*Official Benchmark Thread* - Post it here or it didn't happen :D

[tps3443]

6 minutes ago, electrosoft said:

 

Congrats bro and at a killer price no less!

 

I'm sure the Suprim X 4080 Super will sell relatively quick at that price.

It’s sold 😃 (Going to Canada) 

 

I was very tempted to just keep it as a backup or spare high end GPU. But then that didn’t make much sense. Plus, I’m not that rich. Haha. 
 

The 4090 is nice! I like it a lot so far. It’s strong. I saw a 42% higher Superposition score over my 4080 Super. 🤩

 

Also. What sort of memory overclocks are people getting out of 4090’s?

[electrosoft]

3 minutes ago, tps3443 said:

It’s sold 😃 (Going to Canada) 

 

I was very tempted to just keep it as a backup or spare high end GPU. But then that didn’t make much sense. Plus, I’m not that rich. Haha. 
 

The 4090 is nice! I like it a lot so far. It’s strong. I saw a 42% higher Superposition score over my 4080 Super. 🤩

 

Also. What sort of memory overclocks are people getting out of 4090’s?

 

Just using the stock vBIOS shipped on the original  I usually boost to 2820-2835mhz out of box gaming and with an MSI OC I game around 3120-3150 on my Suprim X Liquid 4090.

 

I will say temps have crept up slightly over the last 19 months since purchase in January 2023 so a repaste would probably do wonders.

 

My memory tops out around +1600. I've seen some samples just blast straight to +2000 right out of the box. Just don't get a dud like my other Suprim X Liquid I tried that couldn't even go +700 and only boosted to 2650ish and OC'd stalled at 2900.

 

 

 

 

[Mr. Fox]

It was 113°F today. Tomorrow the forecast is 115°F with a nighttime low of 103°F. But, it's a dry heat.

 

[tps3443]

3 hours ago, electrosoft said:

 

Just using the stock vBIOS shipped on the original  I usually boost to 2820-2835mhz out of box gaming and with an MSI OC I game around 3120-3150 on my Suprim X Liquid 4090.

 

I will say temps have crept up slightly over the last 19 months since purchase in January 2023 so a repaste would probably do wonders.

 

My memory tops out around +1600. I've seen some samples just blast straight to +2000 right out of the box. Just don't get a dud like my other Suprim X Liquid I tried that couldn't even go +700 and only boosted to 2650ish and OC'd stalled at 2900.

 

 

 

 

So, this card doesn’t have a very amazing core unfortunately 🥲. I use Cyberpunk 2077 at 4K for testing. And I can run 2,955 solid! Anymore is a no go. Even just 2,970 will crash lol. Temps are warm about 69-72C. But that’s its limit using auto fan or even 100% fan. Now, the memory is absolutely solid at +1700 so far, maybe can do a little wee bit more. 
 

My 4080 Super could handle about 3,060-3,070Mhz pretty dang  well. 
 

Anyways, it is what it is. I’m still happy with it! I imagine chilled water will help quite a bit. I know Cyberpunk at 4K puts a “Real” limit for the most part. Those RT cores either can or can’t it seems. 

 

Testing was done with Galax HOF 666 bios. 

[ryan]

Congrats @tps3443 get r done...looking forwards to overclocked/stock gaming benchmarks. Can't imagine how much your CPU and ram will help, but there's a small chance you may end up CPU limited. 4090 is a beast much like the 1080ti was...needlessly powerful..

[tps3443]

28 minutes ago, ryan said:

Congrats @tps3443 get r done...looking forwards to overclocked/stock gaming benchmarks. Can't imagine how much your CPU and ram will help, but there's a small chance you may end up CPU limited. 4090 is a beast much like the 1080ti was...needlessly powerful..

It’s a nice GPU. Not quite CPU limited. Even 1080P and hammering 300+ fps, it’s all GPU limits in the games I play. 
 

@Mr. Fox how did your gaming OC clock on air? And on water? 

[Raiderman]

@electrosoft agesa 1.2 finally dropped for our boards.

 

https://us.msi.com/Motherboard/MPG-X670E-CARBON-WIFI/support

[Papusan]

1 hour ago, Raiderman said:

@electrosoft agesa 1.2 finally dropped for our boards.

