4.37 killed the AMRAAM
-
@EuKeule if you are running DDR4-3600 on a 7th gen Intel chip… with XMP enabled in BIOS or setting 3600 frequency manually – that is, essentially, a brutal “overclock” of the northbridge bus – the part of the CPU that controls communication to/from RAM.
I’m not sure what the fastest DDR4 ram existed when Intel 7th gen launched, but pretty sure it was much less than 3600. (I’m surprised the system would even boot and be stable, that way, so this may not be your problem – but hey it’s easy to check.)
Download your mobo manual and it should say what the various supported RAM speeds are/were for that socket generation… go with the highest one that doesn’t have an asterisk or “OC” warning In you BIOS settings… disable XMP and set the RAM speed manually – or let the default SPD negotiation happen.
I think I have the same high-end Samsung DDR4-3600 that you do… and I had stability problems on 9th gen Intel Core i7. I ended up turning it down to run at 3200 MHz iirc… with water-cooled cpu block.
[Edit: I’m not sure how to read the CPU-Z output … if the “1071.3 MHz” on the Memory tab is accurate, then you’re probably fine… but definitely doublecheck the BIOS settings]
-
What……
I‘ve no idea what you are talking about. This System is untouched since (I guess) 6 years.Unbelievable. I will check your recommendation and make the changes.
Thanks for that. I will let you know the result.
-
@airtex2019
According to Intel the i7 7700k supports up to DDR4-2400
https://ark.intel.com/content/www/us/en/ark/products/97129/intel-core-i77700k-processor-8m-cache-up-to-4-50-ghz.htmlUsing DDR4-3600 with this CPU does not really make sense. Let’s see what @EuKeule finds out after loading the default / optimized BIOS settings.
-
@Seifer said in 4.37 killed the AMRAAM:
@Razor161 interesting. Does it happen only for him? Sp or MP? What is his hardware?
what i know is that the guidance code is using quite precise calculation manipulating quite small numbers especially at the end of guidance…
A calculation error could explain
however i fail to understand why a RElease exe would work and not a LTCG
-
wow @airtex2019 is right… that is extreme overclock on the memory side. Please disable XMP and report back.
-
@EuKeule said in 4.37 killed the AMRAAM:
no idea what you are talking about. This System is untouched since (I guess) 6 years.
Rest well knowing you have very expensive, very fast ram. It was probably crazy, needlessly expensive, 6 years ago.
-
The drop from 3600MHz to 2400MHz might be quite substantial.
I think you will probably need to find the balancing point of least performance drop while avoiding AMRAAM shenanigans.
Maybe drop to 3200 first and do some tests, then to 3000 if you still have issues, then 2666 and lastly to 2400 if there are still issues at 2666.
I remember someone in the FL discord also had some issues with BMS and OC’d RAM, not about AMRAAMs though.
-
@unkindled no, please, let him check stock first. Then if it works, try the other way.
First optimization rule: make it work first, optimize later.
-
@EuKeule Might also be an idea to look into updating your BIOS to the latest version, as you’re seemingly running on the one it shipped with (version 0604, released in 2016), and there’s a lot of updates released since (newest being version 1301, released in 2018).
These updates include system performance/stability fixes and updates to DRAM compatibility:
https://rog.asus.com/motherboards/rog-maximus/rog-maximus-ix-hero-model/helpdesk_bios/I know some are skeptical of BIOS updates, but I’ve had no issue doing it over the last 15 years (most being Asus motherboards as well).
Just thought it was worth mentioning.
-
I manually set the DRAM frequency to 2133 MHz, the maximum for my prozessor. I also recognize, that the BIOS overclocked the prozessor from 4200 MHz to 4500 MHz. I stopped this.
I don’t want to be euphoric yet, but the first tests looking promising that the problem is solved.
So standby for my final conclusion after more testflights under different conditions.
So thanks to all who tried to help slowing the issue.
-
@EuKeule said in 4.37 killed the AMRAAM:
I manually set the DRAM frequency to 2133 MHz, the maximum for my prozessor. I also recognize, that the BIOS overclocked the prozessor from 4200 MHz to 4500 MHz. I stopped this.
I don’t want to be euphoric yet, but the first tests looking promising that the problem is solved.
So standby for my final conclusion after more testflights under different conditions.
So thanks to all who tried to help slowing the issue.
Well you issue was very typical , every of your missiles exhibited a snake pattern when closing to target
So if you don’t see this happening anymore that’s a good sign
-
@Mav-jp it’s a little concerning to me to learn that overclocking (ram or cpu) can make a computer “do math wrong” or differently… and not just bluescreen, or blackscreen / reboot etc
is the root cause of this well understood?
