Slashdot Mirror
Intel Launches 8th Gen Core Series CPUs With Integrated AMD Radeon Graphics (hothardware.com)
MojoKid writes: At CES 2018, Intel unveiled more details of its 8th generation Intel Core processors with integrated AMD Radeon RX Vega M graphics. Like cats and dogs living together, the mashup of an Intel processor with an AMD GPU is made possible by an Embedded Multi-Die Interconnect Bridge (EMIB), which provides a high-speed data interconnect between the processor, GPU and 4GB of second-generation High-Bandwidth Memory (HBM2). Intel is delivering 8th generation H-Series Core processors in 65W TDP (laptops) and 100W TDP (desktops) SKUs that will take up 50 percent less PCB real estate, versus traditional discrete configs. Both the mobile and desktop variants of the processors will be available in Core i5 or Core i7 configurations, with 4 cores and 8 threads, up to 8MB of cache and 4GB of HBM2. The 65W mobile processors can boost up to 4.1GHz, while the Radeon RX Vega M GL GPU has base/boost clocks of 931MHz and 1011MHz, respectively. The AMD GPU has 20 compute units and memory bandwidth checks in at 179GB/s. Desktop processors ratchet the maximum boost slightly to 4.2GHz, while the base/boost clocks of the Radeon RX Vega M GH GPU jump to 1063MHz and 1190MHz, respectively. Desktop GPUs are also upgraded with 24 CUs and 204GB/s of memory bandwidth. Intel says that its 8th generation Core i7 with Radeon RX Vega M GL graphics is up to 1.4x faster than a Core i7-8550U with an NVIDIA GeForce GTX 1050 GPU in a notebook system. System announcements from Dell and HP are forthcoming, with availability in the first half of this year. Intel has also launched a new NUC small form factor gaming mini PC based on the technology as well.
Does it run Crysis?
Meltdown?
Count me in!
Did anyone at Intel even think about the probable reaction to this announcement?
Seems like Meltdown and Spectre fixes might not be making this build.
Who is going to buy these CPUs? Intel knew about the bug for half a year and did not scratch the release? Linus was right then. They do intend to keep selling shit.
How long are you going to spam this?
Until the Intel spin doctors stop posting misinformation. Meltdown is an INTEL ONLY ISSUE. These processors are defective and need a redesign.
So if you buy an Intel CPU you will need to deal with Meltdown.
By using an up-to-date version of Windows, Mac, or Linux. And if you don't buy an Intel CPU, you still need a very similar fix to mitigate Spectre.
Like it or not, this is not going to end in a mass recall - it's a fundamental design flaw with speculative execution. Poor security is a bug, but they delivered everything they promised on the box. This is not a simple patch. It's a complete redesign of a huge section of the die - and even if you do get your recall, it's going to be years before a new design can be fabricated and mass-produced. By then, it will be replaced and obsolete.
I'm glad I don't have to give Intel road maps anymore. This is an 8th gem processor with i5/i7 core in how many chip sets, memory types, gpu types? AMD parts. Why not throw a RISC processor in to be sure.
Cryptocurrency miners will buy up all the supply as usual, due to it being AMD.
Intel says that its 8th generation Core i7 with Radeon RX Vega M GL graphics is up to 1.4x faster than a Core i7-8550U with an NVIDIA GeForce GTX 1050 GPU in a notebook system.
That's actually a very impressive result. If I was buying a gaming machine, a 1050 is about the minimum. A 1050 TI would be better and it seems like this chip is in 1050 Ti territory.
echo -e 'global _start\n _start:\n mov eax, 2\n int 80h\n jmp _start' > a.asm; nasm a.asm -f elf; ld a.o -o a;
...even if he's also my enemy?
I'm confused about AMD's business strategy with this move. They just finally got their foot back in the door with Ryzen being competitive against Intel after a decade of falling behind in performance. Topping this off of Intel not having a competitive solution to AMD's APU with a decent-performing 3D GPU, and AMD finally seemed poised to grow its share in the laptop & desktop market.
But partnering with Intel to create an Intel APU defeats the purpose of buying an AMD APU.
