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

[Mr. Fox]

8 hours ago, Papusan said:

What I don't understand... Why suddenly the need for 50% more pins on a minstream platform. Remeber AMD most liekly have to support AM5 several years down the road. Maybe to add in some gimmick that they need to offer to try compete with AMD's real cores? I just can't see why they need so many more pins on the socket. Or just because this is only a real scam to try lure people "why they need new MB"? 

 

More junk on the PCB may need those extra pins. 

https://wccftech.com/intel-14th-gen-meteor-lake-s-desktop-cpus-confirmed-in-leaked-drivers-tiled-chips-for-mainstream-pc-builders/

It could also be that they're so deeply emotionally involved with pleasuring motherboard manufacturing OEM partners that they have to keep doing this to keep them financially engaged. It requires time and money. That could also explain why it has been challenging to get top quality AM4 support for high-end enthusiast boards. Just think how upset most of them would be if they only had the opportunity to sell a new motherboard once every 5 to 6 years versus 2-2.5 years now because of pin-out compatibility over more than two generations. They'd all probably poop their pants. This is exactly why laptops can't be upgradable and have all their crap soldered to the motherboard. It's only to benefit the people selling laptops. To hell with the people buying them, they don't matter. Plus, they're so stupid they won't know any better.

5 hours ago, Reciever said:

I wonder if its because i never updated my sBIOS version for why I never experienced the USB issues, Hiew either from what I can recall as our systems were similar, actually thats the same for most of our Teamspeak group now that I think about it (AM4, not same SKU)

I think it's just inconsistent quality control and sloppy manufacturing. It seems to be an issue at the silicon level of individual CPUs rather than the motherboard. My own impression is the people that haven't had problems just got lucky in the silicon lottery. It also seems to get worse for those who have tried to do anything other than run their CPU stock. Overclocking seems to exacerbate the issue if you were one of the unlucky drawers of a short straw.

[Mr. Fox]

On 6/3/2022 at 3:47 PM, Mr. Fox said:

It took a while, but I finally got 6800 CL30 dialed in as a daily driver memory overclock.

Some interesting discoveries in the process...

• Setting tRFC and tRFCPB too tight improves read speed, but degrades write, copy and latency and the best overall performance and latency are achieved leaving both of them set to Auto and letting the BIOS train the values.
• Likewise, tRCD and tRP set too tight produced errors. Moving from 38 to 39 reduced errors, moving to 40 eliminated them.
• CPU VDDQ and CPU VDD2 and DRAM VPP work best if left on Auto. Setting them manually to the values programmed by the BIOS when set on Auto causes inconsistency in stability.
• Setting tWR and tWR_MR too tight causes inconsistent results, while leaving them on Auto and letting the BIOS train the values had no effect on performance and eliminated errors in testing.
• Setting tWTR and tWTR_L at 14 and 28, respectively, displays as 8 and 19 using Dragon Ball. Asus MemTweakIt doesn't even read those two values and it does not accurately read the tWR value. Using either software tool can result in erroneous interpretation due to the BIOS values not being displayed accurately for a few settings. Most of the values displayed by Dragon Ball match the BIOS settings.

 

 

Just a follow up. Further testing of these same settings allows me to lower my DRAM voltage to 1.470V with no errors, even running more a aggressive and longer-running memory test. Read/write/copy and latency are excellent.

 

 

[Tenoroon]

Wrong thread, sorry guys!

[Papusan]

Not LGA-2551, Intel Socket V1 (LGA-1851) for 14th and 15th Gen Core Series has been leaked

 

Contrary to the previous information, the new socket for desktop Arrow Lake and Meteor Lake-S series is not LGA 2551 but LGA 1851 instead. The 2551-pin version could be a BGA variant for other series, but as leaked today, it will probably not be used for a desktop LGA socket. As reported by Benchlife, the 2551 pin socket is currently not listed by Intel product planning for desktop series, but it might be used by BGA packaged products some time later.

If this is correct, Intel have learned from the compatibility  mess with Socket LGA 1151 (v2). Yep, I upgraded from i7-6700K-i7-7700K-8700K and 9900K on same old socket. Add a few more pins for same socket then you don't have to talk about compatibility screw up above 2 gen chips. Fantastic work Intel.

LGA 1151 v1 was the socket for the 6th and 7th generation chips. Intel re-used same socket also for 8th and 9th gen chips. This means Intel didn't want people with 4 core chips to upgrade to next gen 6 and 8 cores chips eeven if the socket was good enough for that. How difficult was it to make a new socket with enough pins from the ground of for compatibility with more than 2 gen chips, same as AMD ? Was this nice trickery (add 151 pins) done only due maximizing profits down the road? If AMD can make it work, then should also Intel be able to do as well.

 

 

And how much did Intel pay UL/Futuremark for the help with tuning their new garaphics cards? Both cases show ugly works from Intel. I hope they will have to pay for their greed.

