation-based purchases.
And AMD is actually selling great graphics cards at the moment. At least in relation to how messed up the GPU market has become (thank you, Nvidia and thank you, consumers/miners).
I am on Ubuntu, but my point is more like compared to like even Nvidia on Linux (letās not compare with what you get on Windows), thereās no all in one place to get things done.
Thereās CoreCtl and Radeon Profile (the latter allows for even more finetuning) but unlike the Nvidia Control Panel on Linux, it doesnāt help you to toggle various graphical settings all in one app (resolutions, refresh rates, power levels, fan curves), thus I said it wasnāt as polished. Even to expose options in CoreCtl or Radeon Profile requires additional toggles at the bootloader when probably the software could give a prompt whether you wish to add those flags.
That being said, the ROCm for AMD (and hopefully Intelās oneAPI) has helped further to offload reliance on CUDA towards OpenCL, especially the 5.3 release.
AMD is good with hardware but the software side is still lacking (but improving rather slowly).
But the fact that I much more prefer to be on Linux despite these things sort of says a lot about how much crap there is to deal with on Windows (I know this is a subjective statement, but there are objectively a bunch a problems on Windows indeed). I have been so much happier after switching everything to Linux and my PCs work with me, not against me.
As growth on Linux continues IĀ“m excited about what the future holds. I for one, will not be stuck in a catch 22. There is no catch 22 if people just switch. I will never go back to a proprietary bloat prison with ads where you are the product and arenĀ“t even in charge of your own system.
Youāll probably get worse power efficiency from chiplets! Although if the rumours are correct, weāll see TSMC nodes on some of the tiles which may help!
AFAIK Meteor Lake has twice the amount of GPU Cores as Raptor/Alder/Rocket Lake! So hopefully a bit more than double the performance! Honestly with Intel Drivers I canāt say itās very useful thoughā¦ Mostly hoping for a better media engine, 11th gen video playback power consumption is very highā¦
Just adding another anecdotal data point to the discussion. Iāve been checking in about the availability of AMD powered framework laptop for the last year or so. I am sold on rest of the idea of framework laptop and Iāll buy the amd version the day the store allows me to order one.
Slightly off topic but I am also considering buying a DIY edition without the mainboard (will ask support). Lenovo sent me a new mainboard for my thinkpad x13 by mistake (instead of just a usb breakout board). And it has an AMD cpu and ram embedded on it. So I might be able to make a franken laptop with frameowork IO and thinkpad/amd brains. Need to look into display, keyboard and battery connector compatibility between the two laptops.
As much as I think this sounds like a fun and cool project IĀ“m afraid there are many details that has to fit in order for this to work.
It would be a miracle if screw holes line up. The cooling solution is likely differently designed on the ThinkPad motherboard. The engineers design this in a way that the airflow is guided out through the vents. Then there are potential protrusions on the chassis or motherboard.
So when thatĀ“s sorted you need to know that all the ribbon cables have the same size, pinout and compatibility with whatever it connects to on the motherboard. The monitor is likely the easiest one as eDP ports are standardized. There are 30-pin connectors and 40-pin connectors. The latter is usually reserved for higher resolution panels.
Though, the keyboard, trackpad, fingerprint reader and keyboard backlighting I donĀ“t know if they are standardiized or even use a common protocol.
Then the IO ports on the motherboard needs to line up. The size of the board is likely similar to the one in the Framework laptop. The aspect ratio of the screen is taller and narrower than usual, so will the ThinkPad board even fit? I havenĀ“t compared dimensions.
So bottom line, it might be a pain in the butt and too much work and hassle to even work. Perhaps it would be easier to just mold a composite chassis yourself than to get the Framework Chassis to work.
IĀ“d love to know the answer to these questions, though. ItĀ“s an interesting challenge, indeed.
Yeah, but they will make other improvements that will compensate for it.
The advantages of chiplets are just too good to not take advantage of. Wafer yield and chip design flexibility is massively beneficial. Chiplets is one of the reasons why AMD has been so competitive. It was a brilliant move. Lisa Su really lead the company in the right direction.
Intel definitely agrees about chiplets even though they tried to make fun of it some years back in some embarrassing slides their marketing teams made. xD
AMDās mobile chips have never been chiplet based due to the power defficiency!
I have a hunch chiplets will be limited to desktops, although EMIB is more power efficient than Infinity Fabric.
Seems a bit unrealistic, you canāt simply decide to add on 50% more efficiency in one generation becase you feel like it! These designs take a long time!
Would be nice though, Iām definetly excited for Meteor Lake!
I was under the impression that AMD has chiplets for desktop, server and mobile. I actually had to check that Mobile is really monolithic. So they only share chiplets with their desktop and server CPUs.
Wow does that mean the efficiency focus on Meteor Lake laptop chips would be negated by the chiplet designs?
I donāt believe 14th Gen was targeted at efficiency! That should be more like 16th Gen if I remember? (Arrow Lake?)
Although 14th Gen should bring an significant graphics uplift!
It will not. It retains the same graphics configuration. Meteor Lake (14th gen) is targeted at mobile, Arrow Lake (15th Gen) is targeted to take the mobile efficiency and mobile graphics crown. Arrow lake is expected to have a massive iGPU with hundreds of iGPU cores compared to the 96 we have now on top-end i7s. Lunar Lake is 16th gen, not much is known about Lunar Lake.
Looks like youĀ“re right about AMD mobile being monolithic-only.
I donĀ“t think chiplets will be a desktop-only thing. It makes more sense the bigger the die is, though, so definitely a must on bigger designs. ItĀ“s also confirmed that Meteor Lake will be chiplet-based.
Die stacking will be quite interesting, we already see great benefits of that in other type of chips.
Yep, I know.
Unrealistic? They are already improving efficiency for every new generation.
Efficiency isnĀ“t the only factor at play here. Yield means a lot for production cost and chiplets are a very simple way of massive improving that aspect as well as adding flexibility in designs. It also enables so many more SKUs, especially scaled-up designs.
With chiplets, what you are essentially doing is cutting what would be a single die into, say, 4 dies. Whereas a defect on that monolithic die would make the entire die non-functional (0 % yield from that single die) youĀ“d have 3 out of 4 functioning chiplets (which translates into 75 % yield from the same area). DonĀ“t translate these numbers over to per wafer yields, itĀ“s simply to demonstrate the principle of why chiplets are beneficial.
ItĀ“s pretty obvious how advantageous this is for yields and design flexibility. It also explains why AMD can go crazy with Threadripper and undercut IntelĀ“s pricing by half for something that perform better.
Less relevant for small and low-power chips, yes. But to say it will be a desktop-only thing is not right.
Latest AMD mobile CPUs have USB4 support so eGPUs will work! Also AMD mobile CPU natively have ECC support! So if we could only get motherboard support, then we could install our own ECC!
The 7040 series with Zen 4 and RDNA 3 in an APU look like they could be really great for laptops where you want to do some graphics or gaming, but donāt want to need to be plugged directly into a nuclear reactor and have the fan sound like a helicopter taking off. It also has an AI engine, whatever that is. Itās by far the most modern and interesting CPU in AMDās new laptop lineup, but I imagine it may carry a premium, despite not being available in 16 core variation (fine by meā¦)
Yes but AMDās new Ryzen 7000 mobile processors include the 16-core Zen 4 7945HX, cāmon framework you know its what we all want.
The 7x45 CPUs are probably not going to happen since they have a 55W TDP, but Iāll add here that Iāll jump on any Framework laptop with a Ryzen 7x40 CPU as soon as it releases!
