I'm the author of RAM Test and founder of Karhu Software. I'm a software engineer by trade and by no means an expert in hardware, but given how effective a common-sense approach to stability testing turned out, I think I'm qualified enough to say a few words on the subject. While I agree there's a lot of stuff going on in the background when running a memory test on top of an OS, I'd still argue that all errors detected by RAM Test ultimately boil down to hardware instability rather than software going bonkers on its own. RAM Test is a user space program and has limited knowledge about the underlying hardware. It currently only gives a yes/no answer to whether errors were found. Mapping a virtual memory address to a physical one would require a privileged context on Windows and even then error could have happened elsewhere. The test algorithm itself is very low level and the only software components that could directly interfere with it are the virtual memory and multiprocessing implementations of the OS. But if those were buggy, you'd see program crashes and kernel panics. Not just the error counter ticking up. RAM Test doesn't directly interact with other programs, but of course shares the CPU caches and physical RAM. If other programs indirectly flip bits in the pages allocated to RAM Test, maybe the sticks are vulnerable to row hammer and the other program just happens to be mapped to adjacent DRAM rows? The source of the error might not necessarily be in the DRAM cells at all, but anywhere between CPU registers and physical memory that's just too low level for software to differentiate. There's way too many variables involved to pinpoint the exact source and nature of the error with 100 % accuracy. Anyway, it's true that even if RAM Test detects errors, the system can be stable enough to play games. As a matter of fact, the machine I'm using to develop RAM Test has an unstable overclock most of the time and has never crashed. Just a couple corrupted files, random stuttering and occasional audio artifacts. There's a lot of improvements coming to RAM Test, but development takes time and I don't think I'm getting to quit my day job anytime soon and work on this full-time. Especially if the case is that big memory manufacturers are using the program as a part of their test battery with a non-commercial license.
Hi Shannon from Patriot here, and HARD AGREE!!! In most scenarios your Soiftware picking up errors is absolutely spot on, and one of the reasons I use it as well. I believe there is room for both on this subject. I have seen in many scenarios where when edge overclocking I have the OS perform process calls that caused a false alert on multiple test softwares. There is always chances for the memory or any other hardware to cause actual instability, but in many scenarios it even could be a combination of hardware and software stack that can also cause these sorts of issues. BTW I am not sure how but I think we should connect so we can see how we can better message how your software tests and how to define actual stability, at least for memory.... Even if there are so many variables in the hardware stack alone, including FW which constantly changes with the life of the platform from microcode and feature tweaks and updates. If you are interested, you can find me rather easily, or my first dot last name at Patriotmemory dot com.
Your app is pretty dope. I've been using it for a few years now for system stability and memory OCing. Money well spent. And you're right, most of the time I'd run into errors deep into testing, it was often due to my CPU OC, particularly voltages. Or it was DRAM/SOC Voltage related.
Nice video and good on Patriot for talking so long to you. It will pay dividends for them. I didn't consider their brand previously but I will in the future!
This is nice, our boy Jufes is asking the tough questions.. loved these segments. It's unfortunate that AMD didn't respond to you. That would've been a good discussion point.
Shannon is a great match for Jufus as they talk the same language in terms of testing and FAFO. Please do more content with him if you're able in the future!
Shannon was great and was refreshing to watch. Most important he was Honest. Made me respect Patriot a lot more now knowing people like this work for the company. You should definitely have him back on the channel as i would like to hear more from him. I can honestly say Jufes this interview and the type of questions and topic you guys talked about is exactly what i and many want to here. Such interview would have NEVER been done from any of the other "larger" channels that cover tech. I don't want to sound like i'm fangirling but we are thankful and appreciate your content more then you could ever even imagine Jufes. Sorry for the long comment but i have to say it. It seems most if not all the other tech channels now and prior for so many years is just repeating all the Marketing talk points BS as if they themselves are trying to sell you the product in these interviews. Its Bonkers!!! Watch their interviews and then watching yours. I watched your Nvidia booth video and that guy dint know how to even process your type of questions and kept trying to fall back on the buzz words and main talking points because for so many years he and a lot of these large companies frankly are so use to these larger tech content creators just reading off the marketing script rather then asking real questions that the consumer want to know. Its actually gross that we the consumers have been dealing with that crap from both the company and the content creator. I'm so over it... Thank you for actually asking real question and bring us valuable information because no one does it like you brother. Ii'm so sick and tired of watching those larger tech channels and feeling like i'm watching a paid ad or coming away from it feeling i dint learn a dang thing from their video about the products because all they did was read off the Marketing script. The normal people are so ready for your content Jufes. It couldn't have come any sooner. Cant wait to see all your coverage of CES 2024.