 

https://us.msi.com/Motherboard/MPG-X670E-CARBON-WIFI/support

 

Will it come a new update that makes the new gen Ryzen better than this trash below?

 

9950X results have popped up. Similar disappointment as 9600X and 9700X trash results. Tempting bro @Mr. Fox? You really want that single core performance uplifth? 

 

New benchmark scores for the 16-core AMD Ryzen 9 9950X processor were revealed, showcasing its performance enhancements over previous models

 

In competitive testing against Intel's offerings, the Ryzen 9 9950X demonstrated superior single-core performance compared to the Intel Core i9-14900KS and Core i9-14900K, although it fell slightly behind the 14900KS and was comparable to the 14900K in multi-core tests.

 

https://www.guru3d.com/story/ryzen-9-9950x-processor-singlecore-is-higher-than-ryzen-9-7950x-multicore-is/

 

Will be 38C here in Canary but with high humidity 😞 Equal awful as posted Geek results with Zen5 chips. 

[Mr. Fox]

8 hours ago, tps3443 said:

It’s a nice GPU. Not quite CPU limited. Even 1080P and hammering 300+ fps, it’s all GPU limits in the games I play. 
 

@Mr. Fox how did your gaming OC clock on air? And on water? 

I hardly used it on air. Ordered the water block and switched it as soon as I could. I do remember being impressed with the air cooler's ability to keep it cool. It has never been benched on the chiller so I haven't really done any serious benching with it. Memory overclock on both 4090s is stable at +1700-1800 depending on the task. I use the Galax 1000W vBIOS on both GPUs. I did not have good results with the 666W vBIOS. Both GPUs worked better with the stock vBIOS than they did with the Galax 666W and both work best with the 1000W. For 3000+ on core you need to keep the voltage locked at 1.100V. At least it do. If it drops below 1.100V (the wimpy max) stability gets sketchy. I use Afterburner curve tool to lock it at 1.100V. I think the 4090 Suprim may have a slightly stronger core, but I've not done a good head-to-head comparison on chilled water. The 4090 really needs a core max of 1.200V to shine. The 1.100V limit is pretty wimpy. The OEMs and NVIDA are probably worried about the crappy 12VHPWR sockets melting and castrated the voltage to keep the watts lower.

1 hour ago, Papusan said:

@Mr. Fox? You really want that single core performance uplifth? 

Nope. I don't really care about that. It's the least important measurement to me, and I don't do any single core benchmarks. When I consider the mentality of the people that do care and make a big deal about it, it makes me care about it even less.

22 hours ago, Mr. Fox said:

I think maybe I am going to be holding the line and might be purchasing obsolete hardware going forward.

@Papusan I am working our a deal for a ASUS ROG Rampage IV Gene and 4820K with 32GB (4x8GB) Kingston HyperX 2400MHz quad channel and a Titan GPU. I am doing an even trade for some spare parts I am not using. I should have those parts in another week or two. I will initially install them on the open bench. I think I will use it with the CR360 Dark AIO since that is just sitting on a shelf.

 

After extracting everything the 4820K can muster, I am probably going to grab a 4960X and a Xeon E5-1680 V2 to rack up some extra CPU hardware points. These CPUs are direct cheap on eBay.

 

After I am done racking up some points, I might put everything into the Dark Base Pro 901 case, load it up with mechanical HDDs and stick it under one of my desks to use it as a Windows 7 machine or Linux server. Not sure yet, just an idea. If I do that the Apex Encore and Gigabyte 4090 will move over to the open bench.

 

Since it is HEDT it has 40 PCIe lanes and will support X16 *2 GPU along with NVMe, SATA or NIC add-in cards. I might grab a second Titan or another 2080 Ti for some SLI benching if I can get them cheap enough. I am also thinking about grabbing a cheap 240 radiator to use with the 2080 Ti Waterforce GPU. I already have a spare D5 and reservoir I can connect to it.

 

In a couple of years I expect X299 parts will be just as cheap as these X79 pieces. I still have my X299 CPU delid tool.

