Jump to content

This dont look good, Intel design flaw reduce performance


Recommended Posts

We may not all die just yet.

 

MS has been busy finding out how much their systems are affected and provide gidance here for the common punter:

https://support.microsoft.com/en-us/help/4073119/protect-against-speculative-execution-side-channel-vulnerabilities-in

 

 


Windows 7 Service Pack 1 Still most stable and best FPS in IL2 i tried out of 3 OS versions. Whats gonna happen if i dont update it.

Nothing. Provided you don't access untrusted data ("web pages").

 


They will steal my il2 password?

If you post it on the web, yes. Else, your game rig is probably more interesting for hosting "stuff", mining or "sending Email". Honestly, what is the value of a warezed il-2 password?

 


Steam account?

More likely. But only if you read and execute untrusted data on your rig.

 

Nothing else of importance on pc why would there be.

Depending on what may be hosted on your rig (see your second question), you might revise that assessment. Remember that in this new world, you may have to prove your innocence, when they have caught a terrible villain.

 

Link to post
Share on other sites

So, things get a bit clearer:

 

Intel's CEO Brian Krzanich just announced the microcode updates for their CPU's, they'll provide microcode updates adressing branch target injection issues (Spectre), of course withholding anything specific.

 

There are new BIOS' coming out that patch the CPU's microcode. It will, as Krzanich admits, "reduce CPU performance depending o workload", whatever that means.

 

But as far as I can see now, using this exploit, no one could actually read beyond L1 cache data, so the exploit is not as bad as some people said. Certainly not on your gaming rig. So, if you update your BIOS and CPU, you may want to run chili's benchmark for IL-2 before and after the update and post your results. I suspect, a lot of people will be hesitant to update the BIOS, as you may be able to roll back on the BIOS, but once your CPU microcode is updated, I wouldn't know of a way rolling back.

Link to post
Share on other sites

https://www.bloomberg.com/news/articles/2018-01-08/-it-can-t-be-true-inside-the-semiconductor-industry-s-meltdown

 

Regardless of how it will end up and how it will affect performance it is an interesting story to read. The activity of major software developers in this regard also suggests that it is actually serious. 

If you are running your critical infrastructure on somebody elses computer, and that computer runs *anyones* software also, you should rightfully be concerned.

 

As it stands now, MS is just shoving down patches to everyone, today one came down and Microsoft was getting a bit specific what it means for us.

 

In short, if you are running Win7, expect to get neutered. If you are running Win10, not so much. (Who would have guessed). If you are running older CPU than 6xxx series Core-blabla, expect to get neutered.

 

All depending on the kind of use, and we "may expect benchmark results in the coming weeks." (Or not.)

 

So if you thought your 4790 was a great CPU, well, it *was*.

 

Man, I hope I'm wrong.

Link to post
Share on other sites

 

 

If you are running older CPU than 6xxx series Core-blabla, expect to get neutered.

 

 

So if you thought your 4790 was a great CPU, well, it *was*.

 

Since Ryzen+ is just around the corner and will be available in April I guess I will just switch sides. 2600k served well, reached 5.1 Ghz on water and gave me fun as a platform but after 5 years its time to drop Intel for good.  

  • Upvote 1
Link to post
Share on other sites

Since Ryzen+ is just around the corner and will be available in April I guess I will just switch sides. 2600k served well, reached 5.1 Ghz on water and gave me fun as a platform but after 5 years its time to drop Intel for good.  

 

I bought my 2500k Q1 2011 - about the first month they were available. It has been the best damn investment -as far as PC parts go- I have ever made. It's the CPU that just keeps on keeping on!

I have been contemplating an upgrade in the last month or so, but then the Meltdown/Spectre news hit. So I think it will be worth waiting to see how things pan out in the CPU world over the next couple of months.

 

CVE-2017-5754 (the freq-nasty of the 3 exploits) is a major headache for Intel, so I am looking forward to Ryzen+ in April. I expect I will be switching sides then too.

Link to post
Share on other sites

Since Ryzen+ is just around the corner and will be available in April I guess I will just switch sides. 2600k served well, reached 5.1 Ghz on water and gave me fun as a platform but after 5 years its time to drop Intel for good.

Why? AMD is affected as well. MS clearly stated that.

Link to post
Share on other sites

There are 3 exploits. 1 of them seems to effect all CPU's, including AMD and ARM. This error was patched in Windows, Linux, OSX/IOS and others. The other 2 exploits are Intel only and are much more difficult to fix. Intel is working with motherboard manufacturers to develop BIOS updates to reduce the risk of the exploit, but it's my understanding it can't be truly fixed software side.

Link to post
Share on other sites

Why? AMD is affected as well. MS clearly stated that.

What BeastyBaiter said is true: https://www.amd.com/en/corporate/speculative-execution

I genuinely find Ryzen more attractive and appealing as a new cpu. It offers me more than Intel does and doesnt restrict me in any way. I've used 2600k for years and I saw no reason to change to Ivy Bridge, Haswell, Broadwell, Skylake and even Kabylake. Coffelake sounds tempting but it's still the same arch as Skylake, just improved, refreshed and with added two more cores for the price of massive power draw and need to delid cpu if I wish to go beyond 4.8 Ghz. And the never ending need to change platforms because reasons. 

AMD offers an improved version of their last year success that is backward and forward compatible. It offers twice as many cores and threads and as it turns its also more secure. At least this time, because who knows, maybe next time something new will expose flaws in Ryzen. But until then AMD is just more convincing to me. 

Link to post
Share on other sites

Just be aware guys, there is more than Windows updating the operating system.

 

Motherboards will need bios updates as well, as I understand it.

I know EVGA is currently working on new bios updates for their line of boards down to the X-79 series to address this issue.

 

Edit: As BeastyBaiter said above, did not notice his post before posting this.

Edited by dburne
Link to post
Share on other sites

What BeastyBaiter said is true: https://www.amd.com/en/corporate/speculative-execution

I genuinely find Ryzen more attractive and appealing as a new cpu. It offers me more than Intel does and doesnt restrict me in any way. I've used 2600k for years and I saw no reason to change to Ivy Bridge, Haswell, Broadwell, Skylake and even Kabylake. Coffelake sounds tempting but it's still the same arch as Skylake, just improved, refreshed and with added two more cores for the price of massive power draw and need to delid cpu if I wish to go beyond 4.8 Ghz. And the never ending need to change platforms because reasons. 

AMD offers an improved version of their last year success that is backward and forward compatible. It offers twice as many cores and threads and as it turns its also more secure. At least this time, because who knows, maybe next time something new will expose flaws in Ryzen. But until then AMD is just more convincing to me. 

Maybe you don't know it yet, but it will reduce severely your performance in most VR games, and in IL-2's case, make it sheer impossible to go there. Just a heads-up for you, if you want to go there in the near future, so you don't run into a wall later. :salute:

Edited by 4./JG52_Fenris_Wolf
Link to post
Share on other sites

I'm very aware that AMD usually looses in games that have little to no multicore utilization (such as BoS, DCS and so on) and it also has impact on VR. But since I've been using 2600k and made some tests to compare where it gets me with current generation of Ryzens I am quite sure it wont be a reduction of performance in any way.

It's not, its primary issues are:

- IPC lower by 3-5 % depending on the type of calculations;

- Clock disadvantage due to manufacturing process, there is no doubt that Intels 14nm is superior to what Global Foundaries delivers and thus most Ryzens reached limits at 3.8 to 4.0 Ghz (albeit some managed to sit safely at 4.1 Ghz);

- Cache latency which between each CCX was a bit too high.

All the three, and particularly second and third problems, were of primary concern of AMD and that is why they shifted to 12nm process to allow for higher clocks to narrow the gap. Its also claimed Ryzen+ will come with better ram compatibility. 

What will be results is unknown at this point but I'm very keen on trying that. 

Link to post
Share on other sites

Ok, Intel now gives some Benchmarks, showing what the new microcode (that you get with a new BIOS) plus the MS patches do to your 6th, 7th and 8th generation Intel CPU.

 

In short, an 8700K will lose, depending on the Benschmark, between 0% (mostly around 4%) and 22% performance.

 

It is of note that the heavy impact comes not from loss of CPU processing power (you will not lose much FPS), but from the halved(!) SSD 4k random i/Ops, as shown by the German tech magazine c't. So your Samsung 960 Pro drops from roughly 200k i/Ops (read and write) to about 100k i/Ops.

 

I'm certainly not gonna install the new BIOS. They can side channel me all day long (how much easyer is it just to open the box through other exploits anyway), but halving the quality of an SSD that matters that much is just gaga.

 

Besides, Intel wasn't lying when it said that there is no error in the CPU regarding that. It is just that you could exploit the CPU when it was working as intended. No wonder Intel frowned upon anyone daring to say (for years!) that there might be issues like that.

  • Upvote 1
Link to post
Share on other sites

For those concerned about BoX performance, here are my results for the balapan test track:

 

Pre-patch:

Min: 60

Max: 91

AVG:87

 

Post-patch:

Min: 68

Max: 91

AVG: 87

 

This is with an i7-8700k at 4.7Ghz (no avx offset) with an Oculus Rift. So yeah, no real change.

 

 

Well given that you have the newest i7 processor it is unlikely to be the bottleneck for your system, so performance reduction on this should not show much. On my i4820k i7 with my Gtx1070 the processor and graphics card are quite on par, with the processor a bit behind already. Any reduction on CPU performance on my system will directly translate to lower FPS.

 

In the last 2 days I noticed my FPS having dropped around 30 fps (90-100 to 60-70) and wondered why, maybe the patch installed whithout me noticing it? Will check later. If so I hope I can roll this back, can still make my important stuff from my laptop.

Edited by 216th_Jordan
Link to post
Share on other sites

Well given that you have the newest i7 processor it is unlikely to be the bottleneck for your system, so performance reduction on this should not show much. On my i4820k i7 with my Gtx1070 the processor and graphics card are quite on par, with the processor a bit behind already. Any reduction on CPU performance on my system will directly translate to lower FPS.

 

In the last 2 days I noticed my FPS having dropped around 30 fps (90-100 to 60-70) and wondered why, maybe the patch installed whithout me noticing it? Will check later. If so I hope I can roll this back, can still make my important stuff from my laptop.

The Windows patch is not really bad. Did you also install the new BIOS and thus upgrade CPU microcode? You should meter your game loading times before updating and after that.

 

It is really SSD performance that get tanked. Also things like font rendering in Win7.

Link to post
Share on other sites

Maybe you don't know it yet, but it will reduce severely your performance in most VR games, and in IL-2's case, make it sheer impossible to go there. Just a heads-up for you, if you want to go there in the near future, so you don't run into a wall later. :salute:

Do you have a source for this? How can these assumptions be made before the hardware patches are even out?

Link to post
Share on other sites

Well given that you have the newest i7 processor it is unlikely to be the bottleneck for your system, so performance reduction on this should not show much. On my i4820k i7 with my Gtx1070 the processor and graphics card are quite on par, with the processor a bit behind already. Any reduction on CPU performance on my system will directly translate to lower FPS.

 

In the last 2 days I noticed my FPS having dropped around 30 fps (90-100 to 60-70) and wondered why, maybe the patch installed whithout me noticing it? Will check later. If so I hope I can roll this back, can still make my important stuff from my laptop.

 

Actually, the i7-8700k is the bottleneck in BoS in VR. I suspect it is less effected than older models, but my system certainly isn't immune. We'll have to see what a bios update does. I might wait on that part, I rather like my Samsung 960 Evo speeds.

Edited by BeastyBaiter
Link to post
Share on other sites

Hmm, I played this evening and I noticed that my game seemed very smooth, I was however getting a bit nauseous and it appeared like I couldn't sense the motion of the other planes well.

 

After I uninstalled the update again the game was a little bit more jumpy when doing TrackIR movements as before but I didnt feel nauseous any more and the perception of trajectories was a lot better again. fps remained only a tad better though.

 

It seems like the update has caused some kind of input lag while smoothing out the frames, I really don't know..

Link to post
Share on other sites

Do you have a source for this? How can these assumptions be made before the hardware patches are even out?

Context, my friend. It was referring to Ryzen performance in VR, a reply to an off-topic post in this thread, and not about the CPU issue itself.

 

Our own collected data on performance is the source, look it up yourself in chiliwili 's performance comparison thread in VR. It's not an assumption, it's fact.

 

The underlying issue is that IL 2 as well as many other VR games run their main on a single thread, and are choking on a single core of the Ryzen's multicore architecture. Just FYI. Well, safe flying.

Hmm, I played this evening and I noticed that my game seemed very smooth, I was however getting a bit nauseous and it appeared like I couldn't sense the motion of the other planes well.

 

After I uninstalled the update again the game was a little bit more jumpy when doing TrackIR movements as before but I didnt feel nauseous any more and the perception of trajectories was a lot better again. fps remained only a tad better though.

 

It seems like the update has caused some kind of input lag while smoothing out the frames, I really don't know..

If that is true, that is terrible news, especially for VR users. Will have to test myself when back on Feb 1st.

Link to post
Share on other sites

If that is true, that is terrible news, especially for VR users. Will have to test myself when back on Feb 1st.

 

I think this likely more a problem of older 22nm and up CPUs. But I was always combating this experience in  the last months also and only a few weeks ago found a setting that worked very well for me without being a pain in the eyes and brain. So I'm thinking this is also very setup dependent.

Didn't want to buy myself a new CPU in the next 2 years however...so I'm a bit trapped now.

Link to post
Share on other sites

Has anyone with a venerable Sandy Bridge i7 and win 7 who has installed the Microsoft update and if so what was the FPS impact? The combo of older cpu's and win 7 appears to be the most impacted. My 3820 has no issues running BOX at

Max settings at HD resolution. Wondering if that will continue to be the case if I install the update. This whole thing is such bs!!

Link to post
Share on other sites

Thanks for the report =LD=Hiromachi. Not terrible results on the Sandybridge but not good either. With so many of us using older cpu's (because they still perform well vs newer versions)I would hope there would be more testing of the older units particularly running on Win 7 which most still use. It is aggravating that there is really no testing on older setups. Intel's testing totally excluded anything but their newer units. I am not installing any updates until I have a good idea of how it will impact my rig and how it runs the games I play. Hopefully some with older cpu's who decide to install the updates can comment on how BOX or other games run afterward.

Link to post
Share on other sites

It almost seems like most of the [best]information about this is coming from non-English language sources.

 

"The Wall Street Journal is reporting it got its hands on a confidential memo issued by the company and shared with large companies and cloud providers not to install the patches."

It appears that Intel is not rolling out microcode patches with OS updates because it seems in some instances it sent those machines into reboot cycles.

 

The real issue with Spectre etc. is that now CPUs have become the target of hackers. Before it was the programs and the OS, now hardware has become sufficiently complex (for more than a decade) for „successful unintended use“.

 

What we see here is not the problem with a patch, but the dawn of a new field of attacks, requiring maintenance („updates“) that is not implemented yet. It means that you buy ahardware according to certain specs now, those specs will change over time in an unforseen way, making it much harder to plan required hardware installations.

 

What is also of note that this patch devastated a good margin of IPC advantage it had over AMD. I‘m sure at AMD, Champagne was handed out once those news of Spectre broke. Intel will get hurt where it hurts them most.

Link to post
Share on other sites

At this point, these bugs are not terribly concerning for current desktop computers. The performance losses are small, and those who cannot afford them can easily skip any updates. 

 

However, the future of processors in general is quite worrying. Many people had been hoping for incremental IPC and clock improvements to help with software such Il-2. The recent bugs might be a serious setback. I imagine that unreleased products will still be affected for quite some time. 

 

With the next generation of VR devices seemingly around the corner, any stagnation in CPU performance is very sad. I hope that the industry will be able to deliver improved products in this critical time. The software that we love is always very close to the edge of the performance envelope. 

Edited by Mitthrawnuruodo
Link to post
Share on other sites

 

 

At this point, these bugs are not terribly concerning for current desktop computers

I think the part where javascript code from the web can see anything your browser has in memory is pretty scary.

For the rest, I agree. Today's OSes such as Windows and Linux have a security and separation model that's pretty pointless on the desktop. They make sense in a multi-user environment, but at home it's more important to keep separate programs apart, and doing so without losing comfort of use isn't easy.

Devs reacted strongly to the address space separation issue, but for me, I don't care if processes can see the kernel's memory. I'm only worried about them seeing other proc's memory, or simply reading parts of the disc they aren't supposed to read, such as my private SSH keys, IL-2 accounts passwords...

Link to post
Share on other sites

I think the part where javascript code from the web can see anything your browser has in memory is pretty scary.

And this you can fix with proper code that makes your webbrowser. Besides, would you tolerate just any Javascript pestering your browser? Even, as in case of Spectre, that requires max use of your system capabilities to get anywhere?

 

The real issue is not what damage that attack vector can do to your PC. It is just proven that, in principle, it is a possible vector to reading what is in other parts of the system memory. This does however not mean that it is a convenient way to do (especially regarding the awfully simple alternatives).

 

What really happened here is that with Meltdown and Spectre, we crossed a rubicon.

 

Hardware has been hacked and thus become a "work in progress" thingy. And this goes two ways. For money, you could get some more MHz or higher IPC on the fly (Porsche is considering doing such with their cars, just log in, zip your credit card and *pooff* your car has extra 25 hp or so). And for "security" reasons, things that made your system fast can simply get disabled. Just like that with an update, as with every boot of your Windows, mcupdate_genuineintel.dll can install new microcode.

 

We see this intended obsolescence now with the speed penalty that Sandy Bridge processors are having compared to Skylake / Coffe Lake ones.With the older CPUs, there is just no "replacement function" in place to make up for the ways that made it fast as well. And you really think Intel would be hard pressed giving old CPUs new features?

 

AMD might be happy that Intels lead in IPC took a serious hit, just like that. On the other hand, AMD, and all other processor manufacturers are certainly not looking in the mouth of a gifted horse, as also they enjoy the new way of breaking old parts for the sake of "safety".

 

From now on, they can break your old thing and telling everyone, even the courts, "that it is good for you".

 

Edit: Good news ist that you are most likely not using Apple. Else, your I/O will REALLY take a hit, small files down to 25% of the original performance. Ouch.

Edited by ZachariasX
Link to post
Share on other sites

Well aside we from all the tech if, ands or buts the way I see it is that the Intels, AMD's and Apples are grossly responsible for the mess. I spend top money on an intel CPU based on the performance it offered. Now to cover for their poor designs driven by the need for speed they are taking away an amount of the performance I paid them for. How does that work? I get that as time passes hacks get the best of the industry but this one at least in part was caused by actions taken to make the cpus faster. I just think the whole thing blows and it's us who's taking the hit. Our choices are update and lose performance, upgrade rigs at high cost or remain vulnerable. No good choices.