I appreciate that, just be sure to get something that best matches your build... I know going super fast seems like "the best" but if not matched with your setup, will give you a pretty bad experience.
I'm running 7200 CL34 with my strix z790-e/13900k. Only had to turn on xmp and it runs great in gaming and video editing. IMO it's the spot you want to be in.
Up your trefi to 65536 or even 131072, you will get ton of performance on 7200mhz with that :) It is easiest tweak for you, if you want more then tweak tRDRD_sg (read speed) and TWRWR_sg (write/copy)
Quite like this dude, he comes over as a no nonsense enthusiast. It's actually refreshing to hear a company representative, just talk straight like a normal person.
Ive finally got 8200 running i was stuck on 7800 until today 😂 Im not at 8400 so I don't need to seek help😂😂😂 Great video and I could listen to him all day!
Thanks Jufes what an awesome video and great topic to listen to! Thank you for going over to CES and Shannon Robb if you see this thank You are awesome!
Oh I'm sure there's plenty of folks on various forums and whatnot who are upset by what I said, but that's ok. Disagreement is cool, and I'm always down to change my view if they can prove me wrong :)
for stability the quickest judge is how well it does tRFC and tREFI at heat, because they rely on the capacitance of the cells themselves it gives a quick indicator between the same type of IC ie if one kit of b-die cant do 32768 tREFI but another does 65535 on the same platform then id pick the latter kit, it will likely have better temperature tolerance while having better voltage scaling capabilities much the same for hynix ddr4 ics, havent tried micron yet, havent tried ddr5 yet.
I've just started watching this video only 10 percent in. But I wanted to type this before going further. Jufus be honest you went in and you definitely had in the back of your mind.. Ooh I'm gonna roast this chicken hard.... 😂 Finished watching. informative. What a good guy. Hope we can get more insight on chip level things later when you guys get him on stream
Nice to see videos like that from CES for once. I like the segway with the errors and what could have triggered an error. If you are gaming stable and not experiencing any issues while playing, who cares. He his right on target when he compared with a car. We should have a guy like that for every vendors at CES.
I mean, we can do it, I know most cannot.. But with some skill you can even do it with a PCB 0re-heater and rework station, but I would advise leaving it to someone with a proper setup for reballing. Don't need to bridge the solder balls below the ICs
Been following your content since almost day one amazing how far youve come man. Now you;re at CES great interesting interviews, not cases and fans sick ces content bro wish i could catch more live content sucks cant watch them after streamed afaik#besthardwarereviews ty all supporters
My Z690-F Strix couldn't run my 48gb 7200mhz kit until i hit the lastest bios update. I haven't tried to push further but it was nice i was able to just XMP set and forget it.
Very interesting video. I live in Moldova, here we have officially only Kingston and Patriot, just few models of Goodram, Transcend and Hynix/Samsung Original. Patriot have much much lower prices then Kingston. So I got Viper Venom 2x32Gb DDR5-6400 CL32 sticks, overclocked them to 6800 CL32 and after few weeks one stick just dissapeared, it died :) No problem, I have warranty. So now I wait for PATRIOT VIPER XTREME 5 RGB DDR5 RAM 48GB (2X24GB) 7600MT/s, we dont have such model, I ordered one from official supplier... Price is around 340$, yes Moldova is the TOP country, so we have nice prices over here xD I hope it will run 7600 on Z790 Hero, if not, then at least 7200 :)
5:45 yes, I have those dog water chips, Patriot Viper 4 Series. I blindly bought too much of that branded RAM several years ago for TR system and later regretted it, they used Spectek dies and were hardly overclockable.
I have a lot of respect for this guy and his approach, however I still feel that it's doing a disservice to consumers to not have that trust in advertising. Something needs to be done (apparently industry wide) to address that, despite the current reality of the situation.
I can agree and disagree on this. I think there needs to be a level of responsibility and honesty in advertising in general. I mean like I said, Xtreme 5 is awesome, but if for a gaming build, we have Viper Venom's and Elite 5... Xtreme 5 is made for those who are benching or general overclocking. However if you do get Xtreme 5, you can always turn down the speed and crank down the timings, so not end of the world either way.
