Guru3D Thermal Paste Roundup 2019
Back-to-school Microsoft Office 2016 $18 Win 10 Pro $12
TeamGroup T-Force Vulcan 1TB SSD Review
Corsair K57 RGB Wireless keyboard review
AMD Ryzen 7 3800X review
PowerColor 5700 RED DEVIL review
MSI Radeon RX 5700 XT Evoke review
ASUS Radeon RX 5700 XT ROG STRIX review
ViewSonic Elite XG240R Monitor Review
Promo: URCDKey Summer Sale Windows 10 Pro for $11
AMD Security Announcement on Fallout, RIDL and ZombieLoad Attack
AMD today responded to yesterday's news about the new serious Intel processor vulnerabilities. In this announcement, they confirmed that their processors are not susceptible to this kind of vulnerability.
Yesterday, researchers announced three new security exploits – Fallout, Rogue In-Flight Data Load (RIDL) and “ZombieLoad Attack”. Based on our internal assessment, we believe AMD products are not impacted by these new threats.
Below is our public statement, which you’ll also find available here.
At AMD we develop our products and services with security in mind. Based on our analysis and discussions with the researchers, we believe our products are not susceptible to ‘Fallout’, ‘RIDL’ or ‘ZombieLoad Attack’ because of the hardware protection checks in our architecture. We have not been able to demonstrate these exploits on AMD products and are unaware of others having done so.
You can read up all about it in this whitepaper, titled “Speculation Behavior in AMD Micro-Architectures.”
AMD Security Vulnerability – The Day After - Seems Financially Motivated - 03/14/2018 06:20 PM
It has been a day after the news broke on the acclaimed AMD Security Vulnerabilities. In this news-item, I wanted to recap and report on the current status and overview, as well as my look on things....
AMD Security Statement from CTO and SVP Mark Papermaster - 01/12/2018 03:30 PM
AMD have updated their web page and clarified a bit more on speculative_execution vulnerabilities with some additional details on current state of affairs and AMD actions....
Aura89
Senior Member
Posts: 7563
Joined: 2008-07-31
Senior Member
Posts: 7563
Joined: 2008-07-31
#5669698 Posted on: 05/16/2019 02:45 AM
If you were a hacker an wanted to hack as many computers as you could would you first find a flaw in amd based computers or intel?
intel = 80%
amd = 20%
ARM
Many more devices then Intel and AMD, and so much more potentially juicy, valuable data with what people do with their phones and tablets and various other gadgets.
If you were a hacker an wanted to hack as many computers as you could would you first find a flaw in amd based computers or intel?
intel = 80%
amd = 20%
ARM
Many more devices then Intel and AMD, and so much more potentially juicy, valuable data with what people do with their phones and tablets and various other gadgets.
Alessio1989
Senior Member
Posts: 1311
Joined: 2015-06-11
Senior Member
Posts: 1311
Joined: 2015-06-11
#5669699 Posted on: 05/16/2019 02:48 AM
ARM
Many more devices then Intel and AMD, and so much more potentially juicy, valuable data with what people do with their phones and tablets and various other gadgets.
ARM looks like a clusterfuck but I can bet my balls they share 99.9% of security holes (speaking of the same generation). OEM and builders usually personalize only minor things on the socket, but they do not touch the CPU implementation at all. Moreover on ARM devices usually things are surrounded by a crap OS, usually in an outdated version and usually with tons of OEM useless personalization which adds other security holes.. And on those devices most of the user sensitive data are stored.
ARM
Many more devices then Intel and AMD, and so much more potentially juicy, valuable data with what people do with their phones and tablets and various other gadgets.
ARM looks like a clusterfuck but I can bet my balls they share 99.9% of security holes (speaking of the same generation). OEM and builders usually personalize only minor things on the socket, but they do not touch the CPU implementation at all. Moreover on ARM devices usually things are surrounded by a crap OS, usually in an outdated version and usually with tons of OEM useless personalization which adds other security holes.. And on those devices most of the user sensitive data are stored.
jwb1
Senior Member
Posts: 373
Joined: 2011-03-28
Senior Member
Posts: 373
Joined: 2011-03-28
#5669702 Posted on: 05/16/2019 02:53 AM
Fanboy to the max. Intel deserves any bashing they get from this.
Time to ditch my intel server. Threadripper here I come.
Actually, he pointed out facts. You just don't like them. Difference.