 

INTELARC A730M Practical tests show a discrepancy between 3DMark and games

 

So far, Intel's discrete Arc graphics cards have been characterized by their absence. First tests of sold models are ambivalent.

 

Intel’s High-End Arc A730M GPU Is Barely Faster Than An NVIDIA RTX 3050 In Gaming, BETA Drivers Won’t Start Some Games in DX12 Either
What makes this worse is the fact that the RTX 3050 Mobile is an entry-level solution while the Arc A730M GPU will be aimed at high-end gaming laptops. We don't really see the 'high-end' action in these benchmarks though.

 

https://wccftech.com/intel-high-end-arc-a730m-gpu-is-barely-faster-than-an-nvidia-rtx-3050-in-gaming/

[Tenoroon]

13 minutes ago, Papusan said:

Not LGA-2551, Intel Socket V1 (LGA-1851) for 14th and 15th Gen Core Series has been leaked

 

Contrary to the previous information, the new socket for desktop Arrow Lake and Meteor Lake-S series is not LGA 2551 but LGA 1851 instead. The 2551-pin version could be a BGA variant for other series, but as leaked today, it will probably not be used for a desktop LGA socket. As reported by Benchlife, the 2551 pin socket is currently not listed by Intel product planning for desktop series, but it might be used by BGA packaged products some time later.

If this is correct, Intel have learned from the compatibility  mess with Socket LGA 1151 (v2). Yep, I upgraded from i7-6700K-i7-7700K-8700K and 9900K on same old socket. Add a few more pins for same socket then you don't have to talk about compatibility screw up above 2 gen chips. Fantastic work Intel.

LGA 1151 v1 was the socket for the 6th and 7th generation chips. Intel re-used same socket also for 8th and 9th gen chips. This means Intel didn't want people with 4 core chips to upgrade to next gen 6 and 8 cores chips eeven if the socket was good enough for that. How difficult was it to make a new socket with enough pins from the ground of for compatibility with more than 2 gen chips, same as AMD ? Was this nice trickery (add 151 pins) done only due maximizing profits down the road? If AMD can make it work, then should also Intel be able to do as well.

 

 

And how much did Intel pay UL/Futuremark for the help with tuning their new garaphics cards? Both cases show ugly works from Intel. I hope they will have to pay for their greed.

 

INTELARC A730M Practical tests show a discrepancy between 3DMark and games

So far, Intel's discrete Arc graphics cards have been characterized by their absence. First tests of sold models are ambivalent.

I've heard that most if not all of the extra pins will be used for power delivery to smooth everything out. 

This doesn't justify changing the socket though. They could have used LGA 1851 for ADL knowing that they can stick Raptor Lake and Arrow Lake CPUs in there in the future. 

Intel will probably learn their lesson once they start losing more and more market share to AMD due to the fact AMD can support many generations on the same socket. Only time will tell 😉   

[Papusan]

Ain't it fun with Ryzen processors? @Mr. Fox Pick your poison/cancer! What Agesa version would you prefer? Maybe roll the dice?🙃 And do you enjoy pay some money in different lotteries? You could win as you know. But you have a lot bigger chance to loooose. See it this way, the winner takes it all.... Always the one who offer the lottery. 

 

 

Conclusion
You can say a lot of bad things about this Agesa, but she also does a lot of things right. The RAM runs stably, even if sometimes with incomprehensible hurdles, the performance drops slightly and at the same time it is bought with more temperature. For overclockers who don't use the Curve Optimizer, this Agesa is simply useless. It also brings advantages. The USB problem as well as the TPM stuttering is finally over and I'm particularly happy about that. I never really had any problems with the TPM, but I did with the USB slots. So my mechanical keyboard liked to write “oooooooooooooo” with an o keystroke and then work again without any problems. See you next Mallllllllllll. With the Agesa 1.2.0.7 the problem is off the table. And since I generally only overclock using the Curve Optimizer, Agesa 1.2.0 is my choice. 7 actually welcome. The only thing that triggers me here is the loss of performance paired with the higher temperature.

 

For anyone who just wants to use the new Ryzen, Agesa 1.2.0.7 is mandatory anyway, on an older mainboard paired with Zen3 as well. The Agesa is also recommended for normal users who just create XMP, precisely because every previous Agesa also had problems with XMP profiles, sometimes. And anyone who has problems with USB or TPM stuttering, which is by no means everyone, should definitely use this Agesa. I am ambivalent especially with 5800x users who use a not so potent cooling. The temperature could quickly become a limiting factor.

 

And for overclockers who really want to tickle the last ounce of performance out of the CPU, hands off! Agesa 1.2.0.3abc is and remains the first choice here.

