Intel Might Drop 10nm node for Desktop processors

Published by

Click here to post a comment for Intel Might Drop 10nm node for Desktop processors on our message forum
#5720736
https://forums.guru3d.com/data/avatars/m/181/181063.jpg
JamesSneed:

This was simply wccftech making stuff up to get clicks. They didn't even pull the article. They state stuff as rumor what they really mean is they are starting a new rumor with nothing to back it up.
Yeah, and people start going crazy again about nodes, nanometers, codenames of architectures, Ryzen, AMD, Intel, xxxLake,7, 7++, 9++, 10++, 14++, and so on... Meanwhile in Linux world not even Linux is what is was before ("Joe Vennix of Apple Information Security found a significant security vulnerability (CVE-2019-14287) in the Linux sudo utility that could have allowed other users to gain unauthorized administrative (“root”) privileges on a Linux machine.") and starts to get the bad behavior of MS Windows so clearly the end of the world is approaching...
#5720740
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
kakiharaFRS:

1) yes it does as the trickery (a very brief spike of Mhz at idle advertised as a "fix" is a cheat) is directly linked to the fact that AMD advertised "clocks" the same way Intel does except on an Intel the cores run at max speed all the time and AMD 7nm cannot maintain those high clocks 2) 7820x seems to have no problem to overclock around 4.3Ghz either you got a bad chip or it's coming from your setup (last time I checked using an app to oc is garbage you do it from the bios as each mb/memory/cpu is different and Intel app doesn't account for that) 3) those forums are filled of people overjoyed by the fact AMD has "fixed" their clock speed when it hasn't it created a fake spike so that people who only check the "maximum value" on their statistics page are happy, sorry that's not how I work I read graphs of all the cores and they better do what they should
Honestly. Single thread clock is for simpletons. Check how many cores each application you run uses. Then check your regular scenarios where you run multiple applications at same time. Outside of benchmarking focused on single threaded performance, you are looking at 2 or more cores under load. Regular gaming today is 3+ cores. I get that people are angry about ~1% lower clock in 1 loaded core situation. 50MHz ePeen hurts. But in their real use, they will not measure difference in performance nor clock deviation from marketed information 🙂
#5720741
https://forums.guru3d.com/data/avatars/m/270/270233.jpg
schmidtbag:

I said a while ago how Intel was digging themselves deeper into a hole, where for every generation they couldn't get 10nm out the door, they had to keep increasing clock speeds to make it seem like they were making progress. But every time they cranked up the clock speeds, they were making it that much more difficult for 10nm to match the performance. That being said, Kaby Lake is where they screwed up the most. That whole product lineup basically just a Skylake refresh with higher clocks (and higher prices, IIRC). If they just went for what Coffee Lake is, they could've actually lowered clock speeds, since the additional cores were actually an upgrade. But... we don't know what their 10nm node is capable of. Perhaps Coffee Lake lowering boost clocks by a few hundred MHz wouldn't have been enough.
Intel was betting on their process tech, which isn't a bad bet to be honest - I don't think anybody could have imagined the troubles they would have with it. Intel had always been able to rely on their manufacturing advantage to keep them ahead of the pack but those days are likely over; TSMC has grown rich from Apple and Samsung money and I think they will be the leader in process tech going forward (as long as we keep buying new smartphones, they'll have the money to keep extending their lead). I've said this before but AMD couldn't have chosen a better time for a comeback. Intel still seems to believe in 10nm and wants to release desktop products, but I don't think anybody wants it. Even if Ice Lake at 4.5 GHz beats out a 9900K at 5 GHz, I think most Intel fans would prefer the latter for the boasting rights of higher clocks. The average consumer would also be put off by the lower clocks as they know nothing about IPC. Like I said before, if they release 10nm for desktop it will likely be low power chips, similar to what they're doing with mobile 10th gen.
#5720751
https://forums.guru3d.com/data/avatars/m/270/270008.jpg
D3M1G0D:

Intel was betting on their process tech, which isn't a bad bet to be honest - I don't think anybody could have imagined the troubles they would have with it. Intel had always been able to rely on their manufacturing advantage to keep them ahead of the pack but those days are likely over; TSMC has grown rich from Apple and Samsung money and I think they will be the leader in process tech going forward (as long as we keep buying new smartphones, they'll have the money to keep extending their lead). I've said this before but AMD couldn't have chosen a better time for a comeback. Intel still seems to believe in 10nm and wants to release desktop products, but I don't think anybody wants it. Even if Ice Lake at 4.5 GHz beats out a 9900K at 5 GHz, I think most Intel fans would prefer the latter for the boasting rights of higher clocks. The average consumer would also be put off by the lower clocks as they know nothing about IPC. Like I said before, if they release 10nm for desktop it will likely be low power chips, similar to what they're doing with mobile 10th gen.
Intel was way to aggressive with the density they tried to achieve using quad patterning for there 10nm process. TSMC's first go at 7nm was using quad patterning but was not as dense as Intel's 10nm and on top of that it was only some of the up front processes that were at the 7nm feature size. People bag Intel for their failures on 10nm but frankly I think most seem to miss they were trying to push the envelope too far without moving to EUV. I think you will see all the fabs learn from this and take a TSMC like approach where they iterate the process in small steps year on year because these nodes are extremely complex. For example of complexities, just to create the EUV light source it takes a room sized laser firing at tin droplets that are being shot at 100mph.
#5720799
https://forums.guru3d.com/data/avatars/m/197/197287.jpg
kakiharaFRS:

