Core 9000 series

[humbug]

AMD defence force are out in numbers this week. The thing is though, even with these fixes for the bugs in place. The 9900k is still going to wipe the floor with the 2700x.

Oh no one is denying its a good thing, i rather like competition like this, the harder the competition the better, just bring it all on all out war!

This will put the pressure back on AMD, well if you ignore the price difference... for core count Intel's move here will just insure AMD go 12 or 16 core mainstream for around £350 in a few months.

Then its Intel's turn again, what are they going to do then?

One thing is for sure, we are very soon pushing mainstream core count well into double figures and there is no way Intel can go back to charging £1000+ for such CPU's, that cash-cow is dead and good riddance. Don't you agree?

 

[SiDeards73]

@SiDeards73 With Cascade Lake Intel should be on par with AMD so we'll know then for sure what the difference between the two is IPC wise. Will be very curious to see The Stilt comparing Cascade Lake, Skylake-SP and Zen+.

@humbug you and a few others derailed the thread long before I replied to someone bringing up AMD.

Is Cascade Lake getting all the fixes then? definitely will be interesting to see how much is hampers their performance, from what i understand a lot of the exploits are in the way Intels Branch Prediction works, as well as Memory Space stuff, however the Branch Prediction part like AMD is a huge part of their actual IPC performance, if its patched to make it more secure, it will presumably introduce more steps to validate that security which i imagine is where the slow down comes from.

Any idea what node Cascale Lake is going to be on? 10nm or a 14nm+++ ?

 

[jigger]

I heard it was around 9 months before they went public, i believe both AMD and Intel were made aware, if we look at the state of current CPU's, its very telling. Its also telling BK sold off a ton of shares before this went public, enough telling that he is under investigation for it i believe.

Is it possible Intel was first made aware of the problems after it was pointed out by a third party? I'm not sure it is.

 

[Jamie Archer]

It seems Intel got 6-12 months heads up before the flaws went public.

I'll be honest and i will say that I'd be going out of my depth if i started debating too deeply about these security flaws.
I'm an all round hardware enthusiast and my loyalties are not owed to either AMD or Intel, i just basically go for what is factually the faster cpu core for core even if it is a bit more costly(within reason of course).
I can see what is statistically superior in terms of speed in this field, but that doesnt mean amd doesnt offer value for money or isnt a good choice for people who mostly take advantage of multicore perfomance or are on a budget. Ive always maintained the same outlook, and if amd prove they can be faster, then I'd happily jump ship. Top speed and core performance to me is what is most important as an enthusiast.

As far as i know, there is absolutely zero circumstances (that i've seen or read about) where a cpu vulnerability has been taken advantage of and Intel has had these security flaws for many, many years. These flaws were only conveniently and relatively recently announced to the public near Amds 1st Ryzen releases (as far as im aware).

After little reading on the subject, it appears as though it is nigh on impossible to take advantage of these flaws. Im no pro in this field and not going to pretend to be as i say, but as far as i can tell, it is something to do with the cpu being vulnerable whilst accessing cache level memory access which to take true advantage of most likely requires you to have physical access to the pc or be altering code at bios type levels (correct me if i'm wrong).

 

[SiDeards73]

Oh no one is denying its a good thing, i rather like competition like this, the harder the competition the better, just bring it all on all out war!

This will put the pressure back on AMD, well if you ignore the price difference... for core count Intel's move here will just insure AMD go 12 or 16 core mainstream for around £350 in a few months.

Then its Intel's turn again, what are they going to do then?

One thing is for sure, we are very soon pushing mainstream core count well into double figures and there is no way Intel can go back to charging £1000 for such CPU's, that cash-cow is dead and good riddance. Don't you agree?

I dont think AMD need to go 12 or 16 core in all honesty, the 9900k will be Intels premier chip until the Core 10 series, which will be a few years away, AMD only need to beat that 9900k, if the shrink to 7nm and other low hanging fruit sniping does not bring them in parity with Intels best offering, then they can introduce more cores. Otherwise if they meet parity of performance with the 9900k they will just offer it cheaper, AMD cant really massively push the price up on the next gen 7nm 3700x or whatever it will be called, people wont buy it, and it will be an obvious cash grab. AMD are still atleast 3 - 5 years away from major market share / mindshare, they cant afford any issues on the way or they will not only be back to square one, a lot of people will give them up for good.