Alright, so could you do a post or video on the Warzone FPS difference between Apex 8000 and for example a Z790i Edge 7200 XMP with all the same parts and same 7200 Teamgroup DDR5 kit?
Representative of a memory manufacturer when asked about the hit/miss stability of XMP profiles suggests that stability is "in the eye of the beholder" and that if Karhu has 3 errors in 24 hours you'll likely not hit a problem if you run a different stress test in Linux. Would be surprised if they didn't try to draw focus away from stress tests showing problems with their product rather than tackle the question of how to set an industry wide stability test which can be run regardless of variables.
TUF Z790 PRO here. 2x24GB @ 7200 CL34 1.43v on a 14900K. runs ycruncher VT3 for hours but i had to add SA and IVR transmitter and a little PLL voltages.
Keep holding their feet to the fire. Apart from that, I'm going to have to disagree with most of the comments and say Shannon was not a good interviewee. He's a marketer with a company and that company has more control over him than even the channels/websites they advertise with. And we know how untrustworthy those are. That's his job, fine, but we'll never get unfiltered opinions from him as long as he works there. Also that failed flex when Shannon was all "I machined my own pot for Slot A, did you even have that?" and Jufes responded "like a Pentium 2?" and Shannon responded, "no, the Celeron 300A". Pentium 2, along with the infamous 300A, were BOTH slot CPUs. The SAME slot. Slot 1, not Slot A, to be clear. Gamers, as Jufes has said a million times, are not the same thing as PC enthusiasts. Most of them want the fastest without FAFO. This mindset Shannon has that "if you want to game with the balls out fastest, you shouldn't expect it to be Karhu stable for 10 days" is absolutely wrong. As long as you buy reasonable parts: if you have a 13900K, you are going to get the advertised 5.5GHz all core and if you buy a 4090, you are going to get the advertised 2.2GHz all core and they will just work. But if you buy DDR5 8000 and, as Shannon disclosed, less than 20% of their 8000 kits are stable on Apex with 13th gen, it isn't an 8000 kit. Give us QVL lists that aren't pure lies which is what this channel has been complaining about for years. Even sharing the stats he said he has on an Excel spreadsheet would go a long way. Instead we get hand waving. All the stuff about silicon lottery, drivers, etc. is true but at the end of the day, over 80% of DDR8000 isn't. As for questioning the validity of stress testing and the nonstop excuses "But what about Windows scheduler" "you need to use an oscilloscope" "the OS parked a core" "a draw call happened" "RGB" "polling" etc. doesn't take away from the fact that Karhu will consistently error at consistent times with a given set of RAM timings and not others. That is a memory issue plain as day.
I’m on 14900K z690 dark and Mdie 6200 Corsair 32GB kit and it’s so hard to get 6400+ MHz stable or even tight timings. DDR5 is truly tricky to stabilize…. It’s really overwhelming…. I’ve tried 13900K also. Just really lost on the timings and voltages sweet spots, but the way people say giving it 10mv back and forth is really time consuming of days. There has to be a better guide… and everyone is stroking their A DIE score, it’s like MDIE talks are barely there or it’s AMD drones
I’ll run stability tests for days, but crash with optimization shaders in Battlefield and crash…. So I know it’s not stable… really could use guidance on someone on discord
Thanks to all the real memory makers for the tips, using refresh interval 2000 made input latency nothing even with path crap and deep learning super ship, ray repoopulificaftrity
First time I'm hearing that Linkpack can burn VRM's and gates... it makes sense as I used it as a measure of stability and the CPU seem to be getting worse....
I'm the author of RAM Test and founder of Karhu Software. I'm a software engineer by trade and by no means an expert in hardware, but given how effective a common-sense approach to stability testing turned out, I think I'm qualified enough to say a few words on the subject.
While I agree there's a lot of stuff going on in the background when running a memory test on top of an OS, I'd still argue that all errors detected by RAM Test ultimately boil down to hardware instability rather than software going bonkers on its own.
RAM Test is a user space program and has limited knowledge about the underlying hardware. It currently only gives a yes/no answer to whether errors were found. Mapping a virtual memory address to a physical one would require a privileged context on Windows and even then error could have happened elsewhere.
The test algorithm itself is very low level and the only software components that could directly interfere with it are the virtual memory and multiprocessing implementations of the OS. But if those were buggy, you'd see program crashes and kernel panics. Not just the error counter ticking up.