[Papusan]

51 minutes ago, Mr. Fox said:

4090 really needs a core max of 1.200V to shine. The 1.100V limit is pretty wimpy. The OEMs and NVIDA are probably worried about the crappy 12VHPWR sockets melting and castrated the voltage to keep the watts lower.

 

From what I have read and learn,,,, You need exactly 1.150 to reduce/stop clock stretching om 4090s. People believe they run with correct boost clock but it isn't. This is properly tested from the author of the unlocked nvflash for 4000 series cards. No one should use any org vbios on revised cards with the 1.075. 

 

Edit. Can't remember his name now but i think its Kefi. He also posted about this on his reddit thread. And his youtube channel. 

[tps3443]

1 hour ago, Mr. Fox said:

I hardly used it on air. Ordered the water block and switched it as soon as I could. I do remember being impressed with the air cooler's ability to keep it cool. It has never been benched on the chiller so I haven't really done any serious benching with it. Memory overclock on both 4090s is stable at +1700-1800 depending on the task. I use the Galax 1000W vBIOS on both GPUs. I did not have good results with the 666W vBIOS. Both GPUs worked better with the stock vBIOS than they did with the Galax 666W and both work best with the 1000W. For 3000+ on core you need to keep the voltage locked at 1.100V. At least it do. If it drops below 1.100V (the wimpy max) stability gets sketchy. I use Afterburner curve tool to lock it at 1.100V. I think the 4090 Suprim may have a slightly stronger core, but I've not done a good head-to-head comparison on chilled water. The 4090 really needs a core max of 1.200V to shine. The 1.100V limit is pretty wimpy. The OEMs and NVIDA are probably worried about the crappy 12VHPWR sockets melting and castrated the voltage to keep the watts lower.

Nope. I don't really care about that. It's the least important measurement to me, and I don't do any single core benchmarks. When I consider the mentality of the people that do care and make a big deal about it, it makes me care about it even less.

Sounds like your cards might be okay core wise and memory. I tried three different bios so far, and I did try locking the core yesterday to 1.100V. I physically can’t do 2,970Mhz unfortunately. The Galax Bios was working pretty well for me, I also tried the Gigabyte gaming OC bios with the unlocked 1.100V, and also a new Gaming OC bios from April 2024. I might just leave this one on air cooling and be happy with it like my 4080 Super. I really enjoyed that card on air.

[Mr. Fox]

1 hour ago, Papusan said:

 

From what I have read and learn,,,, You need exactly 1.150 to reduce/stop clock stretching om 4090s. People believe they run with correct boost clock but it isn't. This is properly tested from the author of the unlocked nvflash for 4000 series cards. No one should use any org vbios on revised cards with the 1.075. 

 

Edit. Can't remember his name now but i think its Kefi. He also posted about this on his reddit thread. And his youtube channel. 

How do you get the voltage to 1.150V though? Is there a vBIOS that allows more than 1.100V apart than owning a Galax HOF 4090 with XOC vBIOS and overclock software to increase the voltage? I am not aware of any way to increase the voltage unless you own a Galax HOF GPU or break out the soldering iron and use Elmore hardware mods.

1 hour ago, tps3443 said:

Sounds like your cards might be okay core wise and memory. I tried three different bios so far, and I did try locking the core yesterday to 1.100V. I physically can’t do 2,970Mhz unfortunately. The Galax Bios was working pretty well for me, I also tried the Gigabyte gaming OC bios with the unlocked 1.100V, and also a new Gaming OC bios from April 2024. I might just leave this one on air cooling and be happy with it like my 4080 Super. I really enjoyed that card on air.

Yours just might not be cold enough with the air cooler. I know for a fact that the Gigabyte 4090 will not do 3000+ stable on ambient water on the smaller (unchilled) loop. Core clock speeds since Maxwell forward are capped/throttled based on temperature ranges, and got capped worse from Turing forward. The 4090 Suprim has always been run with a much larger liquid cooling system with the chiller, so I cannot directly compare the two GPU in that scenario. I just haven't been motivated enough to test it on the loop with the 5 gallon reservoir and chiller. And, my office ambient temps are just way too hot to be ideal for regular water cooling, and truly terrible for air cooling. It's 7:30 AM here with the central AC set to 73°F and my office is already almost 80°F. It will be 85°F in my office before the end of the day. And, that is only with my ITX (work) computer running. If I turn on two or all three desktops the ambient temperatures go even higher.

