安裝 Steam
登入
|
語言
簡體中文
日本語（日文）
한국어（韓文）
ไทย（泰文）
Български（保加利亞文）
Čeština（捷克文）
Dansk（丹麥文）
Deutsch（德文）
English（英文）
Español - España（西班牙文 - 西班牙）
Español - Latinoamérica（西班牙文 - 拉丁美洲）
Ελληνικά（希臘文）
Français（法文）
Italiano（義大利文）
Bahasa Indonesia（印尼語）
Magyar（匈牙利文）
Nederlands（荷蘭文）
Norsk（挪威文）
Polski（波蘭文）
Português（葡萄牙文 - 葡萄牙）
Português - Brasil（葡萄牙文 - 巴西）
Română（羅馬尼亞文）
Русский（俄文）
Suomi（芬蘭文）
Svenska（瑞典文）
Türkçe（土耳其文）
tiếng Việt（越南文）
Українська（烏克蘭文）
回報翻譯問題
討論區規矩與守則
檢舉此篇文章
on MB
How to keep things moving along the chain to keep those fps.
SLI power feed on battery woulda been a madhouse. Batteries need a full redo?
Get some things benefit and others won't even work after forcing over 60 refresh and that would be a software ability to optimize even after a driver compromise.
256 bit was figuring see more often and 384 for this raytrace hardwares in gametops. If pins have gotten longer for desktops how come for laptops less then smaller...
All right thanx for the chance for some feedback where ever it leads
Further explanation of those two things...
1. Yes, there are improvements elsewhere that compensate to a degree to allow lowering the bandwidth. Namely, that improvement is more cache. AMD did it with Infinity Cache on GPUs, then nVidia did it on GPUs, and then AMD did it again on CPUs with v-cache (X3D CPUs). Rumors have been that Intel will do it with CPUs but that's been rumors for a couple years so who knows. Cache, as it turns out, is pretty important for performance. Unused RAM gets used for it, RAM gets used for it with storage (yes, even on fast NVMe SSDs), and some older PCs (like Pentium IIIs) allowed you to completely disable to CPU L1/L2 (or maybe it was only L2) cache and you could see how slow things would be without cache.
2. While they are able to compensate for the loss of bus width with cache, part of it is just cutting it down to encourage consumers to buy higher end parts, or to replace hardware more often. nVidia doesn't want consumers getting comfy with the same thing for more than two generations, it seems. On the desktop side, the x60 is now 128-bit (and 8 lanes instead of 16), the x70 is now 192-bit, and the x80 is 256-bit. The first two tiers at least are cut down compared to the past, and if you look at the chip specs/performance relative to the flagship, they are also cut down a full tier. This is why people call the RTX 4060 an x50, and the RTX 4070 an x60, etc. The RTX 4070 Ti occupies the same relative spot in the lineup that the GTX 1060 did. If you think this is bad, rumors are indicating the RTX 5080 will be half of the RTX 5090 (the latter will likely be two chips?), and the VRAM/bus width counts of the lower tiers are not going to increase next generation. Expect x60s with 8 GB, x70s with 12 GB, etc. again, at least until the 3 Gbit RAM releases and allows for 24 GB/18 GB/12 GB at the x80/x70/x60 tiers respectively, which will likely come later as either Ti or Super refreshes.
The CPU (p-cores and e-cores) has nothing to do with it. The CPU is its own thing.
The laptop 4080 is comparable to a desktop 4070, which is cut down in bus width, but it's 192-bit not 128-bit for the 4070. Keep in mind that chips (CPUs and GPUs both) are running towards a wall where they can't keep scaling performance like they used to because node shrinks aren't offering the same gains, so part of the answer has been "throw more power, and thus heat, at the problem". This is more costly, and less practical for laptops especially with their limited cooling. Might be why laptop ones are cut down more? I can only speculate there.
How does this effect virtual memory. If a cap of ram can be used by the GPU separate amount for CPU, has that increased? Important for crashes from limited resources also explains why not everyone has it happening.
32GIG say is the cap, having over 64 gig of ram so both cpu and gpu can have a spillover being averted with MHZ increase of ram and gpu and cpu? since caches are going to be able to use less lanes more efficiently.
I see control panel needing more system information and/or directx going further then 64bit.
where is the upgrade? specialized ram a channel for gpu
Not going gen5 and building say a SLI that is no longer needing updates
full change out?
gen5 ssd with ddr5 ram and double digit cpu core with a gpu for physics needed only?
gpu x16 phase out from lanes being taken. Power savings for cars at stop has become a standard gpus getting a feature?
Power saving involved in gametops over desktop with battery removed options should be implemented or console will become more relevant.
vram bit bus width has nothing to do with system ram
its how many bits it can address and transfer at one time
more cache can make it call the same cached data less, but when games are using 8+g of vram, odds are its not going to be calling the same cached bits every time
to bring up the memory bandwidth, you can either increase frequency, which is already pushing ddr5 limits, or increase the amount of bits it can access in each byte
64-128-192-256-384-512 bit buss width are common for vram
as for the pci-e revs, and fewer lanes
thats another story, its cheaper to make a card with 8 lanes vs 16, but stuffing it into a 3rd gen pci-e rev really hurts it, vs 4th gen
the nvidia x60 / and amd x600 cards are kind of a joke for upgrading older pcs with pci-e 3.0 but ok with pci-e 4.0
the x60 is kinda the entry level for gaming now since games are more demanding, and standard res is raising toward 1440p
4080s mobile have 192bit memory bus and it is equal to RTX4070TI/Super
And the CPU size of i9-14900HX is 45x37.5mm. Which is nothing.
