What's new

Intel slowdown

GreyFox7

Super Moderator
Staff member
This is definitely interesting -- that 2D graphics dropped so considerably. From everything I've been able to discern, that's not an expected result -- the issue that needs to be resolved is in the CPU and relates to accessing system memory. I'd say that if this were an SB2 it might make sense, since the integrated GPU uses system RAM while the dedicated GPU has its own memory. But on an SP3 both 2D and 3D graphics are using system memory -- you'd expect to see the same impact.

Until I see it confirmed elsewhere, I'm going to call this an issue with the graphics driver and something that will be fixed.
Anyone else that want to run it could help shed some light on it.

IDK perhaps there's some kernel switching in that code.:confused:
 

wynand32

Well-Known Member
Anyone else that want to run it could help shed some light on it.

IDK perhaps there's some kernel switching in that code.:confused:

I haven't received the firmware update on any of my machines yet. Once I do, I'll run the tests and report back.
 
OP
G

gman713

Member
I just did the Geekbench test and got a 4553 single core and 12516 multi-core score. I don't know if it matters that my power mode was set to better performance and not best. Hopefully I won't drop that much after all the patches are done.
 

jtrox02

New Member
Well I am or was on the Insiders Release Preview (but they changed the name and I'm not sure what I'm on now) but I got the OS patch a lot later than people not on the Insiders. I saw a post here that said it was out so I just did Windows Update to check and sure enough I got it.

Got it thanks!...I saw that it was already released right after I wrote that message. So I downloaded passmark and ran a test before running Windows update. I too had a significant reduction in results after the update, approx. 15% reduction in Passmark rating. I have an SP4 i7 8gb 256gb. The 3d test failed because of some message about resolution but I wasn't too concerned as I don't play games on it.

Before:
https://xx8evq-dm2305.files.1drv.co...O25Z_5QASA?width=660&height=481&cropmode=none

After:
https://xxp5yq-dm2305.files.1drv.co...eXKvUb-tyg?width=650&height=660&cropmode=none
 

Seneleron

Active Member
I looked at both results. I'm not an engineer, so I don't claim to completely understand all the ins and outs of Meltdown, but I understand enough to know they've removed "speculative execution". As such, it seems to me that the CPU should take longer to load/offload to compensate for the work of security checking every operation that the Kernel processes. Wouldn't this cause a drop in CPU performance numbers based on this kind of workload?

What I saw between the two comparisons is that memory and 2d video performance dropped significantly in the second test. I guess I can understand if the added workload is overtaxing the memory [though I have a hard time believing we suddenly need more than 8 gig for basic operations] but I'm really confused about the 2d video drop. Obviously there's integration between CPU instructions and GPU functionality, but is it really that much to cause a 220ish point drop? I get a 120ish point drop CPU in synthetic benchmarks is more than margin of error but it seems the system overall is taking a bigger hit then I'd anticipated.

Of course I have to balance that against it being a 2 core 15w ULV cpu, but I'm surprised at the overall numbers in general. Does day to day operation feel any different? Obviously synthetic benchmarks are designed to completely max systems, but I'm very curious as to the day to day impact on latency and workflow.
 

GreyFox7

Super Moderator
Staff member
I looked at both results. I'm not an engineer, so I don't claim to completely understand all the ins and outs of Meltdown, but I understand enough to know they've removed "speculative execution". As such, it seems to me that the CPU should take longer to load/offload to compensate for the work of security checking every operation that the Kernel processes. Wouldn't this cause a drop in CPU performance numbers based on this kind of workload?

What I saw between the two comparisons is that memory and 2d video performance dropped significantly in the second test. I guess I can understand if the added workload is overtaxing the memory [though I have a hard time believing we suddenly need more than 8 gig for basic operations] but I'm really confused about the 2d video drop. Obviously there's integration between CPU instructions and GPU functionality, but is it really that much to cause a 220ish point drop? I get a 120ish point drop CPU in synthetic benchmarks is more than margin of error but it seems the system overall is taking a bigger hit then I'd anticipated.