RAM Test doesn't directly interact with other programs, but of course shares the CPU caches and physical RAM. If other programs indirectly flip bits in the pages allocated to RAM Test, maybe the sticks are vulnerable to row hammer and the other program just happens to be mapped to adjacent DRAM rows?
The source of the error might not necessarily be in the DRAM cells at all, but anywhere between CPU registers and physical memory that's just too low level for software to differentiate. There's way too many variables involved to pinpoint the exact source and nature of the error with 100 % accuracy.
Anyway, it's true that even if RAM Test detects errors, the system can be stable enough to play games. As a matter of fact, the machine I'm using to develop RAM Test has an unstable overclock most of the time and has never crashed. Just a couple corrupted files, random stuttering and occasional audio artifacts.
There's a lot of improvements coming to RAM Test, but development takes time and I don't think I'm getting to quit my day job anytime soon and work on this full-time. Especially if the case is that big memory manufacturers are using the program as a part of their test battery with a non-commercial license.
This comment needs more attention.
Hi Shannon from Patriot here, and HARD AGREE!!!
In most scenarios your Soiftware picking up errors is absolutely spot on, and one of the reasons I use it as well.
I believe there is room for both on this subject.
I have seen in many scenarios where when edge overclocking I have the OS perform process calls that caused a false alert on multiple test softwares.
There is always chances for the memory or any other hardware to cause actual instability, but in many scenarios it even could be a combination of hardware and software stack that can also cause these sorts of issues.
BTW I am not sure how but I think we should connect so we can see how we can better message how your software tests and how to define actual stability, at least for memory.... Even if there are so many variables in the hardware stack alone, including FW which constantly changes with the life of the platform from microcode and feature tweaks and updates.
If you are interested, you can find me rather easily, or my first dot last name at Patriotmemory dot com.
Are you on twitter?
I refuse to use PayPal can you add extra options to pay for your software 🇬🇧
Your app is pretty dope. I've been using it for a few years now for system stability and memory OCing. Money well spent. And you're right, most of the time I'd run into errors deep into testing, it was often due to my CPU OC, particularly voltages. Or it was DRAM/SOC Voltage related.
Been following you since you first posted that 3080 fix for warzone 1 and now you’re at CES. Happy for you man what a journey
Real
What a Gent. A 40 minute chat just for us. Respect to Shannon and Patriot for their time.
Please have Shannon back on. I love the way this guy puts things into perspective. I bet he has great takes on multiple things regarding PC Tech.
And you can tell he’s a real enthusiast not a sales guy.
HARD Agree
Yeah he really knows his stuff and it shows
@@CodeStonksfn he used to be at EVGA. Yup he is an enthusiast and not just a sales guy.
Shannons a legend Patriot lucky to have a passionate person like that
Nice video and good on Patriot for talking so long to you. It will pay dividends for them. I didn't consider their brand previously but I will in the future!
Haha thanks to jufes Patriot has sold a lot of b die
This is nice, our boy Jufes is asking the tough questions.. loved these segments.
It's unfortunate that AMD didn't respond to you. That would've been a good discussion point.
Damn wanted to see that
Shannon is a great match for Jufus as they talk the same language in terms of testing and FAFO. Please do more content with him if you're able in the future!
This guy needs more air time either with you or other channels, he has a voice that people will want to hear more!
we finally found someone who loves memory just as much as jufes
You're not wrong, that's why I make it 😅
Shannon was great and was refreshing to watch. Most important he was Honest. Made me respect Patriot a lot more now knowing people like this work for the company. You should definitely have him back on the channel as i would like to hear more from him. I can honestly say Jufes this interview and the type of questions and topic you guys talked about is exactly what i and many want to here. Such interview would have NEVER been done from any of the other "larger" channels that cover tech. I don't want to sound like i'm fangirling but we are thankful and appreciate your content more then you could ever even imagine Jufes.
Sorry for the long comment but i have to say it. It seems most if not all the other tech channels now and prior for so many years is just repeating all the Marketing talk points BS as if they themselves are trying to sell you the product in these interviews. Its Bonkers!!! Watch their interviews and then watching yours. I watched your Nvidia booth video and that guy dint know how to even process your type of questions and kept trying to fall back on the buzz words and main talking points because for so many years he and a lot of these large companies frankly are so use to these larger tech content creators just reading off the marketing script rather then asking real questions that the consumer want to know. Its actually gross that we the consumers have been dealing with that crap from both the company and the content creator. I'm so over it...