1) yes it does as the trickery (a very brief spike of Mhz at idle advertised as a "fix" is a cheat) is directly linked to the fact that AMD advertised "clocks" the same way Intel does except on an Intel the cores run at max speed all the time and AMD 7nm cannot maintain those high clocks 2) 7820x seems to have no problem to overclock around 4.3Ghz either you got a bad chip or it's coming from your setup (last time I checked using an app to oc is garbage you do it from the bios as each mb/memory/cpu is different and Intel app doesn't account for that) 3) those forums are filled of people overjoyed by the fact AMD has "fixed" their clock speed when it hasn't it created a fake spike so that people who only check the "maximum value" on their statistics page are happy, sorry that's not how I work I read graphs of all the cores and they better do what they should
....Please stop, you're hurting my brain just by reading what you write, let alone the face palms i'm doing may just give me physical brain damage
#5720860
data/avatar/default/avatar06.webp
nevcairiel:

And you don't know what happens next year or even in two years.
Yes we do know how the Zen 3 is from the AMD presentation few weeks ago. No more fractured CCD to 2x CCX. The Zen 3 chip will be a single CCD with 8 cores and 32MB+ shared L3. (now is 16GB per CCX) in 7nm EUV process with 20 million more transistors and better ~20% improvement to perf/power ratio. The chiplets going to be tad smaller also. The abolition of CCX would eliminate a lot of latency that exists currently with estimate 8-10% IPC gain over Zen 2 (along side further changes). Also the I/O goes to 7nm EUV with many more improvements to wiring and faster operations. 2021. Zen 4 with DDR5 & PCIe 5.0. New sockets for server & mainstream. No design specs or concrete information if would be on N7 or 6nm process. 2022 Zen 5 at 5nm, is what we know of only. Again the above are known from AMD. Now rumours have it that the first batch of Zen 3 ES chips are clocking 100-200Mhz more than Zen 2. Also there are speculations that with the reduction of the I/O to 7nm frees up space for a 3rd slot with AMD might decide to use. If it doesn't make a more balanced spread of the 3 chips.
#5720900
https://forums.guru3d.com/data/avatars/m/165/165018.jpg
Fox2232:

I get that people are angry about ~1% lower clock in 1 loaded core situation
I actually haven't heard of anyone(who owns one) being all that angry. It seems to be the Intel users who are "angry". None of us are denying the issues was there and are going to welcome the updates that will fix it. In my case my performance was minimal of course butt it was there and I'm not going to complain about the extra performance.
#5721059
https://forums.guru3d.com/data/avatars/m/250/250418.jpg
The next couple of years are certainly going to be interesting! Intel is almost 5 years behind with their 10nm process, making a decision to ditch it means they're finally taking strategic decisions and trying to pull their head out of their respective rear end.
#5721071
https://forums.guru3d.com/data/avatars/m/270/270008.jpg
Silva:

The next couple of years are certainly going to be interesting! Intel is almost 5 years behind with their 10nm process, making a decision to ditch it means they're finally taking strategic decisions and trying to pull their head out of their respective rear end.
Just so you know Intel is not ditching 10nm for desktops. This is in the article: Meanwhile, intel was very quick to out some words on this to Toms hardware: desktop processors based on the 10 nm silicon fabrication node are still on the company's roadmap. "We continue to make great progress on 10 nm, and our current roadmap of 10 nm products includes desktop," the company said in its communication
#5721089
data/avatar/default/avatar11.webp
Fox2232:

Honestly. Single thread clock is for simpletons. Check how many cores each application you run uses. Then check your regular scenarios where you run multiple applications at same time. Outside of benchmarking focused on single threaded performance, you are looking at 2 or more cores under load. Regular gaming today is 3+ cores. I get that people are angry about ~1% lower clock in 1 loaded core situation. 50MHz ePeen hurts. But in their real use, they will not measure difference in performance nor clock deviation from marketed information 🙂
That's not true. While everything uses multiple threads and hence multiple cores a lot of applications are still bottle-necked by a single thread and hence a single core. It's difficult to see because the scheduler will just bounce that thread around all the cores so none will look particularly high but basically your applications performance is still bound by that thread running flat out. Hence you need enough cores so you are never maxing everything out and there is a bit left over to respond quickly to anything new, but beyond that it is single core performance that matters.
#5721135
https://forums.guru3d.com/data/avatars/m/197/197287.jpg
JamesSneed:

