HW News - AMD Zen 3+, Silent NVIDIA Changes, 10GHz DDR5, Intel Goes to Mars
video description
And well, there is a huge difference between running a core consisting of a few hundred million transistors, and running an interface that only has a few serializers/deserializers consisting of a few hundred transistors each.
Also, PCIe v.3 has ran at 8GHz for years. PCIe v.4 is at 16GHz currently.
And HDMI had 12 GHz as its standard for a long time. (The new versions of it is considerably higher though.)
Then there is SFP+ ports that has managed to reach 28 GHz for a fair bit of time now.
And for anyone wanting to say that bitrate and frequency aren't the same thing, yes it isn't strictly the same thing.
Though, for everyone going, 10GHz DDR5 is only running at 5GHz! then no, it is actually sending data at 10Gb/s per differential pair, the data is actually clocked at 10 GHz, but the reference clock sent between the memory controller and the memory chip is running at half that frequency. (Since this increases clock stability.)
But the main reason why DDR memory busses aren't faster is mainly due to their exceptionally parallel nature.
DDR4 for an example has 64 data bits, and 43 address bits if I recall correctly.
DDR5 made the wise decision to split the data part of the bus, effectively making it two independent 32 bit buses, this makes any timing skews less impactful. (or rather, it is easier to ensure that there isn't much skew to begin with.)
Though, DDR5 also does a few other things to combat some other issues of prior iterations of the standard.
Like increasing the burst length, this also increases the latency of the bus since each request gets larger, and one needs to wait for the request to finish before one can start on the next one.
Date: 2021-05-02
Related videos
Comments and reviews: 9
afelias
Proof of Work: Everyone racing to do a task in parallel, fastest submits work and wins.
Proof of Stake: Everyone lines up for task allocations. Tasks are allocated in such a way so that there is little to no racing. Tasks are only allocated to those who pony up some ETH as insurance, so that if you're caught trying to fudge the records (verified as a task allocated to others, tasks overlapping in a way so that everyone is checking each others' work) they take your ETH.
Proof of Stake is meant to (not sure if it will achieve all its goals though) remove the race condition that forces everyone to pump as much power (both electrical and computational) as possible just to keep up with everyone else. This not only means that less overall power will be needed, but also allows smaller and weaker devices to still contribute and mine reasonably. You can still have ETH mining farms, but the cost of buying stakes for every machine/client makes it less viable (possibly discouraging some mining operations, though scaling should still be possible), and they won't need to use the best hardware just to catch up on race conditions.
I'm not really sure how they're going to keep Proof of Stake decentralized, because now you need clients doing the tasks of handing the tasks out to other clients, other clients verifying that tasks are being handed out right, and then the tasks themselves of writing to blocks, and verifying the writing of those blocks. As well as the tasks of taking everyone's ETH payments for stakes. It sounds crazy complicated and honestly I want to see the security issues, like the equivalent of the 51 percent attack in Proof of Work.
reply
Proof of Work: Everyone racing to do a task in parallel, fastest submits work and wins.
Proof of Stake: Everyone lines up for task allocations. Tasks are allocated in such a way so that there is little to no racing. Tasks are only allocated to those who pony up some ETH as insurance, so that if you're caught trying to fudge the records (verified as a task allocated to others, tasks overlapping in a way so that everyone is checking each others' work) they take your ETH.
Proof of Stake is meant to (not sure if it will achieve all its goals though) remove the race condition that forces everyone to pump as much power (both electrical and computational) as possible just to keep up with everyone else. This not only means that less overall power will be needed, but also allows smaller and weaker devices to still contribute and mine reasonably. You can still have ETH mining farms, but the cost of buying stakes for every machine/client makes it less viable (possibly discouraging some mining operations, though scaling should still be possible), and they won't need to use the best hardware just to catch up on race conditions.