Thank you for actually asking real question and bring us valuable information because no one does it like you brother. Ii'm so sick and tired of watching those larger tech channels and feeling like i'm watching a paid ad or coming away from it feeling i dint learn a dang thing from their video about the products because all they did was read off the Marketing script.
The normal people are so ready for your content Jufes. It couldn't have come any sooner. Cant wait to see all your coverage of CES 2024.
Great comment! Agreed.
Had no idea what this guy said 90% of the time but I'm getting Patriot ram for my next build
😂
I appreciate that, just be sure to get something that best matches your build... I know going super fast seems like "the best" but if not matched with your setup, will give you a pretty bad experience.
Where your haters at... They need to watch this!
This is why I sub to you. You know your shit.
Love see more live! I learned lot out of this.
they realized that their hate means absolutely nothing to anyone
I'm running 7200 CL34 with my strix z790-e/13900k. Only had to turn on xmp and it runs great in gaming and video editing. IMO it's the spot you want to be in.
Up your trefi to 65536 or even 131072, you will get ton of performance on 7200mhz with that :) It is easiest tweak for you, if you want more then tweak tRDRD_sg (read speed) and TWRWR_sg (write/copy)
Quite like this dude, he comes over as a no nonsense enthusiast. It's actually refreshing to hear a company representative, just talk straight like a normal person.
These discussion series are really good stuff :-) Thanks for bringing them to us.
Jufus this is excellent thanks.
It's been a pleasure to watch your channel grow!
Yes definitely have him on a stream! Amazing informational video!
Shannon is so passionate and chill. What a great guy.
Incredible interview, what a stand up and passionate guy. Clearly see that he loves what he does.
Best interview yet. This was a great discussion; I love patriot memory.
Golden content, Shannon is a real one for sure
This guy is very cool. So happy I found your channel. keeping it real. love it.
this was an amazing talk.best interview at CES ive seen. good information. still watching, only 15 minutes in.
Ive finally got 8200 running i was stuck on 7800 until today 😂 Im not at 8400 so I don't need to seek help😂😂😂 Great video and I could listen to him all day!
If you're trying to run it 24/7 for your main rig, please still seek help 😂
Shannon "the Patriot guy"
😂😂😂😂
This was great!! Thanks, Jufes!
Yes, we want Shannon on a stream.
Great conversation! Would love to see more content with you two.
What a breath of fresh air. :) Best 2024 CES Interview! Bravo! Well Done!
P.S- My Cow Cat liked it too! :)
What is a cow cat?🤔
This guy was great I'd love to hear more from him!
I would love to see a collab between you and them
Be good to have him on the live stream at some point
I'm down to join, no issues there as long as schedules line up
Thanks Jufes what an awesome video and great topic to listen to! Thank you for going over to CES and Shannon Robb if you see this thank You are awesome!
This is a good interview. Touches on multiple topics that are always contentious over on the OCN forum.
Oh I'm sure there's plenty of folks on various forums and whatnot who are upset by what I said, but that's ok. Disagreement is cool, and I'm always down to change my view if they can prove me wrong :)
for stability the quickest judge is how well it does tRFC and tREFI at heat, because they rely on the capacitance of the cells themselves it gives a quick indicator between the same type of IC
ie if one kit of b-die cant do 32768 tREFI but another does 65535 on the same platform then id pick the latter kit, it will likely have better temperature tolerance while having better voltage scaling capabilities
much the same for hynix ddr4 ics, havent tried micron yet, havent tried ddr5 yet.
Great video, DDR4 Patriots b-die best modules
Appreciaet you getting in there and having a focused conversation this topic. Keep crushing it.
Good conversation, good info.
I've just started watching this video only 10 percent in. But I wanted to type this before going further. Jufus be honest you went in and you definitely had in the back of your mind.. Ooh I'm gonna roast this chicken hard.... 😂
Finished watching. informative. What a good guy. Hope we can get more insight on chip level things later when you guys get him on stream
Nice to see videos like that from CES for once. I like the segway with the errors and what could have triggered an error. If you are gaming stable and not experiencing any issues while playing, who cares. He his right on target when he compared with a car. We should have a guy like that for every vendors at CES.
