Is Y-Cruncher Pass required to be considered stable?

[freeagent]

For me y-cruncher does 240w on the CPU, and 373 peak at the wall, R23 does 213w on the CPU and 317 solid at the wall

 

[Deleted member 185158]

2.5B because it eats up more ram. Only works on 32GB or above. 10B for 64GB. I found ycuncher to be the hardest by accident while trying to submit scores to HWBot when DDR5 first came out. Was wondering why people weren't submitting 2.5B scores, while the lower 1B and 25m was full of super high frequencies. Turns out, its much harder to get a passing result the higher you go.

So while, you could do everything gamers do on the daily. Chances are its not entirely stable. I've resolved this by changing the timings, changing the DRAM voltage and messing with CPU voltages. It could be all 3, just one or none because that MB or CPU just can't handle it.

But if its the ONLY program to fail, is it truly the best program to set the bar to, or is it kinda like Prime95 for CPUs these days?

2.5b uses 11gb of dram and I have no issues running only 16gb for this benchmark.

Voted NO, because a single benchmark is not a tell all of stability on any system.

 

[qubit]

Hi,
Hard to believe you would get a 1600w psu for the system on your spec's page

Oh yes, those are the current specs. And yeah, total overkill in terms of power, but not quality.

I wanted to upgrade my original Corsair HX850 v1 with something really high quality, had the software monitoring and control panel and zero fan mode, but the only PSU to fit all the criteria was this one. The quality of this PSU is second to none and gets superlative reviews everywhere, including TPU. I'd have bought an 850W version if it had been available, but as it is, it's impossible for me to ever stress it, which is important for long term reliability and stability.

I've also noticed something else too: better stability, like perfect. I'd replaced an intermittently faulty RAM stick that had been plaguing my PC for years, which then went good, but it still wasn't 100% glitch-free, just little things. But with this PSU, it's been rock solid like I've never seen before, so perhaps the power from the old PSU wasn't quite so perfect after all. I suspect that the PCIe connectors to the graphics card weren't making perfect contact, in particular, but no such issues now.

 

[RJARRRPCGP]

Unstable CO can also also cause Y-cruncher 2.5B to fail so its not just about the system memory it goes to overall system stability.

OCCT is also a good to run for overall system stability testing but that needs at least 1 hour and it detects WHEA errors which is something other tests don't do.

I would called Y-cruncher 2.5B a quick stability test and 1 hour OCCT a proper system stability.

If its not Y-cruncher 2.5B and 1 hour OCCT stable its unstable end of story. IMHO
View attachment 280639

I agree! I never in fact saw OCCT error out, IIRC. Usually Prime95 errors out or Windows errors out when running Prime95, IMX. Or PC just crashes hard! (freezes and mouse pointer frozen) (was always CPU core in my case and me getting too greedy with CPU OC!)

 

[ir_cow]

Voted NO, because a single benchmark is not a tell all of stability on any system.

Hmm. So you are okay with everything else passing but a single program? or just y-cruncher?

 

[freeagent]

Hmm. So you are okay with everything else passing but a single program? or just y-cruncher?

With a grudge, yes sometimes..

For instance, my 5900X is not core cycler stable at my settings, but they are everything else stable, including time.

My 5600X is core cycler stable, as well as everything else..

 

[Deleted member 185158]

Hmm. So you are okay with everything else passing but a single program? or just y-cruncher?

What I mean is cpu/mem only testing for a gaming rig does not tell if the system is actually stable.

The reason why is that gaming is a bit more intense, obviously depending on the game, where the cpu, gpu, I/O, SSD/HDD are all in use at the same time with various degrees of load.

I think y-cruncher is pretty heavy, but is it as sensitive (more or less) than running a program like PiFast, which tends to be more sensitive than PiMod for example.

So I do not believe in any single stress test to make any claims, be it a benchmark or a burn in.

Does any one even use Intel Burn Test any more? Should they/we?

 

[freeagent]

Does any one even use Intel Burn Test any more? Should they/we?

I use Linpack Xtreme, which is pretty much IBT, but you can use more ram, and its pretty heavy. Its good for testing heavy all core stuff. I don't know if its really needed now a days, I use it for all core stuff, and testing my cooling.. cooking paste.. you know.. that sorta thing

I usually incorporate it into my testing regimen for all of my systems. I stopped using IBT in 2016 I think and moved to Linpack Xtreme.

I'm all about the GFlops

 

[damric]

For RAM/IMC Y-cruncher is part of my stability tests, but it certainly doesn't end there.