-
@airtex2019 said in 4.37 killed the AMRAAM:
@Mav-jp it’s a little concerning to me to learn that overclocking (ram or cpu) can make a computer “do math wrong” or differently… and not just bluescreen, or blackscreen / reboot etc
is the root cause of this well understood?
Missiles guidance requires a lot of accuracies because they are using very small numbers to correct course . When two objects are flying Mach 3.0 heads on , very very small angle corrections are needed to intercept.
Here what we witnessed is a snake pattern which was likely coming from bad calculations and missile over compensing all the time
I m not specialist , better ask seifer but as far as understand OC can alter math accuracies.
-
@Tomcatter31 The one that missed was due to a mechanical failure in the seeker head. I have a FOIA’d document that disscused that.
-
@Razor161 - a compendium of known issues that can be caused by hardware would be a good reference to include in the Documentation.
-
@Stevie said in 4.37 killed the AMRAAM:
@Razor161 - a compendium of known issues that can be caused by hardware would be a good reference to include in the Documentation.
I totally agree. Maybe @Micro_440th could consider this for future releases.
-
The thing that makes no sense with this model is the degree of both range and angular error. For example below:
It make 0 sense that there are angular errors this large, TWS uses the same exact waveform as RWS and SAM. If a radar had this level of angular inaccuracy it would literally be unusable as an FCR. Lets be clear engaging multiple targets with TWS was a thing demonstrated successfully in the 60’s. TWS can have issues with correlation algo’s but ranging and angle accuracy will not be one of these issues.
the above is from U2 also.
-
@nighthawk2174 said in 4.37 killed the AMRAAM:
It make 0 sense that there are angular errors this large, TWS uses the same exact waveform as RWS and SAM. If a radar had this level of angular inaccuracy it would literally be unusable as an FCR. Lets be clear engaging multiple targets with TWS was a thing demonstrated successfully in the 60’s. TWS can have issues with correlation algo’s but ranging and angle accuracy will not be one of these issues.
It’s more than just the waveform with TWS, it’s the computing power to develop and interpolate the tracks, radar power to actually get a usable return every scan to keep a quality track, and then time to go through scan volumes with the mechanical scan, with less time being better to keep accurate tracks.
TWS may have been around since the 60’s, it doesn’t mean it’s as accurate as the other modes. It’s always some guess work, see my previous posts about the AWG-9 vs APG-71. What made the APG-71 better was really the processing power than the radar dish and RIOs say that is when TWS really became a viable mode.
@Snake122 said in 4.37 killed the AMRAAM:
@WPNS24 See this thread I think people got their wires crossed between STT and RWS maybe for a minute and forgot to address the TWS.
@Stevie AWG-9 and APG-71 were different beasts than the F-16’s puny APG-68 and even then I believe I’ve heard a few F-14 RIOs that said that TWS was pretty amazing, but only on the APG-71, not as effective except for bombers on the AWG-9.
@Snake122 said in 4.37 killed the AMRAAM:
@Stevie see my edit above re: F-14. Yes scan volumes and and such are important, but there is also the processing power to develop and maintain the tracks, which was my impression why the AWG-9’s TWS was not the main bread and butter mode like it was for the F-14D.
Check out that thread, you did participate in it some, but especially @OPPOTATO’s reply I found interesting.
-
@Snake122 The angular resolution and accuracy in both range and angle of the radar will be no different between rws and tws. What tws does is it stores the last set of track data in a memory bank then when it goes over the same area it will correlate what it sees with the previous tracks using a correlation algorithm. Beyond this it is not doing anything special that RWS/SAM already don’t do. These radars the 68 and up will have absolutely no issues once so ever with the processing power needed for this task. TWS is set up so that it should go over the same area every 2 seconds or so, this is enough for accurate tracking especially for the case I showed where they were just flying level and straight. What issues a radar may have in its correlation algorithms handling maneuvering targets does not explain the massive angular and range errors being added to the missiles.
-
@nighthawk2174 said in 4.37 killed the AMRAAM:
@Snake122 The angular resolution and accuracy in both range and angle of the radar will be no different between rws and tws. What tws does is it stores the last set of track data in a memory bank then when it goes over the same area it will correlate what it sees with the previous tracks using a correlation algorithm. Beyond this it is not doing anything special that RWS/SAM already don’t do. These radars the 68 and up will have absolutely no issues once so ever with the processing power needed for this task. TWS is set up so that it should go over the same area every 2 seconds or so, this is enough for accurate tracking especially for the case I showed where they were just flying level and straight. What issues a radar may have in its correlation algorithms handling maneuvering targets does not explain the massive angular and range errors being added to the missiles.