So good interview, great Jufes. 😊
Banger video. Definitely have him back for more talks.
like this bro that is what separates you from the rest....thank you
Shannon is legend, I want to hear so much more frrom him. This was by far one of the best interviews I have seen from CES.
Knowledgable people talkin shop is so good. This is the stuff I love watching! Just reball and re-chip a dimm, no probs!
I mean, we can do it, I know most cannot..
But with some skill you can even do it with a PCB 0re-heater and rework station, but I would advise leaving it to someone with a proper setup for reballing.
Don't need to bridge the solder balls below the ICs
Awesome video! Been watching you a couple years. You really cut through the bs and straight to facts. Keep up the great content!
Been following your content since almost day one amazing how far youve come man. Now you;re at CES great interesting interviews, not cases and fans sick ces content bro wish i could catch more live content sucks cant watch them after streamed afaik#besthardwarereviews ty all supporters
haha celeron 300A and Abit BH mobo!! Shannon goes back for sure.
My Z690-F Strix couldn't run my 48gb 7200mhz kit until i hit the lastest bios update. I haven't tried to push further but it was nice i was able to just XMP set and forget it.
For sure live stream w him. This was the best interview so far for me. Very interesting, you should have asked him what platform he prefers.
I think we need a custom windows episode. I'd run anything that you're running😊
Hell yeah, get him on Stream!
Anatoly Frame Chasers looks jacked when not behind a desk.
nice interview. very reasonable guy. manufacturers must have a good grip on reality and optimal vs maximal use case scenarios.
We try, but I try to push our Patriot team over the deep end at times with the XOC stuff, but we manage pretty well in spite of my best efforts 😂
this was a great vid and chat between you guys. love the content bro
They had to know him before this moment because of all the DDR4 kits he’s moved
Very interesting video. I live in Moldova, here we have officially only Kingston and Patriot, just few models of Goodram, Transcend and Hynix/Samsung Original. Patriot have much much lower prices then Kingston. So I got Viper Venom 2x32Gb DDR5-6400 CL32 sticks, overclocked them to 6800 CL32 and after few weeks one stick just dissapeared, it died :) No problem, I have warranty. So now I wait for PATRIOT VIPER XTREME 5 RGB DDR5 RAM 48GB (2X24GB) 7600MT/s, we dont have such model, I ordered one from official supplier... Price is around 340$, yes Moldova is the TOP country, so we have nice prices over here xD I hope it will run 7600 on Z790 Hero, if not, then at least 7200 :)
I guess your "dead" ram stick simply doesn't have RSWP, probably it's cause of your problem :) or PMIC...
Good talk seems like the guy has a lot of experience and knowledge and passion
5:45 yes, I have those dog water chips, Patriot Viper 4 Series. I blindly bought too much of that branded RAM several years ago for TR system and later regretted it, they used Spectek dies and were hardly overclockable.
I have a lot of respect for this guy and his approach, however I still feel that it's doing a disservice to consumers to not have that trust in advertising. Something needs to be done (apparently industry wide) to address that, despite the current reality of the situation.
I can agree and disagree on this. I think there needs to be a level of responsibility and honesty in advertising in general.
I mean like I said, Xtreme 5 is awesome, but if for a gaming build, we have Viper Venom's and Elite 5...
Xtreme 5 is made for those who are benching or general overclocking.
However if you do get Xtreme 5, you can always turn down the speed and crank down the timings, so not end of the world either way.
Great video. Really enjoyed the conversation and information.
Man great job collaborating. I love how you grow and expand and evolve. You are literally overclocking yourself 👌
Great interview. Well done.
Loved this discussion. More content like this!
my next sticks are from patriot
guy knows his shit
Alright, so could you do a post or video on the Warzone FPS difference between Apex 8000 and for example a Z790i Edge 7200 XMP with all the same parts and same 7200 Teamgroup DDR5 kit?
Shannon Robb seems like one of the good ones! great watch enjoyed that one. roll on JJ from Asus...
Representative of a memory manufacturer when asked about the hit/miss stability of XMP profiles suggests that stability is "in the eye of the beholder" and that if Karhu has 3 errors in 24 hours you'll likely not hit a problem if you run a different stress test in Linux. Would be surprised if they didn't try to draw focus away from stress tests showing problems with their product rather than tackle the question of how to set an industry wide stability test which can be run regardless of variables.
Always love Patriots RAM, shannons always a pleasure to watch explain RAM stuff. good video..