I suspect AMD has accepted that they will never ever come close to the market share they had back in the late 90's and early 2000's, and therefore it's better financially to sell large volume's of AMD-GPU-on-Intel-CPUs chips, than small volumes of AMD APU chips.
Intel justifies the higher prices of its CPUs by claiming 5-30% increase in processing speed over their competitors. Intel says this increase in performance really makes a difference.
But now a bug in Intel's CPUs require a 5-30% decrease in performance to fix. Intel says this decrease really makes no difference.
Maybe instead of being a lying sack of shit company, Intel should tell the true, fix the problem, and repair/replace/reimburse people that bought Intel's defective junk.
MELTDOWN and SPECTRE
> And if you don't buy an Intel CPU, you still need a very similar fix to mitigate Spectre.
Fixing meltdown requires kernel unmap/KPTI. This comes with a 20-30% tax for many workloads. I know people at hosting firms with shit melting down because they've installed the patch and now they have degenerate end-user workloads.
Spectre requires a million little fixes that are generally cheap in performance.
According to Intel the 100W TDP parts i7-8809G, https://ark.intel.com/products..., and i7-8709G, https://ark.intel.com/products..., are for mobile use.
Are they now trying to melt there CPU's literally?
https://www.nytimes.com/2018/01/07/technology/intel-chip-security.html
Why not include an integrated AMD CPU while they are at it.
Wrong. Even KPTI won't "fix" Meltdown. It will only mitigate the issue. Fixing requires replacing the CPU.
We need a final solution to the degenerate end-user problem.
echo -e 'global _start\n _start:\n mov eax, 2\n int 80h\n jmp _start' > a.asm; nasm a.asm -f elf; ld a.o -o a;
Change log:
2018/01/01 - Added 14 Useful Links. Disable Intel ME 11 via undocumented NSA "High Assurance Platform" mode, Blackhat Dec 2017 presentation, Intel ME CVEs (CVSS Scored 9.0-10.0)
Intel CPU Backdoor Report
The goal of this report is to make the existence of Intel CPU backdoors a common knowledge and provide information on backdoor removal.
What we know about Intel CPU backdoors so far:
TL;DR version
Your Intel CPU and Chipset is running a backdoor as we speak.
The backdoor hardware is inside the CPU/Bridge and the backdoor firmware (Intel Management Engine) is in the chipset flash memory.
30C3 Intel ME live hack:
[Video] 30C3: Persistent, Stealthy, Remote-controlled Dedicated Hardware Malware
@21:43, keystrokes leaked from Intel ME above the OS, wireshark failed to detect packets.
[Quotes] Vortrag:
"the ME provides a perfect environment for undetectable sensitive data leakage on behalf of the attacker".
"We can permanently monitor the keyboard buffer on both operating system targets."
Backdoor removal:
The backdoor firmware can be removed by following this guide using the me_cleaner script.
Removal requires a Raspberry Pi (with GPIO pins) and a SOIC clip.
2017 Dec Update:
Intel ME on recent CPUs may be disabled by enabling the undocumented NSA HAP mode.
Decoding Intel backdoors:
The situation is out of control and the Libreboot/Coreboot community is looking for BIOS/Firmware experts to help with the Intel ME decoding effort.
If you are skilled in these areas, download Intel ME firmwares from this collection and have a go at them, beware Intel is using a lot of counter measures to prevent their backdoors from being decoded (explained below).
Useful links (Added 2018 Jan 1):
Disabling Intel ME 11 via undocumented mode (NSA High Assurance Platform mode)
Blackhat 2017: How To Hack A Turned Off Computer Or Running Unsigned Code In Intel Management Engine
EFF: Intel's Management Engine is a security hazard, and users need a way to disable it
Sakaki's EFI Install Guide/Disabling the Intel Management Engine
Intel ME bug storm: Hardware vendors race to identify and provide updates for dangerous Intel flaws.
CVE-2017-5689: An unprivileged network attacker could gain system privileges to provisioned Intel manageability SKUs
CVE-2017-5705: Multiple buffer overflows in kernel in Intel Manageability Engine Firmware
CVE-2017-5706: Multiple buffer overflows in kernel in Intel Server Platform Services Firmware
With a Beowulf cluster of these....
Putting AMD hardware next to yours doesn't automatically grant immunity from Meltdown and Spectre. Nice try, though.