[Papusan]

3 hours ago, Mr. Fox said:

How do you get the voltage to 1.150V though? Is there a vBIOS that allows more than 1.100V apart than owning a Galax HOF 4090 with XOC vBIOS and overclock software to increase the voltage? I am not aware of any way to increase the voltage unless you own a Galax HOF GPU or break out the soldering iron and use Elmore hardware mods.

 

Yup, you need the Hof as mine (you know I have access to everything forbidden on the web) or solder and use Elmore tool. There is nothing outside this. Add +1.28v and see your card fry in real time without sub Zero temps. Minus 5 aint enough, LOOL

 

But everyone should have enough voltage available in firmware from Nvidia to avoid clock stretching om Nvidia Cards. Nvidia crippled own Products without reason. Good custom cards should all have 1.150v available in oc mode. The only real limits should be max temp throttling target. At least let the gpu run with max voltage as 1.150 up to 75-80C. This should be enough for most.

 

Be happy one single OEM still can do this. And who knows with 5000 series?  I'm quite sure Nvidia vas damn pissed for the custom nvflash that breaked how firmware with 4000 series worked. And they are certainly not happy people could XRoss-flash VBIOS from all oems custom vbios on their FE cards with this custom mod.

[tps3443]

4 minutes ago, Papusan said:

 

Yup, you need the Hof as mine or solder and use Elmore tool. There is nothing outside this.

 

Be happy one single OEM still can do this. And who knows with 5000 series?  I'm quite sure Nvidia vas damn pissed for the custom nvflash that breaked how firmware with 4000 series worked. And they are certainly not happy people could XRoss-flash VBIOS from all oems custom vbios on their FE cards. 

Maybe they’re mad, or maybe not lol. There are many 4090’s that can run a higher core clock limited to 1.070V than me unlocked at 1.100V 😂

[Papusan]

37 minutes ago, tps3443 said:

Maybe they’re mad, or maybe not lol. There are many 4090’s that can run a higher core clock limited to 1.070V than me unlocked at 1.100V 😂

 

Then let the worse binned cards have voltage raised from 1.1 to 1.150 🤩 Nvidia always use best silicon for own FE cards. And now they even offer 600w. Aka TGP above what some AIB partners offer for own custom cards. Then what next... Offer dual vbios and better colling and screw own partners further 😉Btw... Bought an Android phone for this 1 month summer vacation because my old crapple was near dead and I hate the Android crap. My old iPhone had many weird stupid solutions but damn the Android trash have other stupid flaws. I said to my son... None have good engineers with brain!!! None can make good great jokephones. Sometimes you wish you didn't need more than one old DORO.

[tps3443]

8 minutes ago, Papusan said:

 

Then let the worse binned cards have voltage raised from 1.1 to 1.150 🤩 Nvidia always use best silicon for own FE cards. And now they even offer 600w. Aka TGP above some AIB custom cards. Then what next... Offer dual vbios and better colling and screw own partners further 😉

Yep! That would be nice. GPU’s leave a lot to be desired and it’s tough to land something perfect everytime. Good core, good memory, and no coil whine. 
 

Are you getting a 5090 HOF?

[Papusan]

2 hours ago, tps3443 said:

Yep! That would be nice. GPU’s leave a lot to be desired and it’s tough to land something perfect everytime. Good core, good memory, and no coil whine. 
 

Are you getting a 5090 HOF?

 

Not sure in the beginning. im not an early adopter on new expensive tech as etc gpus. As you know I often have big fun with older cards. And remember Im retired from works due destroyed health. Already on the 15th year retired. And I have spent 20.000 usd on my house last 10 months. And new 20.000 usd will be used the coming 2-3 months. Sometimes you need to prioritize what is most important for you and your family. But it will be new tech, but need to see if it can serve my hobby well. I hate spend money if the product is not what i expect. And I have said it many times... New/newer isn't always better. Or take I wrong (about moderne tech) as usual bro @ryan😁 

 

68% higher IPC on the E-cores and 14% higher IPC on the P-cores than Meteor Lake. 

 

The 16 Baby cores offer 50% of the whole performance of the 14900K

 

So 34% uplifth + 14% on P cores thats half the cpus performance means 7% uplifth for the full cpu if you count in P-cores.

 

So 34+7 = 41% gain. Then pick of the missing HT gain that means around 23% less performance and you see a measly max 18% performance uplifth in MT performance for unlocked 15th gen. Thats not good news in my head. Almost as Intel and AMD trying to share the whole cpu cake. Remember AMD said 16% in MT performance uplifth. This is not real competition!!

 

So this has to be todays joke or reality. Pick what you want. Intel trade performance for better efficiency (no more HT to increase single Core performance to come on par with rest of the stupidity and better power efficiency). Remember single core performance is the new modern trend to compare performance. What a perverse new tech world.

 

Hope you understood 3/4 of my rambling 🙂 But we will see who smile last. Or had correct.

 

https://www.tomshardware.com/pc-components/cpus/intel-upcoming-cpu-models-confirmed-by-diagnostic-tool-arrow-lake-and-lunar-lake-skus-revealed

[electrosoft]

New TM5: https://github.com/CoolCmd/TestMem5/releases/tag/v0.13.1

 

7 hours ago, Raiderman said:

@electrosoft agesa 1.2 finally dropped for our boards.

 

https://us.msi.com/Motherboard/MPG-X670E-CARBON-WIFI/support

 

woot, snagged it. Let's see if it fixes boot times as advertised. Good thing as I was *just* getting ready to flash the last version before a clean install.

 

6 hours ago, Papusan said:

 

Will it come a new update that makes the new gen Ryzen better than this trash below?

 

9950X results have popped up. Similar disappointment as 9600X and 9700X trash results. Tempting bro @Mr. Fox? You really want that single core performance uplifth? 

 

New benchmark scores for the 16-core AMD Ryzen 9 9950X processor were revealed, showcasing its performance enhancements over previous models

 

In competitive testing against Intel's offerings, the Ryzen 9 9950X demonstrated superior single-core performance compared to the Intel Core i9-14900KS and Core i9-14900K, although it fell slightly behind the 14900KS and was comparable to the 14900K in multi-core tests.

 

https://www.guru3d.com/story/ryzen-9-9950x-processor-singlecore-is-higher-than-ryzen-9-7950x-multicore-is/

 

Will be 38C here in Canary but with high humidity 😞 Equal awful as posted Geek results with Zen5 chips. 

 

~14% uplift in ST 7950x vs 9950x but only ~7% uplift in MT? That dog don't hunt. I'll have to see the full reviews but so far nothing super special to compel me to upgrade. I'll still wait for the X3D versions while I tinker away with this new (to me) 14900KS and also wait to see what Intel brings to the table with AL.

 

3 hours ago, tps3443 said:

Sounds like your cards might be okay core wise and memory. I tried three different bios so far, and I did try locking the core yesterday to 1.100V. I physically can’t do 2,970Mhz unfortunately. The Galax Bios was working pretty well for me, I also tried the Gigabyte gaming OC bios with the unlocked 1.100V, and also a new Gaming OC bios from April 2024. I might just leave this one on air cooling and be happy with it like my 4080 Super. I really enjoyed that card on air.

 

Not like it isn't a monster card right out of the box. If you block and chill it, you'll slap on what? 100-150mhz extra?

 

It does make me wonder what mine would do blocked w/ the 1000w vBIOS though since it clocks so high already as is....

 

 

[Papusan]

6 hours ago, electrosoft said:

14% uplift in ST 7950x vs 9950x but only ~7% uplift in MT? That dog don't hunt. I'll have to see the full reviews but so far nothing super special to compel me to upgrade. I'll still wait for the X3D versions while I tinker away with this new (to me) 14900KS and also wait to see what Intel brings to the table with AL.

