Still waiting for a curved, double 4k panel. Gaming may be a bit rough along with 144Hz and HDR, but the real estate... Could be nice for non-FPS titles as well.
As it's likely data center focused with accelerators, security may be less of a concern. Disabling security mechanisms on a closed system in favor of performance isn't a horrible choice.
OpenCAPI should be AMDs open version. Infinity just allowed them to ship it faster. As mentioned above...
Now if someone would just consolidate all the streaming packages into one app with subscribed content made available. Prime is close, but still apps or "channels" for all the services that behave differently.
I'd still consider making an exception for plugs with added power delivery. The ones that could in theory handle 100W to drive some devices. Or possibly for docking stations with lots of bandwidth and IO.
It's not as fast as the pro card or comes with the support though. Pro should have PCIe4 which would be significant in those markets. The 1/4 rate FP64 is fast, but not too fast for a large deployment. As a prosumer/gaming product with HBCC they picked a nice spot. Haven't checked lately if...
That may include modifications. Armoring the door, added bumper, etc may increase the cost a bit.
Doesn't account for selling it back to the grid or using for something else like office lighting.
Keep in mind top speed likely isn't as important as acceleration. Sitting next to a highway...
With a VM and GPU passthrough they would be really close with minimal performance hit. That doesn't account for the security gained by sandboxing the game or app either. A Linux core with Windows/MacOS/Android VM for whatever app is running should remove lots of comparability issues.
Agreed, however a GPU could be interesting. Ideally they could get MCM working, but a design similar to Rome with say 3 processing chiplets and 1 front-end/cache chiplet could be interesting. See if that allows them to really push clocks on the front-end around the serial issues they have now...
No reason not to with the Rome design. Infinity Fabric is based on PCIe signaling and speeding up the interconnect obviously has benefits. Secondly they only have to spin a new IO die to integrate the new standard without bothering with the chiplets if they arrive for consumers.
With NVME and possibly 10GbE pushing into the consumer space the extra bandwidth could be useful.
Another big consideration would be USB3+ and thunderbolt connectors consuming a lot of bandwidth.
As I understand, it should in theory stack. This technique seems to be based on the angle of entry/exit to put it simply. Fire a laser into the fiber at +/-15/30/45 degrees etc. Then account for attenuation based on high angles having to travel further and/or exiting in opposite directions as...
I read that 10Grays/sec more as a fillrate that won't be achievable in most conditions. Even if the dataset fits in memory, the memory access is likely incoherent and limiting performance. As you said, any bounces will curtail performance significantly. Still a nice feature, but the rest of...
At this point Picard has to be an admiral, so the storyline will be interesting. DS9 got interesting once they got a galactic war and space opera in full swing. With modern CGI it could be interesting if they do something similar. Just a question of who the bad guy is. Someone from a new...
If worried about gaming performance, restricting a TR to a single NUMA node shouldn't be all that difficult and get similar results as a single chip. Then still have the upside of all the cores for when needed. Might take some work by Microsoft, but shouldn't be all that difficult to pull off...
Thermal throttling.
That chip will be well worth the money for anyone who can use all those cores. Wouldn't mind seeing how compile times hold up on that thing. That would be the defacto developer platform if all those cores worked for compiling and a remote build setup wasn't available.
The 4 blocks of 16 are the shader engines and likely a more ideal chiplet size for manufacturing. A Vega10 comparable chiplet would be <10mm2 if broken into NCUs. Certainly possible, but as I said it might not be all that practical to manufacture. A SE would be <200mm2 which may be more...
All the high end cards where the memory would be pulling 100W otherwise and really cut into that 300W window. IO is what burns a lot of power on memory and upping the voltage certainly won't help there and won't be all that useful for smaller form factors and mobile where the market appears to...
I'm not sure byte-rewritable is quite the way to go though. The indirection with a TLB could be a real PITA for the caches. More addresses than actual data at that point. Further, I'm of the mindset some form of HBM acts as a high bandwidth cache between the persistent memory and processor...
I think you mean Amazon's Echo Look that comments on your outfit. Just when you thought the mic was, imagine that sending clips to a contact. Rhetorically ponder if someone will think your ass looks fat in a dress and it inquires on your behalf.
I see quite a few good applications for this that aren't being mentioned. The ability to automate certain security and safety functions beyond just facial recognition could be huge. The system could detect armed robberies in convenience stores without any action by employees or using facial...
Considering the usage of the numbers, I don't think double counting is necessarily a bad way to go here. So long as it's counting unique gamers and the hardware they are using. Useless for hardware sales figures, but for a developer targeting a customer base it's relevant.
Except the concern there is cost associated with maintaining multiple soda fountains. Which takes extra space for the dispenser and inventory.
The concern is that APUs and embedded designs displace the majority of dGPU volume if they take off. All the low end designs feature...
One way to create an international fingerprint and facial recognition database for governments. All for the key idea, just perhaps not using fingerprints and faces.
If you completely discount the clockspeed in the 4/clock then I guess you could be right. Few if any titles actually need that much throughput. The Vega issue looks more a temp issue holding back clockspeeds. It's doing well, but not maintaining the clocks that it should which hurts...
So AMD found a preferable solution that is applicable to both IHVs and uses established APIs, and that means it never saw the light of day? Of course they could always go the explicit route and prevent devs from optimizing for Nvidia hardware. Then sink a lot of cash and devrel time into those...
Lots of video will do it. I'd still rather have an external SD card option than internal though. Cards are getting faster and cheaper, plus they can be moved to new phones.
Back to the video, anyone with kids might appreciate it. With offline options for Netflix and eventually other...
Might be worth keeping in mind, drivers aside, the actual games might not be optimized for DX12 on Nvidia hardware. I would think any game with Async On/Off would have similar paths for all DX12 hardware, but that may not be the case.
Any chance of adding CPU utilization to the graphs in the...
Why is that a surprise? Explicit > Portable Compute > Implicit. Makes more sense to focus on the portable option for developers as it works on all hardware. The 1080ti performance has already been achieved when everything is working. Doesn't mean it wont be added eventually either. That...