Still waiting on Serviscope_minor to wake up to fucking reality and realize that Jessica Price isn't going to fuck him.
Intel ME on recent CPUs may be disabled by enabling the undocumented NSA HAP mode, use me_cleaner with -S option to set the HAP bit, see me_cleaner: HAP AltMeDisable bit.
When buying this processeur, you know it haves Spectre and Meltdown flaws... so it's now a feature, not a bug. Is Intel are the one who leak this information before the CES ?
But what is the multiple of Intel + Radeon, compared to Intel + Intel HDxxxx? 30x? 50x? 100x? Snicker... Even Iris Pro was more like Iris The Angry Flower.
Intel graphics have always been the worst in the industry. What's impressive is that Intel seems to finally want to do something about that.
> Spectre requires a million little fixes that are generally cheap in performance.
And it's still not entirely clear that Spectre is that big a problem, since you need to find an application that has this specific vulnerability, and have so much control over it that you can both train the branch predictor, and inject whatever data you want, and be able to time its execution.
All the proof of concepts I've seen so far run in a single process.
AMD RYZEN plus an NVidia, because... why not?
What kind of retards keep building new chips that are already vulnerable to hacks? Goofs and arseholes.
Maybe they'll use the GPU to counter losses in the CPU?
Intel CPUs are the more expensive ones. They run at about the same temperature, and Ryzen is every bit as good at multithreading as Intel chips.
Intel having competition means us consumers win.
eh, I read AMD is only vulnerable to Spectre variant 1
you have source for your claim anything but Intel and certain ARM are vulnerable to Meltdown?
idiot here. why dont we have more chips like this already? i always thought that combining the graphics chip with a cpu was a no brainer. anyway, great slashdot story if its mostly numbers and abbreviations. woo hoo tech!!!
Stop talking about Spectre. There is no fix for Meltdown. Meltdown is much more severe. You cannot fix it in software, only mitigate it. You have been fooled, or you are being intentionally misleading.
This is wrong, you can always unmap all pages a process isn't supposed to have access to. That's not even hard to do, the hard part is doing it in a way that doesn't severely hurt context switching performance. If a page is not mapped, the processor won't read it, and meltdown can only be used to "read" a process's own pages. ... and then there is specter.
If you think these two attacks contain the ONLY examples of transient instructions that can cause data to be leaked through side channel timing attacks, better practice your surprised face now. Neither AMD, Intel, or any other desktop processors are impervious to this class of attack. Duh.
I understand Intel wanting a GPU to pair efficiently with their CPUs for the smallest form factors... but I don't see why AMD and not NVidia. Did NVidia turn them down? Or does Intel really consider NVidia, who doesn't make AMD64 chips, to be a bigger threat than AMD? Or is there something inherent in the GPU platforms that makes AMD possible but not NVidia?
... until I see a PRICE.
Because that is point 1 on why I don't buy Intel.
The other points are currently the IME and the unfixable hardware security leaks that I'm sure won't be fixed, or it would be the first thing mentioned in TFS. ... if I had any.)
(Yes, AMD has a IME equivalent too, as does ARM. So I buy nothing currently. But I would actively get rid of anything Intel on top of that
... with an offer they can't refuse.
I know Intel. I'd also not expect Hannibal Lecter to not try to eat my brain. I'd sooner see a honest politician than Intel not trying to create more monopoly with any psychopathic method physically possible that will let them get away with it.
Nor one thinking further ahead than your goldfish nose.
Yeah, you get some peanuts now. And pay with being tied to your biggest enemy, who has shown the will to murder you in cold blood many, *many* times, which he will obviously abuse at the first chance he gets, and every chance after that.
I bet you're so black-eyed that you still believe in the "the NSA are our friends" anticonspiracy theory.
Yes. It says on the box: "Intel inside".
Does this have the security bug the older CPUs have?
This comes with a 20-30% tax for many workloads.
Except it didn't. All benchmarks point to a 5-10% worst case hit, and an unmeasurable hit in pretty much all desktop / user facing workloads. Despite all the initial reports I've yet to see any benchmark, Windows, Linux, server loads, office applications, gaming, databases, or whatever get into the double digits.