Get it while you can, rumor is Threadripper is getting killed soon by AMD: https://www.forbes.com/sites/antonyleather/2019/05/15/amd-killing-off-threadripper-processors-suddenly-make-perfect-sense/
Fanboy to the max. Intel deserves any bashing they get from this.
Time to ditch my intel server. Threadripper here I come.
Actually, he pointed out facts. You just don't like them. Difference.
Get it while you can, rumor is Threadripper is getting killed soon by AMD: https://www.forbes.com/sites/antonyleather/2019/05/15/amd-killing-off-threadripper-processors-suddenly-make-perfect-sense/
Alessio1989
Senior Member
Posts: 1311
Joined: 2015-06-11
Senior Member
Posts: 1311
Joined: 2015-06-11
#5669704 Posted on: 05/16/2019 03:00 AM
Get it while you can, rumor is Threadripper is getting killed soon by AMD: https://www.forbes.com/sites/antonyleather/2019/05/15/amd-killing-off-threadripper-processors-suddenly-make-perfect-sense/
Probably, but I bet there could be still place for a quad-channel Zen2 CPU platform, especially on the evergreen workstation market. Certainly 4000 and up DDR4 help a bit when on you are tied on dual-channel only...
Get it while you can, rumor is Threadripper is getting killed soon by AMD: https://www.forbes.com/sites/antonyleather/2019/05/15/amd-killing-off-threadripper-processors-suddenly-make-perfect-sense/
Probably, but I bet there could be still place for a quad-channel Zen2 CPU platform, especially on the evergreen workstation market. Certainly 4000 and up DDR4 help a bit when on you are tied on dual-channel only...
Aura89
Senior Member
Posts: 7563
Joined: 2008-07-31
Senior Member
Posts: 7563
Joined: 2008-07-31
#5669705 Posted on: 05/16/2019 03:02 AM
Actually, he pointed out facts. You just don't like them. Difference.
There was no facts in what er557 said, just more nonsense without facts.
"minimal performance impact" can't be a fact, as, again, you have to prove something to be a fact. He can hope, but that's no fact.
"With AMD you get lower per core performance" direct opposite of facts. Now, is this true some of the time? Sure, but it's not universal, and is dependent on the processors on each side as well as what you're doing with, it, such as gaming. So yup, still no facts here.
"low efficiency interconnect," Define "efficiency", as i see no facts here again. Yes, AMDs infinity fabric has its issues, but it also has very strong points as well. But none of it really could be labeled "efficiency", unless you're talking about wattage used, and Intel chips at best match AMD for efficiency, so, nope no facts here either.
"lower performance in games" Again, not universally, and depending on your GPU, your game settings and resolution. Now if he had said "lower performance in games i play at settings i play", then sure! there's some facts, but he didn't. No facts here.
"lower performance in productivity software" Ha? I mean, was this one just a joke? Do i even have to explain why this is not facts? No, really, i shouldn't have to, so i won't, it's that obvious.
The problem with many of you intel fanboys is the fact that you decide that if Intel wins at ONE situation, it therefore wins at all situations.
Then you go around saying how great intel is and how AMD can't compete, these are universal statements, not directed statements, and can only be received as such.
You want to say that Intels i9-9900k is is better at most games with a RTX 2080 ti, low settings and 1080p then a 2700x with same settings? Do it! you'll be correct! Finally, for once, you'll be correct! But no, you just say "Intel is best at gaming" with no regard to what game, what resolution, what settings, what processors and what graphics card is being used
Then you proceed to dig yourselves even more of a hole by stating things, like above "lower performance in productivity software", which literally has no bearing. What are you comparing, an Core i9-9980XE to an Athlon 200ge?
Actually, he pointed out facts. You just don't like them. Difference.
There was no facts in what er557 said, just more nonsense without facts.
"minimal performance impact" can't be a fact, as, again, you have to prove something to be a fact. He can hope, but that's no fact.
"With AMD you get lower per core performance" direct opposite of facts. Now, is this true some of the time? Sure, but it's not universal, and is dependent on the processors on each side as well as what you're doing with, it, such as gaming. So yup, still no facts here.
"low efficiency interconnect," Define "efficiency", as i see no facts here again. Yes, AMDs infinity fabric has its issues, but it also has very strong points as well. But none of it really could be labeled "efficiency", unless you're talking about wattage used, and Intel chips at best match AMD for efficiency, so, nope no facts here either.
"lower performance in games" Again, not universally, and depending on your GPU, your game settings and resolution. Now if he had said "lower performance in games i play at settings i play", then sure! there's some facts, but he didn't. No facts here.
