Discussion i7-11700K preliminary results

Page 22 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

AdamK47

Lifer
Oct 9, 1999
15,544
3,246
136
The Gamers Nexus thumbnail is low level pandering. It's what Youtubers do to drive clicks.

The thumbnail preys on the emotions of the AMD side who are revelling in the jab made. It preys on the emotions of the Intel side with outrage in the belittling.

Either way, the predator got his prey as is evidence here.
 

moonbogg

Lifer
Jan 8, 2011
10,637
3,095
136
I mean how long did it take from inception to a product? 2 years?

That's pretty much a straight backport with no optimization.

It's a backport to an older, worse process, and they rushed it doing so. No wonder it's so bad. Also tells us how important process is and it'll get even more important as shrinks get harder.

They have a presentation about Rocketlake and the team that made it happen, and it sounded like they were pretty proud about it. Not sure if they should have bothered, because there's nothing to be proud about this product.

Yea sure you got it done, but it's questionable if it's better. It's a "we did it cause we can" chip.

Good thing it didn't end up worse. Imagine this chip on mobile.

This is an interesting train of though. Maybe from an engineering perspective it was an excellent exercise to take something designed for a completely different process and do the unthinkable by transforming it to work with 14nm. Maybe when someone mentioned it at first, people laughed inside and thought "um, yeah. No way". Not only did they do it (possibly learning a lot in the process) but they also managed to create a viable product from it. From this perspective perhaps they brought out the champagne? I wonder what other engineering teams think about what Intel was able to do with this. Maybe that's where the glory is and we're all complaining about gaming benchmarks for a product that should have been impossible? Is backporting a common thing? Is it easy and no big deal? Or should this chip have ended up in the trash bin but instead they made a viable product?
 

Mopetar

Diamond Member
Jan 31, 2011
8,104
6,737
136
Maybe when someone mentioned it at first, people laughed inside and thought "um, yeah. No way". Not only did they do it (possibly learning a lot in the process) but they also managed to create a viable product from it.

Hopefully they never have to put that knowledge to use again. If we have to make any more jokes about 14nm+++++++++++++ we'll wear out our keyboards.

It's kind of interesting from an academic perspective, but it's not the type of thing anyone wants to have to do. The real question is that if Intel knew exactly how it would turn out when they made the choice to go ahead with this idea if they'd still go through with it or if they'd do something else instead.
 

Panino Manino

Senior member
Jan 28, 2017
869
1,117
136

"but he needs to learn how to respect people who are doing seriously hard work that he would be totally incapable to do himself, except joining as a marketing person."

The funny thing about this tweet is Francois was a marketing person at Intel.

FP is funny, he blocked me again because I showed how GN had done the same to AMD plenty of times in the past, even very recently as THIS MONTH! I wonder if he blocked everyone who responded on that thread, and they think we are the ones being stubborn?
No surprises here, what surprised me was the sudden change on tune. I don't get it, last time I had seen his account he said this:

2021-03-21 09.49.46 twitter.com 09b5533dcc98.jpg


It's a new "era", right? AMD is changed and a true competitor and even leader in some areas. Intel changed, and is changing as we could see from the news that the new CEO gave this week, but some people are still adapting to this new reality. Even though they recognize and even praise what AMD is achieving, still, Intel is Intel, they expect that every new Intel launch will crush the competition because this is how the word works.
And this is one of the biggest challenges AMD faces. "The better product don't always wins", right? Even if AMD keeps putting in the market better and cheaper products, many will never buy it because it's from AMD.
 
  • Like
Reactions: Tlh97 and CHADBOGA

dmens

Platinum Member
Mar 18, 2005
2,274
959
136
FP is funny, he blocked me again because I showed how GN had done the same to AMD plenty of times in the past, even very recently as THIS MONTH! I wonder if he blocked everyone who responded on that thread, and they think we are the ones being stubborn?
No surprises here, what surprised me was the sudden change on tune. I don't get it, last time I had seen his account he said this:

View attachment 41902


It's a new "era", right? AMD is changed and a true competitor and even leader in some areas. Intel changed, and is changing as we could see from the news that the new CEO gave this week, but some people are still adapting to this new reality. Even though they recognize and even praise what AMD is achieving, still, Intel is Intel, they expect that every new Intel launch will crush the competition because this is how the word works.
And this is one of the biggest challenges AMD faces. "The better product don't always wins", right? Even if AMD keeps putting in the market better and cheaper products, many will never buy it because it's from AMD.