Give credit when is due. Good interview and very interesting stuff.
Still have 2 kits of Patriot Memory Viper II Sector 5 DDR3 2400c9 @ 1.65v Powerchip (PSC-X) IC's New Old Stock in fully sealed packaging
Good stuff, would like to see some Q&A live with Shannon seems a very reasonable guy
Memory stability haunts my dreams
That was awesome!
Well that was enlightening! Loved this!
Nerds! ;)
finally more people are speaking the truth of how monitoring and RGB software can cause issues.
TUF Z790 PRO here. 2x24GB @ 7200 CL34 1.43v on a 14900K. runs ycruncher VT3 for hours but i had to add SA and IVR transmitter and a little PLL voltages.
Really interesting talk!
Jufes, congratz.
you are a great interviewer!
Keep holding their feet to the fire. Apart from that, I'm going to have to disagree with most of the comments and say Shannon was not a good interviewee. He's a marketer with a company and that company has more control over him than even the channels/websites they advertise with. And we know how untrustworthy those are. That's his job, fine, but we'll never get unfiltered opinions from him as long as he works there.
Also that failed flex when Shannon was all "I machined my own pot for Slot A, did you even have that?" and Jufes responded "like a Pentium 2?" and Shannon responded, "no, the Celeron 300A". Pentium 2, along with the infamous 300A, were BOTH slot CPUs. The SAME slot. Slot 1, not Slot A, to be clear.
Gamers, as Jufes has said a million times, are not the same thing as PC enthusiasts. Most of them want the fastest without FAFO. This mindset Shannon has that "if you want to game with the balls out fastest, you shouldn't expect it to be Karhu stable for 10 days" is absolutely wrong.
As long as you buy reasonable parts: if you have a 13900K, you are going to get the advertised 5.5GHz all core and if you buy a 4090, you are going to get the advertised 2.2GHz all core and they will just work. But if you buy DDR5 8000 and, as Shannon disclosed, less than 20% of their 8000 kits are stable on Apex with 13th gen, it isn't an 8000 kit.
Give us QVL lists that aren't pure lies which is what this channel has been complaining about for years. Even sharing the stats he said he has on an Excel spreadsheet would go a long way.
Instead we get hand waving. All the stuff about silicon lottery, drivers, etc. is true but at the end of the day, over 80% of DDR8000 isn't.
As for questioning the validity of stress testing and the nonstop excuses "But what about Windows scheduler" "you need to use an oscilloscope" "the OS parked a core" "a draw call happened" "RGB" "polling" etc. doesn't take away from the fact that Karhu will consistently error at consistent times with a given set of RAM timings and not others. That is a memory issue plain as day.
Question for Shannon, is your buddy Roy . . . . Roy K. aka DUMO by chance ?
Maybe
Awesome convo!
yes lets get a live stream going lol
Love Patriot Viper, have 4 or 5 DDR4 bdie Patriot kits and my old DDR2 or 3 Patriot kit.
We already know Patriot makes good ram. So less segway, more facts! 😁
That look at 8:33
Yes would love a discussion
I’m on 14900K z690 dark and Mdie 6200 Corsair 32GB kit and it’s so hard to get 6400+ MHz stable or even tight timings. DDR5 is truly tricky to stabilize…. It’s really overwhelming…. I’ve tried 13900K also. Just really lost on the timings and voltages sweet spots, but the way people say giving it 10mv back and forth is really time consuming of days. There has to be a better guide… and everyone is stroking their A DIE score, it’s like MDIE talks are barely there or it’s AMD drones
I’ll run stability tests for days, but crash with optimization shaders in Battlefield and crash…. So I know it’s not stable… really could use guidance on someone on discord
very educational
This was awesome, thanks boys
Great video, won’t see this type of video from other creators
Great video, Really enjoyed this!
Thanks to all the real memory makers for the tips, using refresh interval 2000 made input latency nothing even with path crap and deep learning super ship, ray repoopulificaftrity
Great interview! Super informative
Doesn’t ram affect ram click latency and peeking someone in a game really fast by a few milliseconds
First time I'm hearing that Linkpack can burn VRM's and gates... it makes sense as I used it as a measure of stability and the CPU seem to be getting worse....
This guy basically stated everything you been telling people for the longest. 😂 sick vid jufes!
old mates got my idea
just fresh os after tune
We need a live stream with Shannon!