Part numbers of CPU's without the meltdown and spectre bugs?

[DaveSimmons]

Get a Ryzen, they are less vulnerable (only Spectre v1), potential performance impacts are said to be negligible and they are easier to fix with software/OS patches needed.
On Intel's side, motherboard bios fixes are needed in addition to OS patches and the performance impact is potentially much higher.

That might or might not be the better choice. Coffee Lake will probably still offer much better single-core performance, especially at stock. OP should tell us the specific program(s) they use, what kind of budget they have in mind, and whether or not they are willing to overclock.

Reading the OP is important.

But so is understanding that the original poster might be asking the wrong question.

 

[whm1974]

Without knowing your specific applications or workloads I can't really comment usefully. The biggest impacts so far seem to be for I/O heavy workloads. With mitigation patches of unknown performance impact still outstanding.

If the vast majority of your tasks will scale beyond 8 cores and you can actually make use of that power, then you would want Intel or AMD HEDT platforms (Skylake-X / X299 or Threadripper / X399). If the tasks scale beyond 16 cores or you need official ECC support, then Xeon or Epyc (but If you insist on having a processor that isn't affected at the hardware level by Meltdown, you would have to skip on anything Intel).

If you don't need more than 8 cores, then Coffee Lake (i7-8700 or i7-8700K) on the Intel side, or Ryzen 1800X are your best bets for performance.

In any event, we just don't know the full extent of the performance impacts yet.

In any event, we should keep an eye out for upcoming patches and update our OSes. I thank I ran the Manjaro Linux updates like three or four time since this thing started.

 

[aigomorla]

Reading the OP is important.

"I am looking to build a new tower soon and want to know if CPU's from Intel or AMD with the physical bugs fixed are available and what part numbers to look for."

I'm fairly certain that NO Intel CPU will have the physical flaws fixed, except for low performance ones. Software fixes for the next few years.

So your telling him NO intel cpu will be useful for the next few years?

Really?

Wanna see how fast a 5ghz CoffeeLake system is even with the patches?
We dont even need to overclock it... wanna still see how fast a 4.5ghz turbo'd CoffeeLake system is with all the patches?

Do you have any concept of how the performance hits really is in real world application outside Data Servers and High End rendering Farms?
Because a NVMe loses at most 20%, you still are aware that its still about 2x faster then a SATA SSD instead of being 3x faster...

 

[SPBHM]

the old Atoms before BayTrail are the only CPUs I would think are safe.

 

[ZGR]

I think it is the 45nm "Bonnell" Atom architecture that is the latest unaffected CPU from Intel.

 

[SPBHM]

I think it is the 45nm "Bonnell" Atom architecture that is the latest unaffected CPU from Intel.

32nm Cedarview should be good (like the D2700)

 

[Snowleopard3000]

I could build out a server and use

https://ark.intel.com/products/family/451/Intel-Itanium-Processor

Itanium's are still x86 and x64 compatible.... I assume they are affected to? or they are a completely different architecture?

 

[DaveSimmons]

I could build out a server and use

https://ark.intel.com/products/family/451/Intel-Itanium-Processor

Itanium's are still x86 and x64 compatible.... I assume they are affected to? or they are a completely different architecture?

Why? The OS software fix will block the known exploit. Using a "safe" CPU that runs at one-half or one-tenth the speed is just silly. Running a Ryzen if you need single-core performance more than core count is silly.

Build a PC to get things done, not to hide from possible future exploits that might never appear and even then will only affect you if you download and run programs that will trigger them.

 

[LTC8K6]

The performance hit for home users is likely going to be small.

If you really need a new home system now, I would just not worry about it.

If you can wait, then wait a few months and see if the picture clears up.

 

[maddie]

So your telling him NO intel cpu will be useful for the next few years?

Really?

Wanna see how fast a 5ghz CoffeeLake system is even with the patches?
We dont even need to overclock it... wanna still see how fast a 4.5ghz turbo'd CoffeeLake system is with all the patches?

Do you have any concept of how the performance hits really is in real world application outside Data Servers and High End rendering Farms?
Because a NVMe loses at most 20%, you still are aware that its still about 2x faster then a SATA SSD instead of being 3x faster...

Of course not.

I'm saying that you will need software patches to mitigate the vulnerability. He asked about a hardware fix? Why do you keep ignoring that main point?

 

[LTC8K6]

Intel should have chips out this year that are not vulnerable, but really we just have to wait and see what actually happens.

 

[mv2devnull]

He asked about a hardware fix? Why do you keep ignoring that main point?

You are correct that the answer to the exact question is: "There is none." However, this situation is like when a man asks how to cut a branch of a tree. You see that he sits on the branch and will fall to his death, if he cuts the branch. Would you tell him how to use the chainsaw, as he did ask for, whether or not that actually gets him where he wants to be?

 

[scannall]

Intel should have chips out this year that are not vulnerable, but really we just have to wait and see what actually happens.

