News Nvidia Product Page Reveals GeForce MX450 With PCIe 4.0 and GDDR6 Memory

Does it need it? No. But then it would mean that it would require less PCIe lanes on a PCIe 4 CPU meaning things like more lanes or SSD's or maybe less lanes needed on the CPU itself bringing the cost down (Hopefully on that that one).
 

salgado18

Distinguished
Feb 12, 2007
928
373
19,370
PCIe lanes don't cost all that much, going from x16 to x8 likely shaves less than $5 all-inclusive off the GPU's cost.

Also, laptop CPUs and GPUs already operate on cut-down PCIe lane counts to save power.
Then shave it down from 8x to 4x, and save more power and money. Even $5 on a budget gpu means a lot of money, because they are sold in large quantities.

Depending on gpu performance, I'd lower it down to 2x PCIex 4.0 lanes, if it doesn't become a bottleneck.
 

extremepenguin

Distinguished
May 21, 2009
32
2
18,535
I would say for budget laptops it could be a useful thing to have, tracings are in short supply in budget portables so having the ability to cut down on actual PCIE lanes to cut costs with no noticeable performance degradation could be useful in hitting key price points.
 
Jul 3, 2020
6
5
15
Laptop manufacturers love pairing Intel processors with Nvidia's GeForce MX-series graphics cards so you'll likely find the GeForce MX450 alongside some Tiger Lake processor in the foreseeable future.

A mobile intel cpu? Absolutely disgusting and unusable.
 

InvalidError

Titan
Moderator
Then shave it down from 8x to 4x, and save more power and money. Even $5 on a budget gpu means a lot of money, because they are sold in large quantities.

Depending on gpu performance, I'd lower it down to 2x PCIex 4.0 lanes, if it doesn't become a bottleneck.
The first thing that gets pared down on bottom-dollar GPUs is VRAM since it costs ~$10/GB. Once you get to 4GB or less, PCIe bandwidth becomes essential to let the GPU use system memory as an extra layer of cache before applications start constantly having to reload and re-send assets to the GPU. That's why the 4GB RX5500 fares a whole lot better against its 8GB counterparts on 4.0x8 than 3.0x8 by closing gaps most of the way and the 1650S holds up much better than 4GB RX5500/3.0x8 thanks to its 3.0x16 interface.

If you wanted to design the fastest bargain-basement GPU possible, you would end up with 2-3GB of VRAM for absolutely essential local heavy-lifting and 4.0x16 for the fastest access possible to system memory as scratch space since 4.0x16 is much cheaper than the 2-3GB of extra VRAM that would otherwise be required to max out the GPU cores of something approaching an RX5500.
 

nofanneeded

Respectable
Sep 29, 2019
1,541
251
2,090
Low Voltage CPU does not have much lanes available , so yes it is needed to use fewer lanes in any notebook . Add to it , high end Light notebooks are using two thunderbolts each consuming 4 lanes thats 8 lanes alone , two lanes for low voltage GPU , total 10 lanes and thats the max in low voltage CPU lanes.

Moving to Gen 4 will solve alot of light notebooks bottleneck.

More over , Nvidia move hints at a secret mobile intel CPU with Gen4 lanes.
 
PCIe lanes don't cost all that much, going from x16 to x8 likely shaves less than $5 all-inclusive off the GPU's cost.

Also, laptop CPUs and GPUs already operate on cut-down PCIe lane counts to save power.
as those are not powerful chips maybe x2 for lower end, or x4 for higher end ones will be used. Then it can matter on idling power IMHO.
I wonder if higher overhead of v4 did not nullify the gains of half lines.
Also I think pcie4.0 was not just twice the bandwidth right ?
Does anyone recall what was the other benefits that might be a factor here ?
I expect the cost is higher so why they did move lower end gpu's already ?
 
Last edited:

InvalidError

Titan
Moderator
as those are not powerful chips maybe x2 for lower end, or x4 for higher end ones will be used.
If you are going that low-end, you are at sub-IGP levels of GPU performance and a discrete GPU is a complete waste of time, money, power and space.

PCIe 4.0 is only twice 3.0's bandwidth, same with the upcoming 5.0 and 6.0. The odd one out is 2.0->3.0 where bandwidth only went from 5Gbps to 8Gbps.
 
Last edited: