Core 9000 series

[AndreiD]

Has any outlet shown any major performance impact on consumer workloads from all the patching? A lot of them had all of the security patches installed for their Intel systems when the 2700X rolled out and nothing really changed performance wise.

On the Xeon side things are more iffy since the patches impact some of those workloads, but then again I wouldn't cry Intel a river over having to deal with angry data center customers.

 

[Panos]

Has any outlet shown any major performance impact on consumer workloads from all the patching? A lot of them had all of the security patches installed for their Intel systems when the 2700X rolled out and nothing really changed performance wise.

On the Xeon side things are more iffy since the patches impact some of those workloads, but then again I wouldn't cry Intel a river over having to deal with angry data center customers.

Only one run a benchmark for the Spectre V4, which came out post 2700X.

Also a previous one for some of the other patches around January

Not all reviewers are patching their boards, probably that will happen if they use Z370 with the 9000 CPUs. Here we might see the impact, assuming windows is also up to date and not some windows 10 mirror they used from the past.

Also here one more more recent than January as above.

Also look some of the comment, because I have seen this on the 8600K

when i bought the 8600k in nov17 it was inebenching at 1420!!! at 5.0ghz.... after januarys meltdown patch i get 1290 lol... whats next after this 1200??? less..

 

[CAT-THE-FIFTH]

Has any outlet shown any major performance impact on consumer workloads from all the patching? A lot of them had all of the security patches installed for their Intel systems when the 2700X rolled out and nothing really changed performance wise.

On the Xeon side things are more iffy since the patches impact some of those workloads, but then again I wouldn't cry Intel a river over having to deal with angry data center customers.

The Witcher 3 dropped by nearly 10% according to DF,and other games seem more like 3% to 4% at most so not really huge. I would argue certain openworld games and those which are modded will probably be the worst affected as they are more I/O bound. At least on my old CPU,modded FO4 seems to have become more stuttery in specific areas(places which have loads of assets streamed),and after reading a few comments on Reddit,people noticed something similar after the OS patches on older CPUs!

At least I suppose it gives me an excuse to move off socket 1155!!

 

[AndreiD]

The 2nd HU video is with the initial immature patching which Microsoft at one point disabled because it was causing issues: https://www.theverge.com/2018/1/29/...essor-bug-emergency-windows-update-reboot-fix
Intel also recalled those fixes: https://www.computerweekly.com/news/252433630/Intel-recalls-botched-Spectre-fix

The 1st HU video is the one with the mature OS & BIOS fixes, where the performance difference is 1-2%, so I guess you proved the point that consumer workloads didn't see any major performance differences?

The 3rd video doesn't even do any testing and comments from videos are meaningless because they don't have a controlled testing environment. If I leave Chrome running in the background I lose 100 points in Cinebench on my Ryzen.


On a related note, I'm curious if the new 9000 series CPUs will have hardware mitigation for Meltdown and Spectre, they announced Cascade Lake and a "H2 2018 Core" will, which could be another token low volume 10nm part or the 9000 series?

 

[CAT-THE-FIFTH]

HU is actually Techspot:

https://www.techspot.com/review/1659-intel-spectre-variant-4-performance-test/

Closing Remarks
Intel anticipated a performance hit in the range of a 2 to 8% when addressing the Variant 4 vulnerability by disabling Speculative Store Bypass, but in our tests it looks more like 1 to 3%. The impact for Linux users appears to align closer to Intel’s claim and we believe this comes down to the Windows scheduler which is less efficient than Linux's.

When patching Variant 1, 2 and 3 we had found reduced gaming performance on the i7-8700K by up to 5%, though for the most part we saw a 0 to 3% decrease in frame rates. Variant 4 has seen a further 1 to 3% dip, though this time we were testing with the non-K 8700, but the margins should be much the same across the entire range.

In other words, since December the gaming performance of Intel Coffee Lake CPUs is down by 1 to 6%, or about 1-2 fps in games pushing over 60 fps and up to 5 fps for high refresh rate gaming. Definitely not a huge deal overall, but it’s worth keeping in mind that Intel will suffer an IPC hit because of this with future architectures that address these vulnerabilities at the hardware level.

Those of you fortunate enough to own an 8th generation Intel Core processor don't need to worry about your games turning into a slideshow as the performance impact is not significant. But even if it was, we would have strongly recommended you update your BIOS and enable the mitigations as soon as possible.

 

[AndreiD]

I guess it's semantics at this point? 1 to 6% with all of the patches applied doesn't seem like a major hit. It is allowing AMD to catch up more though.

 

[gavinh87]

I guess it's semantics at this point? 1 to 6% with all of the patches applied doesn't seem like a major hit. It is allowing AMD to catch up more though.

A far cry from the 25% claimed in here isn't it?
Honestly, so much fud gets thrown around in this forum.

 

[beany_bot]

A far cry from the 25% claimed in here isn't it?
Honestly, so much fud gets thrown around in this forum.

25% drops in Far Cry what!?!?!

...only joking. And just showing how Chinese whispers happen. Lol

 

[razorpakk]

Sorry for the stupid question, but if the user doesn't apply bios update and whatever else is required to solve these issue what would actually happen?

Like we're supposed to be vulnerable to what exactly? Who's coming for us?

 

[CAT-THE-FIFTH]

I guess it's semantics at this point? 1 to 6% with all of the patches applied doesn't seem like a major hit. It is allowing AMD to catch up more though.

As I said for most games it's under 5% outside the odd open world game like W3 and apparently FO4(well at least on older CPUs),although FarCry Primal was fine.

 

[pastymuncher]

Sorry for the stupid question, but if the user doesn't apply bios update and whatever else is required to solve these issue what would actually happen?

Like we're supposed to be vulnerable to what exactly? Who's coming for us?

As Gigabyte and Asus has abandoned everyone still on the Haswell platform we have had not bios updates for these vulnerabilities, only the software updates through Windows plus the latest versions of Firefox so both our pc's here are still vulnerable. If I remember right though, doesn't someone need physical access to the pc at admin level or something similar to be able to exploit these vulnerabilities?

The 9700k interests me a lot but as long as they keep launching cpu's that are still vulnerable to attack I will not be going anywhere near them. What's going to be their first "fixed" release, Icelake? I should be able to hang on until then and make a choice between that and whatever AMD has then.

 

[kevman]

Quick noob q about cores and hyperthreading. If an older game utilises hyperthreading and has been programmed to run on 4c/8t processor what happens on 8c/8t?

 

[CAT-THE-FIFTH]

Quick noob q about cores and hyperthreading. If an older game utilises hyperthreading and has been programmed to run on 4c/8t processor what happens on 8c/8t?

Games tend to scale better with cores than virtual threads,so if a game can use 8 threads on a 4C/8T CPU,in theory it should run better on an 8C/8T CPU using the same cores,which in the case of Intel is what you are seeing with their last few generations.

The only potential issue is if the 8C/8T CPU were to have much lower clockspeeds,and the games is still limited by one or two primary threads. However,apparently in the case of the latest Intel 8C/8T CPUs the clockspeeds seem to be reasonably high.

 

[kevman]

Games tend to scale better with cores than virtual threads,so if a game can use 8 threads on a 4C/8T CPU,in theory it should run better on an 8C/8T CPU using the same cores,which in the case of Intel is what you are seeing with their last few generations

Thanks! Makes sense

 

[AndreiD]

@pastymuncher The Cascade Lake line that comes out this Autumn has hardware mitigation for Spectre & Meltdown and they said a 'Core' CPU for H2 2018 also has it, could be the 9000 series, but it could have been something else that got delayed with their 10nm.
Remains to be seen, but either way the bios + software mitigation for spectre & meltdown vulnerabilities should take even less performance away from 8 core CPUs than it does from the 4 and 6 core ones.

@kevman Don't think that many games are developed with HT in mind, if at all, but if they can use up to 8 threads then of course the 8 core CPUs will have a performance boost. HT/SMT is just 2 threads contending for the same pipeline since a workload will rarely saturate it, reason why in raw performance numbers HT/SMT won't give you a 2x performance uplift, but closer to the 1.3x mark ideally.
At least until we get more details on Zen 2, these new 8 core CPUs should be really darn good for gaming. Initial leaks point to pretty high stock & turbo clocks indicating Intel further refined its 14nm and the big unified 16MB L3 cache should also bode well for gaming workloads.

 

[Killer7xx]

Intel's cowboy engineering practices will have big implications on their future CPU designs. To the normal consumer it's seemingly not a big deal but the server world are definitely not viewing the security scandals lightly. Server workloads have been affected way more than consumer workloads like gaming, the spectre and meltdown patches caused some drastic decreases in performance and the newly released L1 terminal flaw is another performance killer for servers. When it's recommended to disable hyperthreading for the best protection, you know it's bad.

They will most likely have to make some big redesigns in future architectures, with current architecture they may not be able to completely patch everything in the hardware or patch it without also incurring a loss in performance.

 

[AndreiD]

https://videocardz.com/77356/intel-core-i9-9900k-confirmed-to-be-soldered
Still a rumour, but if true it's looking pretty good.

 

[Killer7xx]

https://videocardz.com/77356/intel-core-i9-9900k-confirmed-to-be-soldered
Still a rumour, but if true it's looking pretty good.

The i9 9900k 8C/16T idea is still ridiculous to me. They're not doing themselves any favours by trying to milk consumers like that. A i7 9700k with 8C/16T and a 9600K with 8C/8T, and in the mid range 6C/12T 9500 and 6C/6T 9400. In the low end they can finally get rid of their dual cores and make everything at least quad core. That would still be quite competitive into the zen2 era.

 

[beany_bot]

The i9 9900k 8C/16T idea is still ridiculous to me.

All milking and ethics aside there is only one question needing answered.

Will it or will it not likely be the fastest gaming CPU in most cases?

Likely yes...Is that ridiculous?

 

[Dirk Diggler]

All milking and ethics aside there is only one question needing answered.

Will it or will it not likely be the fastest gaming CPU in most cases?

Likely yes...Is that ridiculous?

It might be, but would the margins justify buying into bug riddled hardware?