Just so you know Intel is not ditching 10nm for desktops. This is in the article: Meanwhile, intel was very quick to out some words on this to Toms hardware: desktop processors based on the 10 nm silicon fabrication node are still on the company's roadmap. "We continue to make great progress on 10 nm, and our current roadmap of 10 nm products includes desktop," the company said in its communication
The question here in my opinion is what desktop parts? Are they going to release "desktop" parts only in the low TDP section mainly for OEMs? I wouldn't be surprised by this, because at this point, it's clear there is some issue still present on 10nm for high-end processors, otherwise they wouldn't even now be limiting it to laptops. It's typically the other way around: Desktops, highend and servers first, then laptops. So intel denying the rumors, could be correct in the long run, yet the article itself, its purpose, could also be correct, if 10nm never comes to high-end products and servers, but rather AIO systems and low powered OEM systems with possibly a retail release.
#5721140
https://forums.guru3d.com/data/avatars/m/225/225084.jpg
The 3800X is 370 pound, while the the i9900k is 500 pounds. Is it worth an extra 130 pound. Absolutely not in my opinion. Now if Intel lowered the 9900k price to 400 pound then i could do with the bit extra price but @ that price i could get a 3800X and a good mobo for 140ish.
#5721146
data/avatar/default/avatar25.webp
Reddoguk:

The 3800X is 370 pound, while the the i9900k is 500 pounds. Is it worth an extra 130 pound. Absolutely not in my opinion. Now if Intel lowered the 9900k price to 400 pound then i could do with the bit extra price but @ that price i could get a 3800X and a good mobo for 140ish.
9900k was worth it 1 year ago. Now it's worth waiting for the next amd/intel generation. 9900k don't need to lower the prize, because it's the performance leader, due to low latency and high clocks. Best performance always have the extra premium prize 🙂
#5721175
data/avatar/default/avatar34.webp
D3M1G0D:

This makes sense. Releasing 10nm chips on desktop now would mean a significant step back in clocks, which will drive away consumers. Their 14nm process is still competitive and they can coast on it for a while longer (perhaps they'll release some low power or specialty 10nm chips on desktop in the future, but not gaming chips). And the owners of Intel CPUs like to think performance is all about emulators and DOSBOX. It's not 😉 All kidding aside, I bought my Ryzen CPUs for their great value and computing performance. I don't even bother to run benchmark apps at all since theoretical numbers mean nothing - what matters is real world performance, and Ryzen certainly delivers (I can show you my computing numbers if you want 😉).
Join us in Hardware.no F@H team 🙂 Small country, big contributors 😉 https://folding.extremeoverclocking.com/team_list.php
#5722583
https://forums.guru3d.com/data/avatars/m/250/250418.jpg
nizzen:

9900k was worth it 1 year ago. Now it's worth waiting for the next amd/intel generation. 9900k don't need to lower the prize, because it's the performance leader, due to low latency and high clocks. Best performance always have the extra premium prize 🙂
Ya, that extra 10% fps for more 50% in price. Great deal! I just said fack you to Intel and retired my i5 2500k with a R5 2600.
#5722794
https://forums.guru3d.com/data/avatars/m/196/196426.jpg
Silva:

Ya, that extra 10% fps
By comparing overclocked 9900K (power unconstrained 5Ghz) with non-overclocked Zen 2 with older launch bios. What about this: - Give AMD CPU good some good memory that clocks high and push IF speed to that clock ( 3600-3733 ) - Update to 1.0.0.3abba - Set PBO to +200Mhz - Remove TDP limits ( Set to 160W or something instead of the default 65W ) - Use the Overdrive scalar and set it to 5X-7X (when having good cooling), which effectively auto overclocks the AMD CPU (and increases it's power consumption as well unfortunately, so TDP goes out the door) Surprise ! 3700X beats 9900K in most titles (except some heavily Intel optimized ones). 3900X obliterates 9900K in everything I wish reviewers would repeat tests with Zen 2 CPUs after all the findings in last 2 months, because the Launch Day reviews are NO LONGER VALID.
#5724836
https://forums.guru3d.com/data/avatars/m/250/250418.jpg
wavetrex:

I wish reviewers would repeat tests with Zen 2 CPUs after all the findings in last 2 months, because the Launch Day reviews are NO LONGER VALID.
IMHO, I think the reviews should be done out of the box, as it's how most of the products will be used. I know what you'll say: "but Intel cheats with tdp and the 9900k will pull ahead". No, with that crap box cooler it will burn your house instead! Ya, reviewers should not strap an AIO cooler to help Intel when the AMD one just performs as a product.
#5725025
https://forums.guru3d.com/data/avatars/m/196/196426.jpg
Silva:

No, with that crap box cooler it will burn your house instead!
Except that there is no boxed cooler at all on 9900K. Or on the new special edition 9900KS. Well, 9900K(S) are good CPUs of course, but do they worth their money PLUS a good cooler compared to the now-existing and now-improved competition? No, not anymore. Not for the majority of us. 10nm will not change that. So curious what Intel's response will be ... when it will come.
#5725209
https://forums.guru3d.com/data/avatars/m/250/250418.jpg
wavetrex:

10nm will not change that.
Nope, I believe there will not be any 10 nm desktop parts. I forgot they don't include a cooler, my bad. But even if you had a 30€ cooler, the deal isn't great compared to options we have now.

Previous Page

Next Page