 

Yup, nothing in put in stone yet but some results should be close enough, this close to launch. As I said.... Single core performance with the promised Power efficiency as the gods savoir and you see where I go. Single core performance doesnt always show the whole MT performance. And the different benchmarks doesnt show how it perform. Just look at the Geek and the physics scores in 3DM suites benchmarks with Ryzzzzan. I have +31K in TS and 16K in TSE physics. And +64K in FS physics with trash cores disabled. How close is Zen5 to older Intel? Also my physics is 154.000 in older 3DM Vantage. And running the whole benchmark tests. Everything below this is a failure for next gen tech. The hype train continue with the modern 🙂

 

But I always hope I have wrong

@johnksss @Rage Set @Mr. Fox @Raiderman @jaybee83 @Clamibot @Reciever @tps3443 @chew++++.

 

What you think bro @ryan? Are we doomed due AI and or the tech companies greed?🤐

[tps3443]

52 minutes ago, electrosoft said:

New TM5: https://github.com/CoolCmd/TestMem5/releases/tag/v0.13.1

 

 

woot, snagged it. Let's see if it fixes boot times as advertised. Good thing as I was *just* getting ready to flash the last version before a clean install.

 

 

~14% uplift in ST 7950x vs 9950x but only ~7% uplift in MT? That dog don't hunt. I'll have to see the full reviews but so far nothing super special to compel me to upgrade. I'll still wait for the X3D versions while I tinker away with this new (to me) 14900KS and also wait to see what Intel brings to the table with AL.

 

 

Not like it isn't a monster card right out of the box. If you block and chill it, you'll slap on what? 100-150mhz extra?

 

It does make me wonder what mine would do blocked w/ the 1000w vBIOS though since it clocks so high already as is....

 

 

I think so too. I’m debating ordering a waterblock for it. 

[Raiderman]

8 hours ago, Mr. Fox said:

I hardly used it on air. Ordered the water block and switched it as soon as I could. I do remember being impressed with the air cooler's ability to keep it cool. It has never been benched on the chiller so I haven't really done any serious benching with it. Memory overclock on both 4090s is stable at +1700-1800 depending on the task. I use the Galax 1000W vBIOS on both GPUs. I did not have good results with the 666W vBIOS. Both GPUs worked better with the stock vBIOS than they did with the Galax 666W and both work best with the 1000W. For 3000+ on core you need to keep the voltage locked at 1.100V. At least it do. If it drops below 1.100V (the wimpy max) stability gets sketchy. I use Afterburner curve tool to lock it at 1.100V. I think the 4090 Suprim may have a slightly stronger core, but I've not done a good head-to-head comparison on chilled water. The 4090 really needs a core max of 1.200V to shine. The 1.100V limit is pretty wimpy. The OEMs and NVIDA are probably worried about the crappy 12VHPWR sockets melting and castrated the voltage to keep the watts lower.

Nope. I don't really care about that. It's the least important measurement to me, and I don't do any single core benchmarks. When I consider the mentality of the people that do care and make a big deal about it, it makes me care about it even less.

@Papusan I am working our a deal for a ASUS ROG Rampage IV Gene and 4820K with 32GB (4x8GB) Kingston HyperX 2400MHz quad channel and a Titan GPU. I am doing an even trade for some spare parts I am not using. I should have those parts in another week or two. I will initially install them on the open bench. I think I will use it with the CR360 Dark AIO since that is just sitting on a shelf.

 

After extracting everything the 4820K can muster, I am probably going to grab a 4960X and a Xeon E5-1680 V2 to rack up some extra CPU hardware points. These CPUs are direct cheap on eBay.

 

After I am done racking up some points, I might put everything into the Dark Base Pro 901 case, load it up with mechanical HDDs and stick it under one of my desks to use it as a Windows 7 machine or Linux server. Not sure yet, just an idea. If I do that the Apex Encore and Gigabyte 4090 will move over to the open bench.

 

Since it is HEDT it has 40 PCIe lanes and will support X16 *2 GPU along with NVMe, SATA or NIC add-in cards. I might grab a second Titan or another 2080 Ti for some SLI benching if I can get them cheap enough. I am also thinking about grabbing a cheap 240 radiator to use with the 2080 Ti Waterforce GPU. I already have a spare D5 and reservoir I can connect to it.