Of course I have to balance that against it being a 2 core 15w ULV cpu, but I'm surprised at the overall numbers in general. Does day to day operation feel any different? Obviously synthetic benchmarks are designed to completely max systems, but I'm very curious as to the day to day impact on latency and workflow.
On the memory, since they are messing with the Memory address tables I can see how this would impact memory access. It's possible there's something unique about they way they do 2D graphics testing that's causing the low score. AFAIK there are not a lot of 2D graphics tests available. iid say it seems like startup and app launching is slower so far and waking from sleep is slower, login is slower. Touch keyboard might be slower but it's a mess in FCU anyway.
 
Last edited:

wynand32

Well-Known Member
On the memory, since they are messing with the Memory address tables I can see how this would impact memory access. It's possible there's something unique about they way they do 2D graphics testing that's causing the low score. AFAIK there are not a lot of 2D graphics tests available. iid say it seems like startup and app launching is slower so far and waking from sleep is slower, login is slower. Touch keyboard might be slower but it's a mess in FCU anyway.

What I don't understand, generally speaking, is why this would impact 2D and not 3D GPU operations. Certainly, 3D stresses memory at least as much as 2D, and I would think more so.
 
OP
G

gman713

Member
I just did the Geekbench test and got a 4553 single core and 12516 multi-core score. I don't know if it matters that my power mode was set to better performance and not best. Hopefully I won't drop that much after all the patches are done.

After my system updated earlier today, I re-ran Geekbench and got a 4% drop in performance for single core and a 2% drop in multi-core. This is for the Surface Book 2 15".
 

jtrox02

New Member
I looked at both results. I'm not an engineer, so I don't claim to completely understand all the ins and outs of Meltdown, but I understand enough to know they've removed "speculative execution". As such, it seems to me that the CPU should take longer to load/offload to compensate for the work of security checking every operation that the Kernel processes. Wouldn't this cause a drop in CPU performance numbers based on this kind of workload?

What I saw between the two comparisons is that memory and 2d video performance dropped significantly in the second test. I guess I can understand if the added workload is overtaxing the memory [though I have a hard time believing we suddenly need more than 8 gig for basic operations] but I'm really confused about the 2d video drop. Obviously there's integration between CPU instructions and GPU functionality, but is it really that much to cause a 220ish point drop? I get a 120ish point drop CPU in synthetic benchmarks is more than margin of error but it seems the system overall is taking a bigger hit then I'd anticipated.

Of course I have to balance that against it being a 2 core 15w ULV cpu, but I'm surprised at the overall numbers in general. Does day to day operation feel any different? Obviously synthetic benchmarks are designed to completely max systems, but I'm very curious as to the day to day impact on latency and workflow.

I am experiencing the same as GreyFox7. Noticing a bit of a general slowdown on more intensive actions like startup and app launching etc.
 

wynand32

Well-Known Member
Well, whatever the impact, this is one where it's not going to be specific to the Surface line. Every machine, if properly patched, will experience the same, and so it comes down to how secure you want your systems to be. I have an older HP that my wife uses that I'm now thinking should be retired. Although I'm not 100% sure about that -- it's unclear if Intel will be pushing its own patches to all systems that will resolve the issues, if that's what Microsoft is pushing in its patches, or exactly what's going on.
 

Bonj87

New Member
I have a SB1.. I ran the intel checker.. it says im vulnerable.. I tried to DL the patches manually but each patch says its not applicable to my system.. I uninstalled every AV except the windows defender. .I checked my registry and the key is properly set.. dunno why I cant get the patch
 

ncameron

Member
I have a SB1.. I ran the intel checker.. it says im vulnerable.. I tried to DL the patches manually but each patch says its not applicable to my system.. I uninstalled every AV except the windows defender. .I checked my registry and the key is properly set.. dunno why I cant get the patch

I also have an SB1, and I also ran the Intel tool, and it told me that my system is NOT vulnerable. Weird. I bought mine in November 2016, how old is yours?

Neil
 
Top