Just ignore the guy. I have seen his resume. The cover letter was longer than the resume. 🚨
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
Is backporting a common thing? Is it easy and no big deal? Or should this chip have ended up in the trash bin but instead they made a viable product?

If Rocketlake was using a core meant by 14nm and it took lot of planning and thought, then it might have ended up better.

Timeframe suggests it couldn't be. They just ported Sunny Cove and made necessary adjustments to fit on 14nm, likely with repercussions.
 

zir_blazer

Golden Member
Jun 6, 2013
1,192
486
136
I'm actually wondering if the original 10nm Cannonlake (Skylake improved shrink) could have been backported back to 14nm with better results than Cypress Cove. Not sure what was the expected IPC increase on Cannonlake or if the only thing going on for it was adding AVX-512, but Cypress Cove left a lot to be desired by not being substantially faster than what is essencially 2015 Skylake with some Hardware mitigations for Spectre and Meltdown.
This is similar to how AMD could have continued with K10 cores in 32nm like those of the first APU, Llano, instead of going Bulldozer...
 

moinmoin

Diamond Member
Jun 1, 2017
5,064
8,032
136
I'm actually wondering if the original 10nm Cannonlake (Skylake improved shrink) could have been backported back to 14nm with better results than Cypress Cove.
Technically Rocketlake is the exact counterpart to Cannonlake (maybe that's why both names sound like they are related): Whereas Cannonlake is pretty much a straight port of Skylake to 10nm with only minor improvements at best, Rocketlake is pretty much a straight port of Tiger Lake to 14nm with only minor adaptions at best. Like a reverse-tick in pre-Skylake Intel parlance.
 

naukkis

Senior member
Jun 5, 2002
895
773
136
Cannonlake wasn't an option, it barely moved the needle IPC wise as Anandtech observed by running Cannonlake and Kabylake at 2.2Ghz (fixed).

Few percent IPC uplift with AVX512 - at about same size as Skylake. So Intel could have get AVX512, keep 10 or more cores and have plenty of L4 cache for chip as large as Rocketlake - there is at least possibility that it could have much better performing option - at least for gamers.
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,693
136
This is similar to how AMD could have continued with K10 cores in 32nm like those of the first APU, Llano, instead of going Bulldozer...

I doubt that was possible. The last generation of K10 used on Llano APUs were good in their own right, but also really showed the ultimate limit of the K8 architecture. AMD was barely able to get the K10.5 core over 3GHz, and neither were overclockers. A 3GHz 6C 32nm K10.5 CPU would have been a good offer, but the low frequency would likely have hampered it. Especially compared to Sandy Bridge, which did 5.2-5.3GHz on air cooling.
 

zir_blazer

Golden Member
Jun 6, 2013
1,192
486
136
I doubt that was possible. The last generation of K10 used on Llano APUs were good in their own right, but also really showed the ultimate limit of the K8 architecture. AMD was barely able to get the K10.5 core over 3GHz, and neither were overclockers. A 3GHz 6C 32nm K10.5 CPU would have been a good offer, but the low frequency would likely have hampered it. Especially compared to Sandy Bridge, which did 5.2-5.3GHz on air cooling.
Wasn't Llano made in a manufacturing process, design, whatever, that favoured high density to priorize the GPU side of the APU at the cost of the CPU clocks? Then what we saw in Llano was a nerfed CPU. If it was designed as a CPU only, I suppose it should have been at least equal or better than 45nm Thuban. Your comment pretty much reminds me when the tech industry blamed the Prescott issues on the high leakage of the 90nm process even though both Intel Dothan and AMD K8 Winchester showed improvements in all areas compared to their predecessors, so some people reached the conclusion than what sucked was the Prescott design and not the 90nm process itself.
Also, Llano was delayed by like one or two years. Seems like it was problematic for AMD to get the GPU running on the same die than the CPU.
 