 

[gavinh87]

Oh no one is denying its a good thing, i rather like competition like this, the harder the competition the better, just bring it all on all out war!

This will put the pressure back on AMD, well if you ignore the price difference... for core count Intel's move here will just insure AMD go 12 or 16 core mainstream for around £350 in a few months.

Then its Intel's turn again, what are they going to do then?

One thing is for sure, we are very soon pushing mainstream core count well into double figures and there is no way Intel can go back to charging £1000+ for such CPU's, that cash-cow is dead and good riddance. Don't you agree?

I'm not seeing it myself. Once Intel have maxed out the cores on the ringbus then I suspect AMD will sit back too, (with having the higher core count), then wait for Intel's new arc. Who here can honestly say that we need 12 cores or more on a mainstream desktop though?
People who need these will already be on HEDT and average people will have 50% of their cpu never used because software cannot take advantage of it yet. And by the time it does, will be made redundant by faster Cpus

My thoughts anyway.

 

[SiDeards73]

Is it possible Intel was first made aware of the problems after it was pointed out by a third party? I'm not sure it is.

The register says Intel was given 90 days around January i believe to address this, and Intel then formed a Cabal with others to basically keep schtum https://www.theregister.co.uk/2018/02/23/meltdown_spectre_letters_to_congress/

However The Verge is saying it was 7 months before it was common knowledge https://www.theverge.com/2018/1/11/...tre-disclosure-embargo-google-microsoft-linux

 

[jigger]

I'll be honest and i will say that I'd be going out of my depth if i started debating too deeply about these security flaws.
I'm an all round hardware enthusiast and my loyalties are not owed to either AMD or Intel, i just basically go for what is factually the faster cpu core for core even if it is a bit more costly(within reason of course).
I can see what is statistically superior in terms of speed in this field, but that doesnt mean amd doesnt offer value for money or isnt a good choice for people who mostly take advantage of multicore perfomance or are on a budget. Ive always maintained the same outlook, and if amd prove they can be faster, then I'd happily jump ship. Top speed and core performance to me is what is most important as an enthusiast.

As far as i know, there is absolutely zero circumstances (that i've seen or read about) where a cpu vulnerability has been taken advantage of and Intel has had these security flaws for many, many years. These flaws were only conveniently and relatively recently announced to the public near Amds 1st Ryzen releases (as far as im aware).

After little reading on the subject, it appears as though it is nigh on impossible to take advantage of these flaws. Im no pro in this field and not going to pretend to be as i say, but as far as i can tell, it is something to do with the cpu being vulnerable whilst accessing cache level memory access which to take true advantage of most likely requires you to have physical access to the pc or be altering code at bios type levels (correct me if i'm wrong).

Wait for 2022.

 

[CAT-THE-FIFTH]

The advantage of moar cores,is that since many games don't really need like 12 to 16 cores yet,it means most of us should be able to get a cheaper CPU as these will end up with 8 cores,as long as clockspeeds are similar across the range.

 

[AndreiD]

@humbug Who would have thought, competition is good for consumers and when there isn't any competition a corporation, bet it X or Y, will try to increase its margins.

@SiDeards73 It's apparently getting silicon fixes for the variants for which firmware mitigation brought performance impacts, namely variant 2 and L1TF. Branch prediction isn't that different between the two, I think there are more factors to the IPC difference between the two, one of them being how Zen+ got a small IPC uplift by just improving cache latencies.
In the anandtech article they said the performance penalty should be a fraction of what the firmware mitigation brings, so there will most likely be some performance loss with hardware mitigation, the question is just how much.
Cascade Lake is apparently an improved 14nm node, they've been backporting some 10nm R&D to their HVM nodes (22nm & 14nm) according to David Kanter. 10nm next year at the earliest.