"lower performance in productivity software" Ha? I mean, was this one just a joke? Do i even have to explain why this is not facts? No, really, i shouldn't have to, so i won't, it's that obvious.
The problem with many of you intel fanboys is the fact that you decide that if Intel wins at ONE situation, it therefore wins at all situations.
Then you go around saying how great intel is and how AMD can't compete, these are universal statements, not directed statements, and can only be received as such.
You want to say that Intels i9-9900k is is better at most games with a RTX 2080 ti, low settings and 1080p then a 2700x with same settings? Do it! you'll be correct! Finally, for once, you'll be correct! But no, you just say "Intel is best at gaming" with no regard to what game, what resolution, what settings, what processors and what graphics card is being used
Then you proceed to dig yourselves even more of a hole by stating things, like above "lower performance in productivity software", which literally has no bearing. What are you comparing, an Core i9-9980XE to an Athlon 200ge?
chispy
Senior Member
Posts: 8758
Joined: 2006-10-29
Senior Member
Posts: 8758
Joined: 2006-10-29
#5669735 Posted on: 05/16/2019 05:50 AM
Glad to hear i'm safe on my 24/7 AMD Ryzen 2700x pc .
Glad to hear i'm safe on my 24/7 AMD Ryzen 2700x pc .
Fox2232
Senior Member
Posts: 9536
Joined: 2012-07-20
Senior Member
Posts: 9536
Joined: 2012-07-20
#5669777 Posted on: 05/16/2019 09:27 AM
If you were a hacker an wanted to hack as many computers as you could would you first find a flaw in amd based computers or intel?
intel = 80%
amd = 20%
Technically speaking. In most of cases where we are aware of vulnerabilities, they were found by Security Research labs. Their modus operandi is not to cause harm.
Therefore they test and think about possible vulnerabilities in chips regardless of number of devices where they could do harm with found vulnerability.
As result, market share does not really matter.
If you were a hacker an wanted to hack as many computers as you could would you first find a flaw in amd based computers or intel?
intel = 80%
amd = 20%
Technically speaking. In most of cases where we are aware of vulnerabilities, they were found by Security Research labs. Their modus operandi is not to cause harm.
Therefore they test and think about possible vulnerabilities in chips regardless of number of devices where they could do harm with found vulnerability.
As result, market share does not really matter.
Astyanax
Senior Member
Posts: 2627
Joined: 2018-03-21
Senior Member
Posts: 2627
Joined: 2018-03-21
#5669792 Posted on: 05/16/2019 09:58 AM
Intel skipped validation procedures and lost a lot of talent in their validation lab thanks to certain senior intel employee's whining that they were falling behind ARM.
Seems to me like Intel may have been "ignoring" security in the name of speed.
Intel skipped validation procedures and lost a lot of talent in their validation lab thanks to certain senior intel employee's whining that they were falling behind ARM.
xrodney
Senior Member
Posts: 325
Joined: 2015-06-18
Senior Member
Posts: 325
Joined: 2015-06-18
#5669902 Posted on: 05/16/2019 02:29 PM
You realize that's not how that works, right?
Just take GTX 1080 ti vs RTX 2080 ti, a $300 price difference (not 700 like you imply, it'd be 400-500 if you want to take non-MSRP prices which have nothing to do with nvidia, as that price hike goes to fund the 3rd party manufacturer such as MSI, or the retailer such as newegg, or both, so i'm going to stick with $300 difference)
GTX 1080 ti 471mm2
RTX 2080 ti 754mm2
60% increase in die space.
But that doesn't mean that it cost 60% more to make, or that you get 60% less dies per wafer. No, not even remotely.
Being that it's 60% bigger, that means the likelihood of defects rises fairly dramatically, this affects costs. It's likely due to how many less dies they get per wafer that the die itself costs 2-3 times more, could even be more. That being said, we can't know the exact number, since it's publicly not even known.
Then you have to factor in the fact that GDDR6 is more expensive then GDDR5x
But i want to make an extra point on the die cost. One of the biggest reason AMD has been so competitive with its high core count processors is because they don't have a monolithic die, like intel does. This helps TREMENDOUSLY for AMD in regards to costs and being able to undercut intel yet still make good money. Their dies are so small that the amount of defective ones compared to cost isn't a problem, whereas intels much larger dies have a much larger cost due to defects.
Yes, i understand, that's AMD and Intel, and not GPUs, but the principle is the same, large dies have bad yields = higher costs not in regards to the increase in cost per physical die material.