I'm not really sure how they're going to keep Proof of Stake decentralized, because now you need clients doing the tasks of handing the tasks out to other clients, other clients verifying that tasks are being handed out right, and then the tasks themselves of writing to blocks, and verifying the writing of those blocks. As well as the tasks of taking everyone's ETH payments for stakes. It sounds crazy complicated and honestly I want to see the security issues, like the equivalent of the 51 percent attack in Proof of Work.
reply
Jeff
Linux died when it went political /woke anyway, the racism (critical race theory) & the claimed anti bullying bs (banning of free speech and anything not radical left-wing opinions) in the dev community killed it.
Refusing it and messing with its code is common practice among previous supporters, anti-racists and the sane people in europe and asia rejecting the racist extremists on the radical left (such as liberals) haha
At least that's the word from old contacts from the days in computing science in Sweden
I'm more of the crowd waiting for the inevitable blood-soaked streets necessary to cure ourselves from these regressive radical left authoritarians over here in europe but it's nice to hear it's rejected on every level of society hahaha
reply
Linux died when it went political /woke anyway, the racism (critical race theory) & the claimed anti bullying bs (banning of free speech and anything not radical left-wing opinions) in the dev community killed it.
Refusing it and messing with its code is common practice among previous supporters, anti-racists and the sane people in europe and asia rejecting the racist extremists on the radical left (such as liberals) haha
At least that's the word from old contacts from the days in computing science in Sweden
I'm more of the crowd waiting for the inevitable blood-soaked streets necessary to cure ourselves from these regressive radical left authoritarians over here in europe but it's nice to hear it's rejected on every level of society hahaha
reply
TechyBen
My (possibly wrong, but AFAIK summary of the system) understanding of Proof of Stake is:
You no longer mine new Etherium, so no point buying hundreds of GPUs (Proof of Work generates new hashes and new Etherium).
You use your existing stake of Etherium and some GPUs/CPUs or ASYNCs to prove you have the Etherium (by computing a new hash or similar from it that can be checked). So some compute, but much less needed than before (in proof of work).
So this generally suggests it's a system that prevents runs on GPUs as you need less compute, though possibly only happens once the existing ledger gets large enough. I don't think you could start a crypto as proof of stake, unless you first distribute some sort of pre-generated but private ledger.
reply
My (possibly wrong, but AFAIK summary of the system) understanding of Proof of Stake is:
You no longer mine new Etherium, so no point buying hundreds of GPUs (Proof of Work generates new hashes and new Etherium).
You use your existing stake of Etherium and some GPUs/CPUs or ASYNCs to prove you have the Etherium (by computing a new hash or similar from it that can be checked). So some compute, but much less needed than before (in proof of work).
So this generally suggests it's a system that prevents runs on GPUs as you need less compute, though possibly only happens once the existing ledger gets large enough. I don't think you could start a crypto as proof of stake, unless you first distribute some sort of pre-generated but private ledger.
reply
Satan
0:59 while i like the idea i think the diagonal line on the wasd set could be a problem, it would be prefurable to see a belt as well as reflective entiriors for the keys, maybe even some good sarilic options and the option of a black reflective belt frame to make the keys easyer on the eyes. I also dont understand why noone in the mechanical keyboard market has thought of adding a usb female with a volium whiel, the best aproach to my mind would be something like the delux 9050, master volium top right, secondery volium to the side, both sides, a module with wheel and special button would be awesome. I'm also worried about the sensor going off if I'm watching a movie?
the number one thing I want to see is a cat mode button
reply
0:59 while i like the idea i think the diagonal line on the wasd set could be a problem, it would be prefurable to see a belt as well as reflective entiriors for the keys, maybe even some good sarilic options and the option of a black reflective belt frame to make the keys easyer on the eyes. I also dont understand why noone in the mechanical keyboard market has thought of adding a usb female with a volium whiel, the best aproach to my mind would be something like the delux 9050, master volium top right, secondery volium to the side, both sides, a module with wheel and special button would be awesome. I'm also worried about the sensor going off if I'm watching a movie?
