HW News - NVIDIA Adds 'Smart Access Memory,' Zen 3 Delid & Die Shots, Intel Add-in GPU
video description
Date: 2020-11-15
Related videos
Comments and reviews: 9
TheGerudan
About the supposed paper launch of the new Ryzen: One of Germany's biggest online retailer claimed after the launch, that it was by far the biggest CPU launch they ever saw (in about 24 years of business) and that they sold thousands of units in only two hours.
They also write to every product on their online catalogue how often they sold it and when the 5600X was still in stock at the evening of the 5th that number for that one CPU already was in the thousands, if I remeber correctly. So if one German retailer alone really had thousands of CPUs at launch, it really doesn't look like a paper launch. They probably sold more Ryzen CPUs at launch day then they were able to sell 3000 series cards in two months.
reply
About the supposed paper launch of the new Ryzen: One of Germany's biggest online retailer claimed after the launch, that it was by far the biggest CPU launch they ever saw (in about 24 years of business) and that they sold thousands of units in only two hours.
They also write to every product on their online catalogue how often they sold it and when the 5600X was still in stock at the evening of the 5th that number for that one CPU already was in the thousands, if I remeber correctly. So if one German retailer alone really had thousands of CPUs at launch, it really doesn't look like a paper launch. They probably sold more Ryzen CPUs at launch day then they were able to sell 3000 series cards in two months.
reply
Charlie
Can someone explain to me how resizeable BARs, which seem to give the CPU efficient access to the whole of VRAM rather than just 256Mb, give such significant kind of performance uplift? My understanding is that accessing VRAM via the BAR is more efficient than other methods when doing lots of small operations on VRAM, as you reduce the IO bottleneck. So:
1) What are the use cases for the CPU accessing VRAM, as anything the GPU can do directly will be much more efficient
2) Why is 256Mb not enough, assuming that this is only beneficial when doing lots of small operations, that implies lots of small operations that need to be done over a very large portion (>256Mb) of VRAM.
reply
Can someone explain to me how resizeable BARs, which seem to give the CPU efficient access to the whole of VRAM rather than just 256Mb, give such significant kind of performance uplift? My understanding is that accessing VRAM via the BAR is more efficient than other methods when doing lots of small operations on VRAM, as you reduce the IO bottleneck. So:
1) What are the use cases for the CPU accessing VRAM, as anything the GPU can do directly will be much more efficient
2) Why is 256Mb not enough, assuming that this is only beneficial when doing lots of small operations, that implies lots of small operations that need to be done over a very large portion (>256Mb) of VRAM.
reply
Kaosreaper
The artificial scarcity theory for RTX 3000 essentially summed up to drive demand up by limiting availability enough that people can get it, but not easily , which has merit as a way to artificially increase demand.
Step two in that theory's plan is to flood the market with RTX 3000 GPUs in late november/december. This is both to get a large influx of inflated sales, that'll taper off into regular msrp sales but at high volume still; and, to remove demand for Radeon 6000 series cards because suddenly Nvidia's are extremely available.
Some of these are based off of insider info gained from MoresLawIsDead, who is a very good leaker/leak-compiler.
reply
The artificial scarcity theory for RTX 3000 essentially summed up to drive demand up by limiting availability enough that people can get it, but not easily , which has merit as a way to artificially increase demand.
Step two in that theory's plan is to flood the market with RTX 3000 GPUs in late november/december. This is both to get a large influx of inflated sales, that'll taper off into regular msrp sales but at high volume still; and, to remove demand for Radeon 6000 series cards because suddenly Nvidia's are extremely available.
Some of these are based off of insider info gained from MoresLawIsDead, who is a very good leaker/leak-compiler.
reply
SuperFirebolt
AMD is no way in hell a paper launch, PC Case Gear in Australia here has been very upfront on stock for the CPUs. With the only CPUs that need to be pre-ordered being the 5890x and 5800x yet people ordering them will be receiving them around the first 2 weeks of order. Yet if you pre-ordered the 3080 you wouldn t even get it in the first month of ordering in most cases.
AMD has done a much better job than NVIDIA and is no way a paper launch just simply is a week or 2 behind on demand. And this is expected when a new component launches especially in this year.
reply
AMD is no way in hell a paper launch, PC Case Gear in Australia here has been very upfront on stock for the CPUs. With the only CPUs that need to be pre-ordered being the 5890x and 5800x yet people ordering them will be receiving them around the first 2 weeks of order. Yet if you pre-ordered the 3080 you wouldn t even get it in the first month of ordering in most cases.