 

[NinjaCool]

At what point did Intel know about the flaws is the big question.

I heard it was around 9 months before they went public, i believe both AMD and Intel were made aware, if we look at the state of current CPU's, its very telling. Its also telling BK sold off a ton of shares before this went public, enough telling that he is under investigation for it i believe.

I believe it was in July 2017 when independent researchers found the Meltdown bug, (later Spectre too) so they probably contacted Intel right after their discovery.

 

[humbug]

I dont think AMD need to go 12 or 16 core in all honesty, the 9900k will be Intels premier chip until the Core 10 series, which will be a few years away, AMD only need to beat that 9900k, if the shrink to 7nm and other low hanging fruit sniping does not bring them in parity with Intels best offering, then they can introduce more cores. Otherwise if they meet parity of performance with the 9900k they will just offer it cheaper, AMD cant really massively push the price up on the next gen 7nm 3700x or whatever it will be called, people wont buy it, and it will be an obvious cash grab. AMD are still atleast 3 - 5 years away from major market share / mindshare, they cant afford any issues on the way or they will not only be back to square one, a lot of people will give them up for good.

AMD understand that they need to push Intel into a corner they cannot get out of, AMD know they have a technology advantage where they can produce much higher core count CPU's for much less money.

Here it is: for years Intel have been charging insane amounts of money for high core count CPU's, less than 2 years ago £2000 for 10 cores, this is a revenue stream AMD can easily and should deny Intel outright.

So when we are talking about what AMD need to do, i think AMD need be as aggressive as they possibly can, they have the power to deny Intel their cash-cow revenue stream and i think AMD shouldn't pussy-foot around here, they need to take the biggest lump-hammer they have and fully utilise it to stove Intel's head in.
And i think they will, they will force Intel to lose the core count war and at a great expense to them.

 

[jigger]

The advantage of moar cores,is that since many games don't really need like 12 to 16 cores yet,it means most of us should be able to get a cheaper CPU as these will end up with 8 cores,as long as clockspeeds are similar across the range.

Yep prices will be driven down regardless.

 

[SiDeards73]

I'll be honest and i will say that I'd be going out of my depth if i started debating too deeply about these security flaws.
I'm an all round hardware enthusiast and my loyalties are not owed to either AMD or Intel, i just basically go for what is factually the faster cpu core for core even if it is a bit more costly(within reason of course).
I can see what is statistically superior in terms of speed in this field, but that doesnt mean amd doesnt offer value for money or isnt a good choice for people who mostly take advantage of multicore perfomance or are on a budget. Ive always maintained the same outlook, and if amd prove they can be faster, then I'd happily jump ship. Top speed and core performance to me is what is most important as an enthusiast.

As far as i know, there is absolutely zero circumstances (that i've seen or read about) where a cpu vulnerability has been taken advantage of and Intel has had these security flaws for many, many years. These flaws were only conveniently and relatively recently announced to the public near Amds 1st Ryzen releases (as far as im aware).

After little reading on the subject, it appears as though it is nigh on impossible to take advantage of these flaws. Im no pro in this field and not going to pretend to be as i say, but as far as i can tell, it is something to do with the cpu being vulnerable whilst accessing cache level memory access which to take true advantage of most likely requires you to have physical access to the pc or be altering code at bios type levels (correct me if i'm wrong).

Have a watch

https://www.youtube.com/watch?v=kwnh7q356Jk
https://www.youtube.com/watch?v=bReA1dvGJ6Y
https://www.youtube.com/watch?v=RbHbFkh6eeE
https://www.youtube.com/watch?v=0kHFvUcQsWQ

Those are a few examples of whats possible, admittedly they wont be easy to execute without forms of access to the client machine, yet you can garauntee there are people already trying to figure out how to package these to malware to harvest data.

 

[gavinh87]

Have a watch

https://www.youtube.com/watch?v=kwnh7q356Jk
https://www.youtube.com/watch?v=bReA1dvGJ6Y
https://www.youtube.com/watch?v=RbHbFkh6eeE
https://www.youtube.com/watch?v=0kHFvUcQsWQ

Those are a few examples of whats possible, admittedly they wont be easy to execute without forms of access to the client machine, yet you can garauntee there are people already trying to figure out how to package these to malware to harvest data.