I find that unlikely. It's an architectural flaw, a bad decision as it were. To spin up a brand new architecture, test it, validate it and then produce the masks and get it into production takes years. Not a few months. They will be stuck with OS and microcode crutches for quite a while. 2020ish would be my guess, unless they have something new in the labs they can pull forward. But even that would take more than a few months.

 

[LTC8K6]

I find that unlikely. It's an architectural flaw, a bad decision as it were. To spin up a brand new architecture, test it, validate it and then produce the masks and get it into production takes years. Not a few months. They will be stuck with OS and microcode crutches for quite a while. 2020ish would be my guess, unless they have something new in the labs they can pull forward. But even that would take more than a few months.

Intel says the issue has already been fixed in hardware.
Possibly this is the reason 10nm was delayed so long?

Intel also is fixing the problem in future chips, starting with products that will arrive later this year, Smith said. Intel is effectively taking the software fixes being released now and building them directly into hardware, he said.

"We're putting those mitigations in our designs," Smith said. "We're not turning off the benefits of speculation."

The problems occur only when the chip is switching from one level of privilege to another, for example the change from running a computer user's software like Photoshop to the computer's operating system, which gets deeper access to the processor. The Intel fix will address those "corner cases... where you're moving from one level of protection to another," Smith said.

https://www.cnet.com/news/meltdown-spectre-intel-ceo-no-recall-chip-processor/

 

[scannall]

Intel says the issue has already been fixed in hardware.
Possibly this is the reason 10nm was delayed so long?




https://www.cnet.com/news/meltdown-spectre-intel-ceo-no-recall-chip-processor/

10nm was/is delayed because of performance and yield issues.

And I am skeptical of their statement. Seems more like damage control and spin. I'd like to be wrong on it and all, so I guess we'll all have to wait and see.

 

[LightningZ71]

Itanium is a very different CPU architecture and handles the concept of speculative execution in a very different way. As a result, it is effectively immune to these attacks.

 

[maddie]

Intel says the issue has already been fixed in hardware.
Possibly this is the reason 10nm was delayed so long?




https://www.cnet.com/news/meltdown-spectre-intel-ceo-no-recall-chip-processor/

Wow. What a post. Quick, send PM, I've got a couple bridges to sell.

I assume you've read the info from the ex-Intel engineer explaining the difficulty in redesigning hardware to overcome these flaws.

If you still are pushing this official line, then let's extrapolate a bit.

Intel need years to remedy this at the transistor and logic layout level, thus a sub-year fix implies knowing several years now about this.
They continued to sell flawed processors without informing clients, and (this is the kicker) they apparently attempted no effort at remediation.

The only way to prevent massive lawsuits is if when the flaw was discovered, certain influential parties wanted it to remain hidden and covered the company.

Your attempt to trivialize this actually exposes Intel even more.

edit:
I hope you realize that even if Intel is given retroactive immunity as were the telecoms with monitoring, then every foreign and most US customers will not trust them, for at least several years. Talk about a Pyhrric victory.

 

[DaveSimmons]

Did you read how they are fixing the chips?

"Intel also is fixing the problem in future chips, starting with products that will arrive later this year, Smith said. Intel is effectively taking the software fixes being released now and building them directly into hardware, he said."

Designing a safe version of speculative execution that is as fast and aggressive as it is now might take years, but no imaginary bridges are required to hack in some cache clearing.

 

[LTC8K6]

Why can't we just wait and see what happens with the chips Intel releases later this year, as claimed?

What's the point of pointing and jeering at me now for posting what Intel says?

I don't get it.

Intel knew for at least 7 months, and maybe more if Intel already knew before they publicly knew.

 

[AdamK47]

I'd like to point out that there isn't a bug. There isn't a defect. The architecture is working as intended. This is an exploit.

 

[DaveSimmons]

I'd like to point out that there isn't a bug. There isn't a defect. The architecture is working as intended. This is an exploit.

Good point. To repeat myself from another thread, this is information leakage through an unexpected method, like bouncing a laser off a window to listen to a conversation, or looking at the wear pattern on a keypad to narrow down the possible passwords.

 

[LTC8K6]

I'd like to point out that there isn't a bug. There isn't a defect. The architecture is working as intended. This is an exploit.

Yes, that point is getting lost...

 

[richaron]

I'd like to point out that there isn't a bug. There isn't a defect. The architecture is working as intended. This is an exploit.

I'd like to point out this is wrong. Intel CPUs have a bug.

The architecture is designed so it doesn't allow lower level processes to access privileged data. Intel put lots of time and effort and silicone into not allowing lower level processes to access privileged data. And everyone thought they achieved their design goals. But they missed something which resulted in an exploit:

Therefore intel CPUs have a bug.

 

[IEC]

I'd like to point out it doesn't matter what you name it, the impact does not change.

 

[whm1974]

I'd like to point out it doesn't matter what you name it, the impact does not change.

In any case how many of us are actually even affected by this anyway? Maybe it is just too soon to tell right now.