Link to post
Share on other sites

AMD seems for now to be holding the better side of the stick. At the release of Ryzen they also mentioned some built in security features, particularly for content creators and server chips. Maybe Zen arch is just that different. Shame Sandy Bridge will suffer, it was from my perspective the best cpu since I got into pc stuff. So many years of flawless work.

Afaik, here is another one:  https://www.extremetech.com/computing/262031-researchers-found-another-major-security-flaw-intel-cpus 

Link to post
Share on other sites

AMD seems for now to be holding the better side of the stick. At the release of Ryzen they also mentioned some built in security features, particularly for content creators and server chips. Maybe Zen arch is just that different. Shame Sandy Bridge will suffer, it was from my perspective the best cpu since I got into pc stuff. So many years of flawless work.

Afaik, here is another one:  https://www.extremetech.com/computing/262031-researchers-found-another-major-security-flaw-intel-cpus 

You can boot and press "Strg P" and then enter the AMT. It is before the BIOS and the problem here is that almost no one knows that such a thing as the AMT BIOS extension (for remote administration in company networks) exists on their business notebooks and if they do the think "they don't use it". Hence, it remains set as "admin". Having this login, you can conveniently override BIOS settings, even BIOS password. So if you don't use AMT, you have to disable it, not just "let it be". Or at least set a good password. Intel knows about that and feels it is up to the notebook vendors to inform their clients.

 

What is more of an issue is that the Intel ME controls everything, is always there and it is broken by design, creating a remote exploit a layer beneath whatever the user can do. And you cannot / must not buy Intel stuff without having the ME. It is just tragic.

 

There is really no need to get worked up too much about someone pestering your browser by means of Spectre when you can have your whole system just like that, because if you're talented enough to make efficient use by means of Spectre, I would bet that a you have access to a kit that can acesses any PC through the ME as well.

Link to post
Share on other sites

Please don't listen to anyone who claims that going AMD Ryzen suddenly causes IL2 to "choke" in VR or anything else.

 

I play at 4k and VR and noticed no serious impact on FPS going from a 4.6GHz 4770K to an 1700X clocked at 3.8GHz.

Link to post
Share on other sites

Ryzen falls short of the current unlocked Intel chips in BoS and DCS in VR (and only VR). However, you have to overclock them for that to be the case. BoS and DCS care about single thread performance, my 1600x and i7-8700k were roughly equal at 4.0 GHz, but the 1600x tops out at 4.0GHz while the 8700k will push up to 5.0 GHz without much trouble. That extra 25% clock speed makes a very real difference in VR (difference between a locked 45 fps and a mostly 90 fps with dips to 45).

 

That said, I like Ryzen. If you are playing properly threaded VR games or non-VR games, the value of the Ryzen chips is unrivaled. The R5 1600x is a seriously sweet CPU for the money and I highly recommend it unless playing BoS/DCS in VR. And it really is just those two games, everything else in VR runs great with Ryzen. With the recent Ryzen price drops, especially at the top end, the situation is even better when looking at AMD. But this is OT tbh.

 

Also, a 4th Intel exploit was uncovered last week, unrelated to specter and meltdown.

Link to post
Share on other sites

Please don't listen to anyone who claims that going AMD Ryzen suddenly causes IL2 to "choke" in VR or anything else.

 

I play at 4k and VR and noticed no serious impact on FPS going from a 4.6GHz 4770K to an 1700X clocked at 3.8GHz.

Well he should listen, because it is right. It's no match for an 7700k or 8700k. Now we all now that DCS and IL-2's thread handling are the reason for this (see BeastyBaiter's post), but this does not change the result of outcome. As sad as it is.

 

I know your comment "please don't listen to anyone who says this" was about me, obviously, I'm direct and blunt, and that trigger(ed) some folks online - not IRL among Germans though. I may inform you: Saying "please don't listen to anyone who says this" is neither an argument, nor correlating with the documented performance results, nor productive nor helpful to the discussion.

  • Upvote 1
Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...