the number one thing I want to see is a cat mode button
reply
Dragon
gamersnexus Could you guys consider doing a deep dive into high quality displayport cables, particularly ones longer than 10 feet? Im guessing im not the only who has been pulling their hair out attempting to find high quality cables, especially with the proliferation of high refresh rate monitors. The more I dive into it, the more it appears there are some serious per-unit quality tolerance issues. It would be nice to have a specific list of cables that consistently meet their actually rated spec unit after unit, as well as maybe a list of manufacturers who we can actually trust to consistently produce high quality cables that meet or exceed claimed specs.
reply
gamersnexus Could you guys consider doing a deep dive into high quality displayport cables, particularly ones longer than 10 feet? Im guessing im not the only who has been pulling their hair out attempting to find high quality cables, especially with the proliferation of high refresh rate monitors. The more I dive into it, the more it appears there are some serious per-unit quality tolerance issues. It would be nice to have a specific list of cables that consistently meet their actually rated spec unit after unit, as well as maybe a list of manufacturers who we can actually trust to consistently produce high quality cables that meet or exceed claimed specs.
reply
Cristian
Hmmm. That research paper and authors deserve more than being the spot of public shaming. You might know (or find out) that social engineering techniques are all too effective and Linux project is not at all imune to them. That paper would unsurprisingly reveal exactly that and the Foundation and GH in particular as a maintainer for stable branch have a lot to lose on reputation without necessarily being on fault. The project works on the premise of good faith contributions. They do not intend and are not prepared to switching to an adversarial context.
reply
Hmmm. That research paper and authors deserve more than being the spot of public shaming. You might know (or find out) that social engineering techniques are all too effective and Linux project is not at all imune to them. That paper would unsurprisingly reveal exactly that and the Foundation and GH in particular as a maintainer for stable branch have a lot to lose on reputation without necessarily being on fault. The project works on the premise of good faith contributions. They do not intend and are not prepared to switching to an adversarial context.
reply
Michael
In response to Google just now adding search to Stadia...
They probably didn't have it before because they are just now introducing web search results code to the Google Assistant code on Stadia.
Before all you could use GA for on Stadia is personal search results. (Calendar entities)
Hopefully this is the beginning to the introduction of using GA, in Stadia, to search for how to complete a level / defeat a boss, like what was introduced at the launch / announcement of Stadia.
reply
In response to Google just now adding search to Stadia...
They probably didn't have it before because they are just now introducing web search results code to the Google Assistant code on Stadia.
Before all you could use GA for on Stadia is personal search results. (Calendar entities)
Hopefully this is the beginning to the introduction of using GA, in Stadia, to search for how to complete a level / defeat a boss, like what was introduced at the launch / announcement of Stadia.
reply
pino
we need a decrease of people wanting to buy new cards. so miners buying less cards and more old mining cards being sold that way there wont be supply issues and normal pricing. and ehterium being less profitable by mining is irrelevant theres 9000 other coins they just switch to the next.
Micrsoft : small subset.
User with a brain : Small subset of a billion users is tens of millions of users. GJ Microsoft.
reply
we need a decrease of people wanting to buy new cards. so miners buying less cards and more old mining cards being sold that way there wont be supply issues and normal pricing. and ehterium being less profitable by mining is irrelevant theres 9000 other coins they just switch to the next.
Micrsoft : small subset.
User with a brain : Small subset of a billion users is tens of millions of users. GJ Microsoft.
reply
BlowUpDaChr0n
Nividia just shot itself in its foot .. why would you neuter your best selling product.. miners are buying up the GPUs not gamers .. trying to sell us a tore down for a version of a 3060 TI for 600 that can't even play a game with lol then they're wondering why we won't buy that card let's see your gamers keep that profit up while we move over to amd..
reply
Nividia just shot itself in its foot .. why would you neuter your best selling product.. miners are buying up the GPUs not gamers .. trying to sell us a tore down for a version of a 3060 TI for 600 that can't even play a game with lol then they're wondering why we won't buy that card let's see your gamers keep that profit up while we move over to amd..
reply
Add a review, comment
Other channel videos