And lastly, none of this takes into consideration R&D for the features that are in the RTX lineup that are not in the GTX lineup. You can say "oh that doesn't matter" all you want, but it does. A company doesn't just put millions and possibly billions into R&D and not factor that into the costs of a product. A CUSTOMER may not care about this cost, but if the company wants to survive, they HAVE to worry about this cost even if the customer refuses to. It's not JUST the physical material costs that need to be taken into consideration but also the expected amount of units to be sold at which margins to pay for the R&D. Both of these numbers as consumers we will never fully know.
At RTX launch 2080ti was hard to get and real price was $1200-$1400 at same time you could get GTX 1080ti between $540-$700 so if you take higher prices there is your $700 price difference, noone cares about MSRP when price is higher due to low availability.
As for yield and cost for chips 754mm2 should have at least 40% Yield which makes 26 perfect and 38 damaged dies, Thats if Nvidia using full dies for final product which is not case here. Instead they disable part of chip. This help with yields as 80% of chip is non critical for defects and even in remaining part there are certain redundancies to help salvage chips to increase yield in case of defect.
If we assume that 80% of those defective dies are usable with part of chip disabled you get to additional 30 usable dies which makes overall yield to 87.5%.
Cost per wafer should be between $6000-$8000 which makes your cost per chip in $(120-150) range (cost for 1080ti between $60-$80).
Difference in cost between GDDR5x and GDDR6 should be in 30$ range so we are looking at $90-100 higher costs per GPU.
Yes there are R&D costs, but there are here for every single chip or generation so RTX is no different from past generations.
As for comparison to Intel/AMD CPUs its not exactly same, CPUs are hit much more critically by defects, there are less redundancies for defect and as well higher impact on silicon quality that affects clocks.
You realize that's not how that works, right?
Just take GTX 1080 ti vs RTX 2080 ti, a $300 price difference (not 700 like you imply, it'd be 400-500 if you want to take non-MSRP prices which have nothing to do with nvidia, as that price hike goes to fund the 3rd party manufacturer such as MSI, or the retailer such as newegg, or both, so i'm going to stick with $300 difference)
GTX 1080 ti 471mm2
RTX 2080 ti 754mm2
60% increase in die space.
But that doesn't mean that it cost 60% more to make, or that you get 60% less dies per wafer. No, not even remotely.
Being that it's 60% bigger, that means the likelihood of defects rises fairly dramatically, this affects costs. It's likely due to how many less dies they get per wafer that the die itself costs 2-3 times more, could even be more. That being said, we can't know the exact number, since it's publicly not even known.
Then you have to factor in the fact that GDDR6 is more expensive then GDDR5x
But i want to make an extra point on the die cost. One of the biggest reason AMD has been so competitive with its high core count processors is because they don't have a monolithic die, like intel does. This helps TREMENDOUSLY for AMD in regards to costs and being able to undercut intel yet still make good money. Their dies are so small that the amount of defective ones compared to cost isn't a problem, whereas intels much larger dies have a much larger cost due to defects.
Yes, i understand, that's AMD and Intel, and not GPUs, but the principle is the same, large dies have bad yields = higher costs not in regards to the increase in cost per physical die material.
And lastly, none of this takes into consideration R&D for the features that are in the RTX lineup that are not in the GTX lineup. You can say "oh that doesn't matter" all you want, but it does. A company doesn't just put millions and possibly billions into R&D and not factor that into the costs of a product. A CUSTOMER may not care about this cost, but if the company wants to survive, they HAVE to worry about this cost even if the customer refuses to. It's not JUST the physical material costs that need to be taken into consideration but also the expected amount of units to be sold at which margins to pay for the R&D. Both of these numbers as consumers we will never fully know.
At RTX launch 2080ti was hard to get and real price was $1200-$1400 at same time you could get GTX 1080ti between $540-$700 so if you take higher prices there is your $700 price difference, noone cares about MSRP when price is higher due to low availability.
As for yield and cost for chips 754mm2 should have at least 40% Yield which makes 26 perfect and 38 damaged dies, Thats if Nvidia using full dies for final product which is not case here. Instead they disable part of chip. This help with yields as 80% of chip is non critical for defects and even in remaining part there are certain redundancies to help salvage chips to increase yield in case of defect.
If we assume that 80% of those defective dies are usable with part of chip disabled you get to additional 30 usable dies which makes overall yield to 87.5%.
Cost per wafer should be between $6000-$8000 which makes your cost per chip in $(120-150) range (cost for 1080ti between $60-$80).