 

Apparently, it turns out to me, the Agesa 1.2.0.5 and .06 were testagesas to get the stuttering problem solved. It seems to me that with some processors the signal rates have collapsed during load changes and that is why the stuttering. I have no idea how AMD solved this exactly, but less clock plus higher temperature, paired with "caps" says a lot.

 

Agesa 1.2.0.7 Can she what? And if yes or no, what can it do? A comparison to Agesa 1.2.0.3. deskmodder.com
 

 

 

[Mr. Fox]

29 minutes ago, Papusan said:

Ain't it fun with Ryzen processors? @Mr. Fox Pick your poison/cancer! What Agesa version would you prefer? And do you enjoy pay some money in lotteries? You could win. But you have a bigger chance to loooose.

 

Conclusion
You can say a lot of bad things about this Agesa, but she also does a lot of things right. The RAM runs stably, even if sometimes with incomprehensible hurdles, the performance drops slightly and at the same time it is bought with more temperature. For overclockers who don't use the Curve Optimizer, this Agesa is simply useless. It also brings advantages. The USB problem as well as the TPM stuttering is finally over and I'm particularly happy about that. I never really had any problems with the TPM, but I did with the USB slots. So my mechanical keyboard liked to write “oooooooooooooo” with an o keystroke and then work again without any problems. See you next Mallllllllllll. With the Agesa 1.2.0.7 the problem is off the table. And since I generally only overclock using the Curve Optimizer, Agesa 1.2.0 is my choice. 7 actually welcome. The only thing that triggers me here is the loss of performance paired with the higher temperature.

 

For anyone who just wants to use the new Ryzen, Agesa 1.2.0.7 is mandatory anyway, on an older mainboard paired with Zen3 as well. The Agesa is also recommended for normal users who just create XMP, precisely because every previous Agesa also had problems with XMP profiles, sometimes. And anyone who has problems with USB or TPM stuttering, which is by no means everyone, should definitely use this Agesa. I am ambivalent especially with 5800x users who use a not so potent cooling. The temperature could quickly become a limiting factor.

 

And for overclockers who really want to tickle the last ounce of performance out of the CPU, hands off! Agesa 1.2.0.3abc is and remains the first choice here.

 

Apparently, it turns out to me, the Agesa 1.2.0.5 and .06 were testagesas to get the stuttering problem solved. It seems to me that with some processors the signal rates have collapsed during load changes and that is why the stuttering. I have no idea how AMD solved this exactly, but less clock plus higher temperature, paired with "caps" says a lot.

 

Agesa 1.2.0.7 Can she what? And if yes or no, what can it do? A comparison to Agesa 1.2.0.3. deskmodder.com
 

 

 

AMD is just an open invitation to participate in a comedy of errors, providing little more than an endless cycle of nonsense IMHO. I have extremely low confidence in anything that comes from them. Literally everything I have ever owned that was made by AMD had one or more things wrong with it.
 

 

Nothing is perfect, but in stark contrast to my experience with AMD, I cannot remember owning anything with severe engineering defects or blatant flaws that was made by Intel or NVIDIA. I don't care about brand and I am not a fanboy for anyone. I just have a low tolerance for nonsense.

 

 

[Papusan]

1 hour ago, Mr. Fox said:

Nothing is perfect, but in stark contrast to my experience with AMD, I cannot remember owning anything with severe engineering defects or blatant flaws that was made by Intel or NVIDIA. I don't care about brand and I am not a fanboy for anyone. I just have a low tolerance for nonsense.

Yep, I stay away from hyped products. Either they are defective, flawed or just pure junk. But people want it so……. That’s what we’ll get. 

Don't Believe the Hype: Apple's M2 GPU is No Game Changer

https://www.tomshardware.com/news/apple-m2-gpu-analysis

[Talon]

What I love is the tech tubers eating up and promoting AMD as the company that supports their sockets for extended periods of time blah blah blah. While it is true they now do support all of the Zen chips across the socket this was not always the case and wasn't without massive public outcry and pressure, especially from the tech tubers to get it done. 

 

AMD attempted to dump Zen2 support on B450/X370 and only after pushback did they allow support. Then they completely dumped X370/B450 support for Zen3 for a year, and only after sales slumped due to Alder Lake being a success did they decide to backtrack on that and released BIOS updates to support it. AMD isn't a white knight, they're no different than any other company out to maximize profits and drive up share prices. Supporting a CPU family for 2-3 generations vs 2 isn't really that remarkable for most people. Getting newer boards generally does yield some newer features.  It can allow for better compatibility and support without the need or headache of chasing down AGESA versions, figuring out why you're still logging thousands of errors or still having USB disconnects. 

[Papusan]

On 6/7/2022 at 7:29 PM, Papusan said:

And how much did Intel pay UL/Futuremark for the help with tuning their new garaphics cards? Both cases show ugly works from Intel. I hope they will have to pay for their greed.