So 99% of users of mainstream CPUs with the Microsoft/bios updates are fine?

 

[AndreiD]

Phoronix released an article with the performance impact with all security issue mitigation installed on Linux: https://www.phoronix.com/scan.php?page=article&item=linux-419-mitigations&num=1
Of course that's for data center workloads, not for consumer workloads, but it gives some insight as to why they rushed to release Cascade Lake, which is apparently already shipping for hyperscalers/HPC: https://www.nextplatform.com/2018/08/29/cascade-lake-heart-of-2019-tacc-supercomputer/

 

[SiDeards73]

AMD understand that they need to push Intel into a corner they cannot get out of, AMD know they have a technology advantage where they can produce much higher core count CPU's for much less money.

Here it is: for years Intel have been charging insane amounts of money for high core count CPU's, less than 2 years ago £2000 for 10 cores, this is a revenue stream AMD can easily and should deny Intel outright.

So when we are talking about what AMD need to do, i think AMD need be as aggressive as they possibly can, they have the power to deny Intel their cash-cow revenue stream and i think AMD shouldn't pussy-foot around here, they need to take the biggest lump-hammer they have and fully utilise it to stove Intel's head in.
And i think they will, they will force Intel to lose the core count war and at a great expense to them.

Still not buying it, if your AMD and watching Intel right now, you know they are under fire and need to address the security issues, they are failing a 10nm transition and having to keep on their existing node with lots of refinements, there is only so far that it can be pushed.

AMD have the advantage of more cpus per wafer when they go to 7nm, sure the wafer cost is going to be higher, but not so much so that AMD wont be making decent money by only increasing the cost of the 3700X by £20 or something.

Why follow Intel into the high price tiers, especially if your matching their IPC, you land the killer blow by offering the same performance, but drastically cheaper, Intel would need to change their entire product stack pricing to compete with that.

Intels problem now is they simply offer too many CPU's and will be even worse with the 9900k...

2c/4t,4c, 4c8t, 6c, 6c/12t, 8c, 8c/16t etc and thats just in the non HEDT domain.. and then factor in there are also K versions of many of those chips.. its ridiculous..

AMD on the other hand has done the right thing, keep all your CPU's unlockable, and keep it simple in what your offering. Yes AMD's product stack is also quite big but nothing in comparison to Intel, if AMD have to change pricing its less onerous for them.. if Intel suddenly needed to drop the price of the 9900k, they'd pretty much need to adjust everything else below it, and thats a lot of cpu's.

 

[jigger]

So 99% of users of mainstream CPUs with the Microsoft/bios updates are fine?

99% users? You mean 1% of users?

 

[SiDeards73]

So 99% of users of mainstream CPUs with the Microsoft/bios updates are fine?

The key here is "Microsoft / Bios Updates" the average guy on the street buying his PC from high street retailer is not doing BIOS updates, the average person buying their laptop does not even know what the BIOS is, let alone how to update it.

I know tons of people, literally tons, who never ever ever do firmware / bios updates. Hackers also know this.

So yes it is a major problem, just because you and i and other members who and enthusiasts update our bios, do not for one naive moment think we are the majority, we are the extreme minority.

How many corporate PC's out there do you think there are that never ever get bios updates? i can tell you from personal experience its a huge amount.

 

[jigger]

I'm not seeing it myself. Once Intel have maxed out the cores on the ringbus then I suspect AMD will sit back too, (with having the higher core count), then wait for Intel's new arc. Who here can honestly say that we need 12 cores or more on a mainstream desktop though?
People who need these will already be on HEDT and average people will have 50% of their cpu never used because software cannot take advantage of it yet. And by the time it does, will be made redundant by faster Cpus

My thoughts anyway.

So the master race should stop buying hardware, but if they do buy hardware they shouldn't be concerned with performance or security, only buy the fastest Intel they can afford.

Again you shame with your level of logic and lateral thinking.