AMD has done a much better job than NVIDIA and is no way a paper launch just simply is a week or 2 behind on demand. And this is expected when a new component launches especially in this year.
reply
eivis13
Wonder why intel is doing tec? As if air cooling couldn't handle it. Maybe it's the whole 5ghz and 2kw coller on a xeon thing that brought that up? At the time power efficiency wise, having a dozen or so fans and a larger heataink is way better. If they come up with a way that keeps similar temps (2-4 deg C) with a tec but can get rid of the water cooling aspect (from what's known now it's unlikely) and let's say only a 25% increase in power consumption (compared to air) and a huge reduction in dB (compared to server fans) then maybe it's worth it for some.
reply
Wonder why intel is doing tec? As if air cooling couldn't handle it. Maybe it's the whole 5ghz and 2kw coller on a xeon thing that brought that up? At the time power efficiency wise, having a dozen or so fans and a larger heataink is way better. If they come up with a way that keeps similar temps (2-4 deg C) with a tec but can get rid of the water cooling aspect (from what's known now it's unlikely) and let's say only a 25% increase in power consumption (compared to air) and a huge reduction in dB (compared to server fans) then maybe it's worth it for some.
reply
Conor
Regarding supply, I think it s wrong to point at AMD and Nvidia and ha there the same .
You gave Nvidia a lot of benefit of the doubt when they launched and now it two months later and they still have supply.
The is a big difference between paper launch and selling out your first day.
You should at least be saying we will see how bad it is in a month. If it s the same as Nvidia by then I think we can look at both of them in the same light.
reply
Regarding supply, I think it s wrong to point at AMD and Nvidia and ha there the same .
You gave Nvidia a lot of benefit of the doubt when they launched and now it two months later and they still have supply.
The is a big difference between paper launch and selling out your first day.
You should at least be saying we will see how bad it is in a month. If it s the same as Nvidia by then I think we can look at both of them in the same light.
reply
Ken
There were definitely Ryzen 5000's at Microcenter well after the stores opened. I bought a 5600x at almost noon, the earliest I could get to the store. I have no idea how long the supply lasted but they must have had quite a lot of chips at the Chicago store there were people in front of and behind me in line to check out with all the stuff to build new systems so I assume they also had bought Ryzen 5000.
reply
There were definitely Ryzen 5000's at Microcenter well after the stores opened. I bought a 5600x at almost noon, the earliest I could get to the store. I have no idea how long the supply lasted but they must have had quite a lot of chips at the Chicago store there were people in front of and behind me in line to check out with all the stuff to build new systems so I assume they also had bought Ryzen 5000.
reply
Spoot
Could you safely delid one by adding some heat to melt the solder just enough to crack the lid and release the heat spreader..... generally not a lot of reasons why you'd want/need to...but, can it be done. Would liquid metal (grizzly kryonaut) perform better over solder? Could you get great thermals ( really low )with heatspreader off direct die, considering 2 interfaces as shown...direct die LN2?
reply
Could you safely delid one by adding some heat to melt the solder just enough to crack the lid and release the heat spreader..... generally not a lot of reasons why you'd want/need to...but, can it be done. Would liquid metal (grizzly kryonaut) perform better over solder? Could you get great thermals ( really low )with heatspreader off direct die, considering 2 interfaces as shown...direct die LN2?
reply
Jimmy
Gamers Nexus: Nvidia rushed out Ampere TO REVIEW SITES in order to DISRUPT RDNA2's buzz. At the point YOU were reviewing those Ampere cards Nvidia did NOT know the exact performance of RDNA2, assuming performance much, much less than later demonstrated.
Do you not get it??? --> it's all about mindshare, and of course, it royally backfired on Nvidia.
reply
Gamers Nexus: Nvidia rushed out Ampere TO REVIEW SITES in order to DISRUPT RDNA2's buzz. At the point YOU were reviewing those Ampere cards Nvidia did NOT know the exact performance of RDNA2, assuming performance much, much less than later demonstrated.
Do you not get it??? --> it's all about mindshare, and of course, it royally backfired on Nvidia.
reply
Add a review, comment
Other channel videos
34:27