 

In a couple of years I expect X299 parts will be just as cheap as these X79 pieces. I still have my X299 CPU delid tool.

I agree wholeheartedly on single thread scores. It doesn't interest me in the least. I'd rather kiss my sister than to sit and endure a single threaded benchmark.

On another note, besides my hardware addiction, I've just purchased another money sucking hobby.

 

[Mr. Fox]

3 hours ago, electrosoft said:

New TM5: https://github.com/CoolCmd/TestMem5/releases/tag/v0.13.1

Nice! Thanks for the link. Maybe it will be more optimal for DDR5.
Here was the Russian text: Часть секции ресурсов перекинута из TM5.EXE в TM5.DLL, чтобы тупые антивирусы заткнулись.
Google Translate: Part of the resource section is moved from TM5.EXE to TM5.DLL so that stupid antiviruses shut up.

12 minutes ago, Raiderman said:

I agree wholeheartedly on single thread scores. It doesn't interest me in the least. I'd rather kiss my sister than to sit and endure a single threaded benchmark.

On another note, besides my hardware addiction, I've just purchased another money sucking hobby.

 

Wow, I am jealous. Sweet new toy.

[cylix]

Interesting, let's see how it will develop

 

 

[electrosoft]

1 hour ago, cylix said:

Interesting, let's see how it will develop

 

 

 

I watched this last night and found it really interesting. I do wonder if there is any correlation between the SA Bug and the overall failure rates of CPUs long term considering these data centers are targeting lower power profiles as they don't need screaming edge performance.

 

In the end, I think we're going to find Intel pushed these chips too hard for their fab and power/heat levels in an attempt to stay with and/or beat AMD. SA Bug is just icing on the cake.

 

 

[tps3443]

1 hour ago, electrosoft said:

 

I watched this last night and found it really interesting. I do wonder if there is any correlation between the SA Bug and the overall failure rates of CPUs long term considering these data centers are targeting lower power profiles as they don't need screaming edge performance.

 

In the end, I think we're going to find Intel pushed these chips too hard for their fab and power/heat levels in an attempt to stay with and/or beat AMD. SA Bug is just icing on the cake.

 

 

I have been getting this “GPU out of video memory” error since 13th gen first launched, all of those errors in the video are probably from me lol. I would use it as my quick and easy stability test lol. I knew that if BF2042 would not spit that GPU out of VRAM error, or The Last of US, during their shader decompression, then my system was 100% stable lol. The largest reason for this error is simply from voltage/clock speed/ power/heat. There’s some sort of silicon issue with that initial launch load going from 10-180 watts in 0.3 seconds and if that voltage isn’t enough, or if those temps aren’t cold enough. It’s going to fail. I actually had one last night lol. This one popped up after closing the game though so it did not bother me 🤣

 

I always undervolt using LLC and AC/DC etc, so I stretch the CPU‘s to their lowest possible voltage required to save on power and reduce heat. And this is where I have always seen the errors. (Not from running the chips stock) now, I use a chiller so these chips stay cold so that would probably explain why I never see the errors when running stock/auto voltage. 
 

Asus updated the bios and moved away from their LLC3 standard to now using LLC5 which made many of them go away. If your load voltage is high enough you won’t get that error at all. Pretty wild seeing this though. It fixed the problem for most. But it always came down to just a weird silicon sensitivity problem. You can always force the error to happen on any chip out there I doesn’t matter how amazing it is. They all spit the same error if you starve them too much on voltage with these very quick ramping shader compiling loads. 
 

I’m not sure who to blame here. But it’s something wrong with the CPU’s I guess. I can always make them work either way though. Hopefully next gen won’t have the issue. I recommend anyone with an Intel 13th/14th gen to run Cinebench R15 a few times. If they can get through that, they should be okay. Something about R15 and the instruction set it uses is much harder to run on a 13th/14th gen Intel than R23/R24. If they can’t get through R15, add more voltage lol. Then once it can run reliably, add another +0.010 and you should be good for never having “Out of GPU memory errors”