Difference in cost between GDDR5x and GDDR6 should be in 30$ range so we are looking at $90-100 higher costs per GPU.
Yes there are R&D costs, but there are here for every single chip or generation so RTX is no different from past generations.
As for comparison to Intel/AMD CPUs its not exactly same, CPUs are hit much more critically by defects, there are less redundancies for defect and as well higher impact on silicon quality that affects clocks.
HWgeek
Senior Member
Posts: 439
Joined: 2005-04-04
Senior Member
Posts: 439
Joined: 2005-04-04
#5669906 Posted on: 05/16/2019 02:34 PM
Actually, he pointed out facts. You just don't like them. Difference.
Get it while you can, rumor is Threadripper is getting killed soon by AMD: https://www.forbes.com/sites/antonyleather/2019/05/15/amd-killing-off-threadripper-processors-suddenly-make-perfect-sense/
Don't think so, If AMD gonna have SMT4 then it makes perfect sense to limit TR parts to only SMT2.
This way EPYC parts will stay as the best performance parts.
Moreover, you are paying double the price per core on 2990WX vs Ryzen 2700($1700 vs 4*$220), so why would AMD loose all this extra money?
Actually, he pointed out facts. You just don't like them. Difference.
Get it while you can, rumor is Threadripper is getting killed soon by AMD: https://www.forbes.com/sites/antonyleather/2019/05/15/amd-killing-off-threadripper-processors-suddenly-make-perfect-sense/
Don't think so, If AMD gonna have SMT4 then it makes perfect sense to limit TR parts to only SMT2.
This way EPYC parts will stay as the best performance parts.
Moreover, you are paying double the price per core on 2990WX vs Ryzen 2700($1700 vs 4*$220), so why would AMD loose all this extra money?
Astyanax
Senior Member
Posts: 2627
Joined: 2018-03-21
Senior Member
Posts: 2627
Joined: 2018-03-21
#5669908 Posted on: 05/16/2019 02:38 PM
Ryzen getting SMT4 isn't a thing.
Ryzen is per core ipc on par with intel, they are behind where clocks count.
Ryzen getting SMT4 isn't a thing.
Ryzen is per core ipc on par with intel, they are behind where clocks count.
vbetts
Moderator
Posts: 14533
Joined: 2006-07-04
Moderator
Posts: 14533
Joined: 2006-07-04
#5669909 Posted on: 05/16/2019 02:40 PM
SMT4 isn't a thing.
Ryzen is per core ipc on par with intel, they are behind where clocks count.
Not sure what I would call it, but if we're referring to 4 threads per core Intel and IBM both have iterations of this.
SMT4 isn't a thing.
Ryzen is per core ipc on par with intel, they are behind where clocks count.
Not sure what I would call it, but if we're referring to 4 threads per core Intel and IBM both have iterations of this.
Astyanax
Senior Member
Posts: 2627
Joined: 2018-03-21
Senior Member
Posts: 2627
Joined: 2018-03-21
#5669912 Posted on: 05/16/2019 02:48 PM
On cpu's that are designed for throughput at the expense of latency, not for high performance servers and consumer chips.
I think its called Coarse Grain Multithreading btw.
Not sure what I would call it, but if we're referring to 4 threads per core Intel and IBM both have iterations of this.
On cpu's that are designed for throughput at the expense of latency, not for high performance servers and consumer chips.
I think its called Coarse Grain Multithreading btw.
vbetts
Moderator
Posts: 14533
Joined: 2006-07-04
Moderator
Posts: 14533
Joined: 2006-07-04
#5669913 Posted on: 05/16/2019 02:51 PM
Xeon Phi yes, but IBM only makes PPC based CPUs for high output server performance. In any case, the technology does exist.
On cpu's that are designed for throughput at the expense of latency, not high performance servers and consumer chips.
Xeon Phi yes, but IBM only makes PPC based CPUs for high output server performance. In any case, the technology does exist.
Click here to post a comment for this news story on the message forum.
Senior Member
Posts: 4264
Joined: 2012-11-10
I myself have a laptop with an i3-4100U. For many years, it's been able to keep up just fine with my daily workloads, but ever since 2019 when pretty much all of the mitigations have been applied, I've been noticing performance dips, and my CPU usage when watching Youtube videos has gone up dramatically. It's still usable but I think I'm ready for an upgrade now.
The only reason I never mentioned any of this earlier is because I don't like to use personal anecdotes as evidence to back up a point. Haha so, the only reason I'm saying any of this now is to say I feel your pain.