 

INTELARC A730M Practical tests show a discrepancy between 3DMark and games

 

So far, Intel's discrete Arc graphics cards have been characterized by their absence. First tests of sold models are ambivalent.

 

Intel’s High-End Arc A730M GPU Is Barely Faster Than An NVIDIA RTX 3050 In Gaming, BETA Drivers Won’t Start Some Games in DX12 Either
What makes this worse is the fact that the RTX 3050 Mobile is an entry-level solution while the Arc A730M GPU will be aimed at high-end gaming laptops. We don't really see the 'high-end' action in these benchmarks though.

 

https://wccftech.com/intel-high-end-arc-a730m-gpu-is-barely-faster-than-an-nvidia-rtx-3050-in-gaming/

Yep, a lot more difficult make it works in games than benchmarks from UL/Futuremark. Probably the main reason Intel delayed the release several times and also prefered to test it on the Asian market first. Good luck to those buyers of Clevo's new high end gaming laptop (270X) that will replace the older real Clevo gaming laptops as 170Xx and P870xx. They will pay the ultimate sacrifice for helping Clevo maxing out profits with guinea pig graphics cards from Intel. 

 

 

 

Intel Arc A730M: Even with official driver support, the gaming performance remains miserable

 

Intel Arc A730M High-End Mobility GPU Still Slower Than RTX 3060M Despite Latest Drivers, Reviews Show Lackluster Gaming Performance
https://wccftech.com/intel-arc-a730m-high-end-mobility-gpu-slower-than-rtx-3060m-despite-latest-drivers/

[Mr. Fox]

I figured that Banshee was overdue to stretch her legs a bit before putting her nose back to the grindstone again. This wicked girl is my daily workhorse. She spends most of her time crushing Excel and PowerBI. Yes, she does that 8-10 hours per day running the clock speeds used here, LOL. Because she can.

 

[jaybee83]

1 hour ago, Mr. Fox said:

I figured that Banshee was overdue to stretch her legs a bit before putting her nose back to the grindstone again. This wicked girl is my daily workhorse. She spends most of her time crushing Excel and PowerBI. Yes, she does that 8-10 hours per day running the clock speeds used here, LOL. Because she can.

 

dude, for serious excel sheets ull need all the clocks u can get, especially from cpu and ram 😄 (and yes, i AM being serious here, no sarcasm!)

[Ashtrix]

Big News Drop by AMD

 

They are firing on all cylinders now. Here's some nice slides for you to understand.

 

 

 

 

 

 

 

 

To summarize

 

Zen 4 is basically going to be a beast no question esp in MT workloads. Also Zen 4 with 3DV is also in pipeline plus Genoa-X 1GB L3 cache lol epic... and Zen 4 shares both 5nm and 4nm, not sure which will end up on 4nm maybe those BGA junk, they also confirmed Zen 5 on 4nm / 3nm with 3D V Cache too. I do not think we will see all of these on consumer side of things, since X3D is hotter, and with 5GHz+ clocks on all cores with heat is a no go maybe another locked down Gaming only chip ? no idea. But I think they might do a short term socket ? Zen 4 will have AM5 and Zen 5 will have AM6 ? Gotta see if they hold AM5 a long lived one or not.

 

Also Zen 4c and 5c is interesting, this is not big,Little junk, these are Bergamo (4c) server parts and 5c is unknown. They are doing this because ARM's Neoverse V1 and N2 platforms are coming soon with big ST performance across a huge number of cores without SMT technology. So AMD is prepping up x86 for that, Intel delayed XEON Sapphire Rapids it was known news as Anandtech reported it, And Intel's XEON with E cores is in 2023-2024.

 

Also AMD confirmed Zen 4 Threadrippers. but it will be probably Pro series garbage only, I hope I'm wrong. But Intel existing SPR itself is delayed and W class is put in favor more, same for AMD they axed basic Threadrippers and using TR Pro with Octa channel memory.

 

For GPU RDNA is looking mighty interesting but I think those ridiculous rumors of 2.5x RDNA2 i think might not be true lol, same for Nvidia, too much hype. Still gotta see what is in the store for us. Oh also chiplet design for GPUs too, very interesting and finally they confirm it haha.

 

 

Anandtech covered all in detail if anyone is interested.

 

P.S - No idea why the images are appearing squished if I use the high resolution ones... the sources for these are from Videocardz, @Reciever maybe you can look into it ? Update - Using URL fixes it, thanks for the PM.

[Papusan]

56 minutes ago, Ashtrix said:

Also Zen 4c and 5c is interesting, this is not big,Little junk, these are Bergamo (4c) server parts and 5c is unknown. They are doing this because ARM's Neoverse V1 and N2 platforms are coming soon with big ST performance across a huge number of cores without SMT technology.

Damn disgusting (Intel). AMD will offer AVX512 for their coming mainstream chips and Intel prefer fuse it off because the (small baby cores) Hybrid chips is meant for Jokebooks. This will be a choice between pest and cholera (Intel vs AMD). 

 

It seems Intel is happy enough offer the best single threaded performance. Become the single threaded KING. And I expect the notebook/Jokebook/tablet market is what they will put all their eggs into. Why bother offer more than absolutely necessary? 

[electrosoft]

15 hours ago, Ashtrix said:

Big News Drop by AMD

 

They are firing on all cylinders now. Here's some nice slides for you to understand.

 

 

 

 

 

 

 

 

To summarize

 

Zen 4 is basically going to be a beast no question esp in MT workloads. Also Zen 4 with 3DV is also in pipeline plus Genoa-X 1GB L3 cache lol epic... and Zen 4 shares both 5nm and 4nm, not sure which will end up on 4nm maybe those BGA junk, they also confirmed Zen 5 on 4nm / 3nm with 3D V Cache too. I do not think we will see all of these on consumer side of things, since X3D is hotter, and with 5GHz+ clocks on all cores with heat is a no go maybe another locked down Gaming only chip ? no idea. But I think they might do a short term socket ? Zen 4 will have AM5 and Zen 5 will have AM6 ? Gotta see if they hold AM5 a long lived one or not.

 

Also Zen 4c and 5c is interesting, this is not big,Little junk, these are Bergamo (4c) server parts and 5c is unknown. They are doing this because ARM's Neoverse V1 and N2 platforms are coming soon with big ST performance across a huge number of cores without SMT technology. So AMD is prepping up x86 for that, Intel delayed XEON Sapphire Rapids it was known news as Anandtech reported it, And Intel's XEON with E cores is in 2023-2024.

 

Also AMD confirmed Zen 4 Threadrippers. but it will be probably Pro series garbage only, I hope I'm wrong. But Intel existing SPR itself is delayed and W class is put in favor more, same for AMD they axed basic Threadrippers and using TR Pro with Octa channel memory.

 

For GPU RDNA is looking mighty interesting but I think those ridiculous rumors of 2.5x RDNA2 i think might not be true lol, same for Nvidia, too much hype. Still gotta see what is in the store for us. Oh also chiplet design for GPUs too, very interesting and finally they confirm it haha.

 

 

Anandtech covered all in detail if anyone is interested.

 

P.S - No idea why the images are appearing squished if I use the high resolution ones... the sources for these are from Videocardz, @Reciever maybe you can look into it ? Update - Using URL fixes it, thanks for the PM.

 

14 hours ago, Papusan said:

Damn disgusting (Intel). AMD will offer AVX512 for their coming mainstream chips and Intel prefer fuse it off because the (small baby cores) Hybrid chips is meant for Jokebooks. This will be a choice between pest and cholera (Intel vs AMD). 

 

It seems Intel is happy enough offer the best single threaded performance. Become the single threaded KING. And I expect the notebook/Jokebook/tablet market is what they will put all their eggs into. Why bother offer more than absolutely necessary? 

 

AMD looks like it is coming in guns ablazing and while Intel will retain the single core performance crown, AMD will win at everything else.

 

What is the logic behind fusing off AVX512?

 

 

[Mr. Fox]

20 minutes ago, electrosoft said:

What is the logic behind fusing off AVX512?

They probably enjoy giving bipolar whiners things to mitch and boan about. Some complained about AVX512 existing and now they can complain about it going away.

 

https://hardware.slashdot.org/story/20/08/22/0550242/intel-defends-avx-512-against-critics-who-hope-it-dies-a-painful-death

 

And, since the industry seems obsessed with trashy low-power turd technology, eliminating that can enhance the wuss factor of the new energy-efficient garbage the sheeple crave.

[Papusan]

15 hours ago, Papusan said:

Damn disgusting (Intel). AMD will offer AVX512 for their coming mainstream chips and Intel prefer fuse it off because the (small baby cores) Hybrid chips is meant for Jokebooks. This will be a choice between pest and cholera (Intel vs AMD). 

 

50 minutes ago, electrosoft said:

What is the logic behind fusing off AVX512?

From bro Fox article.... "Former Intel engineer Francois Piednoel also said the special instruction simply didn't belong in laptops,"

 

Then re-read my post. The new  Hybrid chips is mobile laptop chips that is re-used in desktops. And Intel forgot to fuse it of in the validation process. Remember Intel was so afraid for AMD's progress that they put all work into chips that will work best for laptops. OEM and notebooks is their bread and butter.

AMD increases notebook CPU share during record quarter for quarter.........

Intel could fuse of AVX 512 inrstructions for all chips that goes directly to notebooks. But nope. Unisex is todays word. 

35 minutes ago, Mr. Fox said:

They probably enjoy giving bipolar whiners things to mitch and boan about. Some complained about AVX512 existing and now they can complain about it going away.

 

https://hardware.slashdot.org/story/20/08/22/0550242/intel-defends-avx-512-against-critics-who-hope-it-dies-a-painful-death

 

And, since the industry seems obsessed with trashy low-power turd technology, eliminating that can enhance the wuss factor of the new energy-efficient garbage the sheeple crave.

 

[Mr. Fox]

Only time will tell if AMD breaks from tradition and doesn't suck in the next generation, but I certainly wouldn't hold my breath on that happening. All of their stuff looked great in theory last generation with Ryzen 9, but it turned out to be just another goofed-up cluster cloaked behind fancy-looking Cinebench scores. I take all of the talking head chatter and theoretical greatness about what's coming next with a grain of salt. It will be interesting to see what remains broken and what new things they can find to mess up.

[Papusan]

59 minutes ago, Papusan said:

From bro @Mr. Fox article.... "Former Intel engineer Francois Piednoel also said the special instruction simply didn't belong in laptops,"

"HOW TO FIX INTEL"

 

 

 

 

What’s wrong with Intel, and how to fix it: Former principal engineer unloads
The principal engineer who worked on CPUs from the Pentium III to Core i7 says Intel has lost focus, and it's "lucky" AMD can't make enough CPUs to do more damage.
https://www.pcworld.com/article/393285/whats-wrong-with-intel-and-how-to-fix-it-former-principal-engineer-unloads.html

 

 

[electrosoft]

So are the bulk of you happy or sad about the AVX512 being disabled? I'm indifferent as it didn't make or break my leisure or workflow 🤷‍♀️

[Ashtrix]

11 hours ago, Papusan said:

"HOW TO FIX INTEL"

 

 

 

 

What’s wrong with Intel, and how to fix it: Former principal engineer unloads
The principal engineer who worked on CPUs from the Pentium III to Core i7 says Intel has lost focus, and it's "lucky" AMD can't make enough CPUs to do more damage.
https://www.pcworld.com/article/393285/whats-wrong-with-intel-and-how-to-fix-it-former-principal-engineer-unloads.html

 

 

 

Bro Papu, I remember this guy, and it was not a good memory of this fellow.. I recall him being not a good Engineer to speak or anything. Better watching Dr, Ian Cutress, even though Ian never criticizes any corporation directly in his articles or such. This guy comes off like some guy who got kicked out of Intel and is now trashing them hard.

 

My take on Intel's failure is simple - Investors did not care as they had other prospects which is other corporations,  esp what is happening under Brian Krzanich, so they let everything go on and on and rot. Because the x86 was entire computing industry. So nobody cared as long as their profits soared.

 

Let's go back , Intel's actual talent like Gordon Moore, and CEO Andy Groove, this man was a master he did some great things during his tenure and he predicted there will be only one pureplay foundry in the end and almost it's true now we have just 2 - Samsung, TSMC. Rest all vanished. Also he owned that issue of Intel back in 1995

Intel recalled a major chip in 1995 and turned them into keychains inscribed by the CEO — and the message speaks to Intel's current crisis

 

 

 

Going back again, so yeah since the HPC, Servers, Mainstream all were occupied by Intel the Investors did not care, and what happened ? braindrain. Slowly Intel starting losing focus which is why they put a lot of money into 5G and failed there, then they even lost focus on their NAND, I remember when M18xR2 was there Intel's SSD was top class MLC technology. now ? Sold off dead to SK.Hynix rebranded all of their IP called as Solidgm. 5G thing failed hard, but the suits were fine. Since majority of these corporations are run by same sect of people who have deep pockets and run Banks too. Their new love was found - Apple corporation.

 

So the entire focus was on Apple. And also meanwhile Intel stagnated since Mobile chip industry was booming. Finally the big shots left leaving lesser frys, who demanded change at Intel which caused them to rattle when AMD started with Zen hitting at 6950X HEDT class with mainstream processors. The left over talent took the weight, pressure and tried to keep the boat float. Intel was squandering over 10nm failure which we all know. They sold off their entire 5G smartphone to Apple for peanuts, (Wait now how Apple will claim in-house lol just like their GPU which is basically PowerVR stolen or better put poached from Imagination Technologies)

 

Ultimately they even axed Optane. It's the world's best Memory technology. Destroys SLC to oblivion and 3D NAND as well.

Intel Optane SSD DC P5800X Review: The Fastest SSD Ever Made

Again it needed a ton of cash. So Intel after reeling their focus back trying to find a CEO for years, failed upwards with 14nm while TSMC caught up. When Haswell started Intel was first on the planet to debut FinFet. Now look where they are. Now Optane for consumers is gone as well sold off entire 3D-XPoint technology to Micron and now Micron abandoned it sold off Utah plant to Texas Instruments.