AMD Ryzen 9 7945HX3D - 45x45mm size.
With touchpads getting larger and after removal of disk drives and even flash drives USB port c being tasked with charging. Removal of speaker system with jack in for larger systems time for?
If ssds are being adapter pci slotted raided GPUs having it done to and losing some overall for redundancy. The monitors refresh fps are just the same scene over and over.
course this is just theorycraft to find whether time to go back or head into newer architecture.
Trying to figure in graphics cards as the sound card since the hdmi port has that function and those missing 64 bit is specifically 256-192=?? and so it is 256?
each gpu core has its own ram and set bus width
the laptop gpu is not the same as a desktop gpu
some high end laptops have mobile pic-e x16 slot for their gpu
please, just stop with all the buzzwords you dont understand
And the GPU chip is soldered to the motherboard like they do on an external PCB.
If 192 bits seems strange to you, check out the Xeon E5 1650 whose memory controller has triple channel mode or 3 dimms x64bits per dimm making it 192 bits. Same story with GPUs.
SLI is dead! You can't just take two GPU chips and put them together (Well, you could if it was a chiplet, but it isn't), you also have one chip with one memory controller.
What SLI does is basically take two chips, make them work together, but the memory management IS TERRIBLE due to having TWO memory controllers that rely on their own memory and ALL THE DATA NEEDED has to be DUPLICATED.
No idea what do you even mean.
And this is BS. GPUs are not audio cards, they are outputs.
you can use hdmi as audio only to a receiver for 2/2.1/5.1/7.1/atmos
mobo or onboard can do audio to the hdmi if you use the igpu
but again, none of that has anything to do with the gpus memory bandwidth or bits per byte the gpu uses for its image processing
256
192
128
The point of Post
128 bandwidth for 4080s across dozens of gametops,
questioned why?
is it the ddr5 ram? is virtual memory capped and now higher with 4080?
is it clock rates?
is it the space needed for Double digit core CPUs?
Is audio of hdmi accounted for with the GPU bit?
Was responded with it's
Caches
Got to thinking
Is audio of hdmi accounted for with the GPU bit?
explained was not so, was explained is so
Wondering now what is left to phase out to fit in a higher grade of gpu for a gametop
reminded things aren't as they appear
recapped.
Thank you for refreshing.
The mobile version uses the AD103 processor, while the desktop version uses the AD102.
AD103 is used for RTX4080/Super-4070/Super, and I don't know how to say that the problem is not why the manufacturers didn't put more RAM banks, but because the GPU chip has a memory controller that depends on it.
DDR5 has nothing to do with this. Virtual memory, they use it, but they don't rely on it, because relying on it means a drastic loss of performance. Just look at the memory bandwidth of GPUs and 16X PCIe 4 or even 5, a hundred times slower.
The CPUs, as I mentioned, don't take up much space, even the X3D CPU which come with a lot of cache I gave as example, as you can see, is not a big deal.
HDMI audio has nothing to do with it.
Clock speeds, partially.
Here's the problem.
Laptops have always had and always will have this problem. Heat.
So why should they put in a real RTX 4090 if they can't use it properly because it pulls out 300W of heat? They can downclock it and undervolt it, which makes it possible to use it, but here's the other problem. Is it worth spending tons of money to get a system that will be about 3% faster. And why should a GPU have bigger memory bus width when it can't take much advantage of it?
laptops are not only limited by heat, but also by wattage
the battery and power brick can only provide a limited amount of power to the entire system, including its display
lcd panels are around 20-40w by themselves
as for the mobile gpus in the mobile pci-e x16 slot, they are all the exact same size and gpu core positions, so they can be swapped and use the same cooling as the orig card
as stated above, the mobile gpu is a cut down version of the desktop gpu
virtual memory is using drive space as memory
not really sure what you think it is
hdd/ssd is so much slower than system ram, and thats so much slower than dedicated gpu memory
ddr5 is only 64bit, 2 channels would be 128bit width (2x 64bit calls per ram freq tick)
gddr5 can come in many flavors
the sound card on the gpu has nothing to do with any other components on the gpu
its a separate thing, with no analog conversion, and does not need hardly any pci-e lane bandwidth
onboard sound cards, share 1 lane with many other things, usb, lan, and other controllers
4060
These are 4080 with 128 bit. A bug occurring in the UI. That's a biggy.
When windows had not been able to calculate for ram sizes and everything was askew.
DXdiag, system advanced display settings, control panels. if don't dismantle and check manual.
on a gametop
heat and wattage will not be allowed to improved at the mid range why it is mid range
configuration of cpu, usb, lan, sound doesn't effect gpu's lane usage
monitor could effect gpu for wattage usage, going that route for improvement
ram is a buffer
with cpu cache going up l1 l2 l3, gpu cache is there
there are switches for direct usage by gpu over the cpu to monitor for gametops
there is software caches for directx and gpus