It's also good for testing static clocks.

Good for observing thermal and boosting behavior, but obviously not good for testing the whole clock curve for stability.

Core Cycler is a step in the right direction, but it's nowhere near capable in its current iteration. A better version of that is needed to test the whole clock/load curves for these newer processors that dynamically change boost clocks based on load and temperature.

Overall I'd say Y-cruncher is good software and I am satisfied with its entertainment value considering what I paid for it.

 

[ir_cow]

I use Linpack Xtreme, which is pretty much IBT, but you can use more ram, and its pretty heavy. Its good for testing heavy all core stuff. I don't know if its really needed now a days, I use it for all core stuff, and testing my cooling.. cooking paste.. you know.. that sorta thing

I think I need to switch to Linpack because none of these mem programs can test 128GB lol. It just says "needs more threads".

 

[freeagent]

I think I need to switch to Linpack because none of these mem programs can test 128GB lol. It just says "needs more threads".

For testing mem, have you tried TM5? I like the Anta777Extreme and Absolut profiles.. Linpack Xtreme has a limit as to how much it can test, I don't recall off hand but it was under 20.. unless it has been updated or there is a way to do it..

 

[ir_cow]

For testing mem, have you tried TM5? I like the Anta777Extreme and Absolut profiles.. Linpack Xtreme has a limit as to how much it can test, I don't recall off hand but it was under 20.. unless it has been updated or there is a way to do it..

I have used Absolute profile before. Seems to catch the errors just fine. Both karhu stress and MemtestPro says not enough threads to test 128GB. I will try TM5 next and see if I get the same message.

update: TM5 works with 128GB

 

[The King]

I have used Absolute profile before. Seems to catch the errors just fine. Both karhu stress and MemtestPro says not enough threads to test 128GB. I will try TM5 next and see if I get the same message.

update: TM5 works with 128GB

I use the 1usmus_V3 profile vs Absolute profile for 25 cycles because there is a troubleshooting spreadsheet - TM5 Errors Deciphered.

tRFC mini DDR4/DDR5 WIP

TM5 Error Description Cheatsheet,1usmus_V3 TM5 Errors Decyphered,SOURCE 1usmus_V3,Error Type,Error Description ERROR #0,RefreshStable 0Mb,Voltage cutoff choke, suspect tRRD & tWTR Nearly always tRRD & tWTR but can also be too low tRP or tiny bit too low tRC (if user used > -3 on tRC) Start by ...

docs.google.com

Good luck running 25 cycles on 128GB. 32GB test took me just over 3 hours.

 

[ir_cow]

I just noticed the reason why it works is because the program isn't using all the RAM at once like Karhu stress and MemtestPro. It say 880x32. Memtest86 has the same problem in not testing all the ram at once. So is TM5 only good for 32GB?

 

[The King]

Try

I just noticed the reason why it works is because the program isn't using all the RAM at once like Karhu stress and MemtestPro. It say 880x32. Memtest86 has the same problem in not testing all the ram at once. So is TM5 only good for 32GB?

Try looking at the MT.CFG or profile *****.cfg file and see if it can be changed to run on 128GB.

 

[ir_cow]

Ah nice!

 

[Deleted member 185158]

Here's another memory test you guys can play with.

 

[Artem1us]

I use Y-Cruncher for two things,
first is for "cpu cores only oc" stability stesting and validation. Also CoreCycler for this job.

After i'm done with that, i move on to FCLK first, and RAM second, using for both TM5 w/ anta absolut .cfg , 3 cycles to see if there's a chance, then HCI Memtest at 3200%, and then back to TM5 12 cycles this time, and if all that checks out, 48hrs Prime95 Large FFT at least.

Then i return to Y-Cruncher for its second use, complete system stability testing/validation. I run it AT LEAST 48hrs. for that.

This program is a real powertool , in 'all' meanings.

And speaking of power, NOTHING has pulled moar juice in my 5700X than this. At least with a meaning (i consider Prime95 for "cpu cores only" and overall stability somewhat outdated).

I saw someone say that Y-Cruncher is "predictable". I beg to differ. It's not "only SSE" or "only AVX" , it's variable.

 

[ir_cow]

Here is a prime example of why I'm asking this question. Working on a memory review. Using the Z690 Tachyon I cannot pass 7800 (8000 not stable). Switch over to the Z790 Apex and 8400 boots, 8200 passes ycruncher at least (haven't done full testing). So in that example, the motherboard was the limiting factor and the system in question is not fully stable.

For a easier comparison, One MB will pass these memory test programs, play games just fine, but it fails ycruncher at DDR5-8000. Switch MBs and the issue goes away (in fact lower timings too). It is odd though, that it seems pretty stable besides that program.

ir_cow`s y-cruncher - Pi-2.5b score: 51sec 761ms with a Core i9 13900K

The Core i9 13900K @ 5800MHzscores getScoreFormatted in the y-cruncher - Pi-2.5b benchmark. ir_cowranks #null worldwide and #null in the hardware class. Find out more at HWBOT.

hwbot.org

 

[Deleted member 185158]

Here is a prime example of why I'm asking this question. Working on a memory review. Using the Z690 Tachyon I cannot pass 7800 (8000 not stable). Switch over to the Z790 Apex and 8400 boots, 8200 passes ycruncher at least (haven't done full testing). So in that example, the motherboard was the limiting factor and the system in question is not fully stable.

For a easier comparison, One MB will pass these memory test programs, play games just fine, but it fails ycruncher at DDR5-8000. Switch MBs and the issue goes away (in fact lower timings too). It is odd though, that it seems pretty stable besides that program.

ir_cow`s y-cruncher - Pi-2.5b score: 51sec 761ms with a Core i9 13900K

The Core i9 13900K @ 5800MHzscores getScoreFormatted in the y-cruncher - Pi-2.5b benchmark. ir_cowranks #null worldwide and #null in the hardware class. Find out more at HWBOT.

hwbot.org

For each board, you would try multiple bios's. Some times earlier bios version works better than new ones and visa versa.
Some boards are better for memory OC because how they are wired.
4 slots vs 2 slots.
And so forth.

There can never be a promise from a review, so "results will vary" should be a common comment in all your reviews.

 

[ir_cow]

There can never be a promise from a review, so "results will vary" should be a common comment in all your reviews.

Thats a given. But I guess isn't said enough in reviews. So when a memory kit on a QVL list doesn't work? Same deal, since technically a OC, everyone gets a pass?

 

[Deleted member 185158]

Thats a given. But I guess isn't said enough in reviews. So when a memory kit on a QVL list doesn't work? Same deal, since technically a OC, everyone gets a pass?

Just to many variables at play. Even with identical hardware, any two systems just are not alike.

For you, pick w/e kit may be handed out by a vendor and you do the thing.
All I'd ever expect is honest but open minded opinions about hardware and what the meaning over OC (over-clocking) really means.

First off, most manufacturers see XMP as an overclock and therefor isn't covered in promises or warranty.
Because of some reasons that would include how the bios is written. Unreasonable amounts of voltage to any device may happen just by enabling XMP/DOCP.
Thus Whea errors may occur. System stability may be at risk, and so forth. These are statements given by the motherboard manufacturer.
Which would also include cooling suggestions, such as using water cooling.

Over-clocking. Not always just setting XMP at X clock will be at a systems peak efficiency. I did demonstrate this with another member in the benchmark threads.
Which I may as well link for a read I guess, if that's cool with ya'll.....

Y-CRUNCHER - PI-2.5B Bench

www.techpowerup.com

Since Y-cruncher is pretty cpu heavy, it's not in the cpu frequency unless you're doing 1000mhz jumps on the top clock of your chip.
It's about using particular memory dividers. Get in a little BCLK on any chip you can. And of course tweaking the timings.

Basically, I look at XMP timings and frequencies as a base point for a learning overclocker. Obviously everyone wins on first post on Jedec standards. You don't NEED XMP, you just want it lol.

 

[ir_cow]

@ShrimpBrime you have good points.

 

[Deleted member 185158]

@ShrimpBrime you have good points.

Sometimes.

Doing reviews on unreleased hardware has gotta be tough. Especially if it's motherboards. How the heck are you supposed to know what in the future may hold for your motherboards with the ram kits until a bios is later released stating " Improved memory compatibility " or " Improved System Performance " . Wow. System performance is one that I've yet to wrap my head around. Sometimes it's worse I swear.

 

[Kodehawa]

No just because a single program has no way of measuring absolute stability. You'd need to test with way more stuff.

That aside, it *is* a good baseline to start with, but it could be unstable with more obscure loads, or just random stuff that might cause a chain reaction in other parts of your system. Also if your system has been stable for, say, a few weeks without issues, that would also count in my book.

I guess it depends on your definition of stable. I imagine if it doesn't pass the stress test (but works with everything else you throw at it), it might crash randomly, even if it's just once in a blue moon. Whether that is really important to you is a personal thing.