Here's just some top google results:
http://www.guru3d.com/articles...
https://www.techspot.com/artic...
And here's some Linux ones on KPTI:
https://www.phoronix.com/scan....
No, the example code only affected Intel. The underlying exploit exist in AMD and ARM64 as well.
The original paper postulated as to why their code wouldn't run on AMD and even went so far as stating it may work on AMD with some optimization or on chips with different execution pipeline sizes.
https://meltdownattack.com/mel...
Because otherwise I'm not buying it. J/K, I'm never buying Intel again regardless.
Easy there Intel shill; your employer's PR is bad enough.
And a Partridge in an insecure Kernel memory tree!
Yes, they postulated because they had no clue about the implementation.
Intel decided to skip the permission check when reading memory, leaving it to the exception handler to clean the resulting mess up. AMD and all except A75 Arm processors check the permissions BEFORE and thus are not vulnerable no matter how much you twiddle and optimize.
What is it with you people that you take a "might" in a paper of people with no idea about the microarchitecture as a proof that they are affected, when the people who actually KNOW say it's not possible?
The only thing the paper says btw. is that other processors also execute instruction streams beyond an exception. That is obvious, otherwise out-of-order won't really work. It also is not a problem. The problem is actually giving the thread access to data IT IS NOT ALLOWED TO ACCESS. That is the serious mess-up on Intel's side (and, one Arm CPU).
Don't think its enough to game on and if your doing occasional video editing it might be OK. I want to know how much more I would pay over a Intel GPU solution which these days isn't that bad for the basics and a bit more. I suspect the best selling point would be battery life, over a separate dedicated solution. This ended up being what AMD did with its APU's. Not so much great gaming rigs, but better power management with a bit more GPU power.
If you're after gaming performance, it makes sense to go with an AMD CPU and kick the savings towards a better GPU. Dumbass.
Honestly I think the IME debacle is a far bigger issue for anyone planning to buy Intel. A difficult-to-leverage hardware flaw that allows logical boundary traversal is absolutely worrying, but a hidden and secret firmware designed specifically for spying and remote control is fucking terrifying.
Then you've not looked very hard. Things that are very syscall intensive, like du across a filesystem, have been shown to pay a 50% tax. Tun/tap stuff in userspace pays a ~35% tax. Or if you want something higher level:
https://www.postgresql.org/mes...
Trivial PostgreSQL txns (where you don't end up IO or scanning bound, but instead are measuring system call path) pay a 17% tax. On Skylake. Where the penalty is comparatively less.
The fact that you constantly spam this over and over is only making it look like you are the one spreading misinformation.
My 486DX2/66 is unaffected. Apparently so is my Raspberry Pi. So is the Commodore 64, though with that machine you can read whatever memory you want any time you feel like it without any fancy tricks.
it's a fundamental design flaw with speculative execution. Poor security is a bug, but they delivered everything they promised on the box. This is not a simple patch. It's a complete redesign of a huge section of the die
It is a design flaw with Intel's implementation of speculative execution which does speculative loads before checking permissions. The only necessary redesign is to check the permissions before executing the speculative load instead of at instruction retirement.
The only benefit to what Intel did is to execute speculative code that causes faults more quickly.
Stop talking about Spectre. There is no fix for Meltdown. Meltdown is much more severe. You cannot fix it in software, only mitigate it. You have been fooled, or you are being intentionally misleading.
Meltdown is the one which can be fixed in software. Flush the TLBs and do not share them between privilege levels. Unfortunately doing so incurs a significant performance penalty during privilege changes.
No, the example code only affected Intel. The underlying exploit exist in AMD and ARM64 as well.
The original paper postulated as to why their code wouldn't run on AMD and even went so far as stating it may work on AMD with some optimization or on chips with different execution pipeline sizes.
https://meltdownattack.com/mel...
However, for both ARM and AMD, the toy example as described in Section 3 works reliably, indicating that out-of-order execution generally occurs and instructions past illegal memory accesses are also performed.
The paper shows a timing variation indicating that speculated instructions caused memory access but *not* that the specific memory was accessed, For instance those may have been page table accesses needed to load a TLB before permissions were checked. Nothing indicates that speculated instructions which relied on a load from the targeted protected memory location were executed.