 

Also that guy Murthy Renduchintala, was probably eying for a CEO position after the ousting of Brian on sexual misconduct ROFL. He also left the corporation.  Then they even announced Jim Keller leaving too. See how many holes are there in this company history, this is supposedly a new uArch design entirely and that project is now axed or we will never know the reason. Edit - I missed mentioning how Keller and Murthy probably had clashes. No wonder Keller left and that guy also left. Also Keller is a notable expert on the CPU design, the other person nope.

 

Also I want to put this too. Intel's all SVP and VP roles are now Women only. Sorry but I cannot keep this fact in dark. It's the CA political spectrum. The company is really if you ask me rotting inside. They replace tech and IT guys with MBA hires. Their interim CEO was an MBA guy to keep the company afloat.

 

Quote

Bryant’s replacement is Intel EVP and Chief Revenue Officer Michelle Johnston Holthaus. She is another Intel lifer, having spent 25 years at the company, joining in 1996. She has held multiple roles in reseller management, HQ central marketing and operations, global account management, and currently sits as GM of the Sales, Marketing, and Communications group. In recent memory Holthaus was the keynote speaker at Intel’s Partner Summit. Bringing her to the role is likely a move to strengthen Intel’s bonds with its OEM partners, an aspect that is seeing increased competition in traditional client OEM markets. It also means that Intel has two women leading its largest business units – Holthaus for Client Computing, and Sandra Rivera for Datacenter.

 

Intel is a massive corporation and it's too big and too many politics inside. Also now with all this mega mess of a situation at Intel corporation. Finally they fixed 10nm and it's now denser than TSMC 7N. But their CORE architecture is still from old era, basically Nehalem / Westmere (Core i7 brand 2008) and major change with Sandy Bridge where the RING bus evolved and then added a ton of Cache to it and etc, it was hardly scalable after 2016 so they created MESH Skylake era XEON and HEDT which was worse in every way as it drank power and wasted a ton of electricity plus heat.

 

Still miracle how they pulled off Comet Lake 2020 on that RING bus beyond 8C with 10C20T, but it couldn't take more which is why Intel 10th gen has Cache errors and OCing cache far too high 24.7 might result in Cache degradation (We can see it in latest Z590 BIOS where ASUS AI Algorithm doesn't put insane Cache Ratios vs launch BIOS). Bet that's why RKL also got axed 2C along with 14nm backport (double whammy).

 

Now 2022, they are stuck with P cores 8 Max for 2024-2025, they will add cache, improve front end add more Clocks more Voltage regulation and more as much as they can optimize and then they will debut a new Architecture.

 

After Pat Gelsinger came, he started hiring the old brass, real people who made Intel CORE series. Ian covered it here at Anandtech & here

 

Quote

News in the last 24 hours, based on public statements, states that former Intel Senior Fellow Glenn Hinton, who lists being the lead architect of Intel’s Nehalem CPU core in his list of achievements, is coming out of retirement to re-join the company. (The other lead architect of Nehalem are Ronak Singhal and Per Hammerlund - Ronak is still at Intel, working on next-gen processors, while Per has been at Apple for five years.)

 

That guy right there, he is one behind the uArch not that Francis guy who acts like that on media bashing Intel and claiming. So yeah I hope that Intel's talent whoever left and rehired back put x86 in higher position esp Overclocking and Tuning of Hardware which we buy & own (Ignoring that Management Engine CIA spyware).

 

I hate ARM processors they are junk because they rely heavily on garbage IP blocks to do their work and on top the Software is a gigantic mess. On Android for eg if you do not get the Baseboard firmware and blobs your hardware will never work on newer OS and Kernel. On x86 what we see ? Alder Lake not officially supported on Windows 7 2009 OS, running it scott free. And ability to run Windows 3.1 .exe on Windows 10. That's what called Computing is, not this BS Smartphone addiction of crappy Social media junk with constant algorithm fed propaganda.

 

Anyways felt to post this lol sorry.

 

AVX512, it's used for some AI math, yeah we do not use much but from the Igor's Lab testing AVX512 enabled makes Alder Lake scream fast (Faster Uncore, Ring, IMC and everything) Not disable and waste the damn transistors (irony here is we disable E cores, yeah unfortunate but that's what it is, beefier silicon yes but weak nope).

 

I welcome anything more for x86 processors NOT LESS. Anything new that works ? Sure sign me up. But why cut off when you already have. Pathetic move by Intel because Toy Trash cores cannot handle the AVX512, and since very few care to run ADL with E junk disabled, so Intel thought it's fine to axe it plus they save money on validation process too esp on Desktop, they give the people 8P here game on them lol pathetic and BGA junk ofc E core = more cores marketing to fool dumb and easy cash milking with tons of SKUs. Big shame, since AMD is going big on AVX512, anything given to us mere mortals is good move no matter how you dice it.. See how pathetically Nvidia castrates GeForce for that licensed Quadro Drivers ? awful.

[Papusan]

2 hours ago, electrosoft said:

So are the bulk of you happy or sad about the AVX512 being disabled? I'm indifferent as it didn't make or break my leisure or workflow 🤷‍♀️

As bro @Ashtrixsaid. Make new silicon then afterwards fuse off features is disgusting. And as bro Ashtrix again said.... The editor at Igors lab scrapped the idea upgrade to the better KS chips because of Intel's retarded move removing AVX512 support. 

 

As a note. My quite good 12900K have AVX512 support. Not that I use it but i never know later. Better have more than less, isn't it? 🙂 Imagine you bought a car with 5 step automatic gearbox, then the car manufacurer said no you don't needed it (no need for reaching 200 mph) and then fused of the 5 gear after your last visit to the car service. 

1 hour ago, Ashtrix said:

I hate ARM processors they are junk because they rely heavily on garbage IP blocks to do their work and on top the Software is a gigantic mess. On Android for eg if you do not get the Baseboard firmware and blobs your hardware will never work on newer OS and Kernel. On x86 what we see ? Alder Lake not officially supported on Windows 7 2009 OS, running it scott free. And ability to run Windows 3.1 .exe on Windows 10. That's what called Computing is, not this BS Smartphone addiction of crappy Social media junk with constant algorithm fed propaganda.

 

Anyways felt to post this lol sorry.

 

AVX512, it's used for some AI math, yeah we do not use much but from the Igor's Lab testing AVX512 enabled makes Alder Lake scream fast (Faster Uncore, Ring, IMC and everything) Not disable and waste the damn transistors (irony here is we disable E cores, yeah unfortunate but that's what it is, beefier silicon yes but weak nope).

 

I welcome anything more for x86 processors NOT LESS. Anything new that works ? Sure sign me up. But why cut off when you already have. Pathetic move by Intel because Toy Trash cores cannot handle the AVX512, and since very few care to run ADL with E junk disabled, so Intel thought it's fine to axe it plus they save money on validation process too esp on Desktop, they give the people 8P here game on them lol pathetic and BGA junk ofc E core = more cores marketing to fool dumb and easy cash milking with tons of SKUs. Big shame, since AMD is going big on AVX512, anything given to us mere mortals is good move no matter how you dice it.. See how pathetically Nvidia castrates GeForce for that licensed Quadro Drivers ? awful.

This is Intel today....  Intel's new line of low-power, low-cost and low-performance x86 and x86-64 microprocessors. I expect they see profits into this. 

 

Intel Silently Slips in new Atom C5000 Series Processors

By Hilbert Hagedoorn on: 06/09/2022 10:15 AM | 3 comment(s) ]

Intel has announced the first six models in the Intel Atom C5000 series. All of the new parts have QAT onboard, which will be a bigger feature later in 2022 and into 2023. PCIe generation for all of the parts is listed as only PCIe Gen3 on Ark. Read more

[Mr. Fox]

2 hours ago, electrosoft said:

So are the bulk of you happy or sad about the AVX512 being disabled? I'm indifferent as it didn't make or break my leisure or workflow 🤷‍♀️

I am not sure of a reason I should care. There could be some reason that I should care about it that I am not aware of. Maybe just as a act of defiance, since Intel doesn't want me to have it. Other than that, probably not.

[Papusan]

On 6/11/2022 at 2:29 AM, Mr. Fox said:

I am not sure of a reason I should care. There could be some reason that I should care about it that I am not aware of. Maybe just as a act of defiance, since Intel doesn't want me to have it. Other than that, probably not.

If Intel got rid of the baby cores meant for Jokebooks and instead added more real cores and on top went back and offered AVX512 support you could gain some performance uplipt in 3DMark Time Spy Extreme.... And AMD will of course have it. See it this way... Intel double castrated their newer chips (Loads of E-cores and fused off AVX512 support). Can't beat stupid. Or can you?

 

https://www.anandtech.com/show/11895/a-first-look-at-futuremarks-time-spy-extreme-dx12-benchmark-in-4k

 

Intel continue help AMD to take market shares due own stupidity/Greed. Not the way it should have been. Intel should have been the frontman and offer people more and better features/even more performance. Not the opposite. And they are taken with their pants down, time after time from AMD. 

[Papusan]

Does this mean we will get a much cheaper KF chips if Intel don't pack in the GFX tile on all PCBs? I doubt it. All too many would scrap pay premium for an K chips if that become an option.

 

 

Intel 14th Gen Meteor Lake-P Mobility CPU Pictured, Gets Close Up Die Shot With 6 P-Cores & 8 E-Cores
https://wccftech.com/intel-14th-gen-meteor-lake-cpu-die-shot-6-p-core-8-e-core-intel-4-process-node-leak/

 

Yep, its an mobile Cpu. But the the desktop chips is in short same Hybrid mess. @Ashtrix