  • Like
Reactions: lightmanek

Topweasel

Diamond Member
Oct 19, 2000
5,437
1,659
136
Honestly this is just a gen later than I thought we should see this. CometLake should have seen our first registerable drop in per core performance. But Ill give Intel kudos (with my usual problem with their outright abuse of TDP) for pulling out all the stops in thinning out the HS and hard capping certain boost modes to maximizing general boost modes to keep their benchmark scores high. But realistically there were only going to stay at the top still using 14nm so far. Specially as Margin driven they are and their desire to cap die size to prioritize server chip manufacturing. Personally I think at this point in time de-prioritizing core count is a massive miss. The extra cores extend their useful lifespan, allow for more defined segmentation. They probably felt that desktop sales would still be pushed by SC performance and felt that their projected IPC increase would outpace AMD. But Steve isn't wrong, outside actual availability, they are going to be missing any real killer reason to purchase their higher end chips, cept maybe quicksync.
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,693
136
Wasn't Llano made in a manufacturing process, design, whatever, that favoured high density to priorize the GPU side of the APU at the cost of the CPU clocks? Then what we saw in Llano was a nerfed CPU.

Far as I remember, it was the other way round. AMD apparently had trouble with the IGP being built on a CPU prioritizing process, and using SOI. But I can't remember where I read that, so I may be wrong.

I think the main frequency limiter wasn't so much the process, but the rather large (for the time at least) L2 cache.

Your comment pretty much reminds me when the tech industry blamed the Prescott issues on the high leakage of the 90nm process even though both Intel Dothan and AMD K8 Winchester showed improvements in all areas compared to their predecessors, so some people reached the conclusion than what sucked was the Prescott design and not the 90nm process itself.

The Prescott design sucked. Simple as that. Proof? Look at what Dothan achieved on the same process. Especially on desktop (f.x. using the CT479 adaptor). It's not even funny. 115W Prescott matched by a 37W mobile chip.
 

zir_blazer

Golden Member
Jun 6, 2013
1,192
486
136
The Prescott design sucked. Simple as that. Proof? Look at what Dothan achieved on the same process. Especially on desktop (f.x. using the CT479 adaptor). It's not even funny. 115W Prescott matched by a 37W mobile chip.
I never said that Prescott didn't sucked. What I said is how the tech industry justified Prescott power issues due to the high leakage of the 90nm process, then arguing about how it was going to become even worse in lower nodes (I don't recall all those issues materializing at all, it was business as usual). Dothan and Winchester showed that 90nm itself worked fine, leaving the previous high leakage statements as excuses to try to justify Prescott being horrible. Thus, what I said is that I don't really believe than a 32nm K10 has any reason to clock lower than a 45nm one, and that Llano is a bad example of that due to the compromises made to have the GPU on the same die.
 

Hitman928

Diamond Member
Apr 15, 2012
6,123
10,530
136
I never said that Prescott didn't sucked. What I said is how the tech industry justified Prescott power issues due to the high leakage of the 90nm process, then arguing about how it was going to become even worse in lower nodes (I don't recall all those issues materializing at all, it was business as usual). Dothan and Winchester showed that 90nm itself worked fine, leaving the previous high leakage statements as excuses to try to justify Prescott being horrible. Thus, what I said is that I don't really believe than a 32nm K10 has any reason to clock lower than a 45nm one, and that Llano is a bad example of that due to the compromises made to have the GPU on the same die.

Leakage power was an increasing problem until we got to finfets. The 20 nm bulk nodes were pretty much ignored because the leakage at that node became too high. Moving to finfets by and large alleviated the leakage issue. Using FD-SOI also mitigates the leakage issue for planar processes (e.g. GF 22 nm FD-SOI). Obviously the P4 architecture had major issues outside of leakage, but the leakage issue was real and increased until the shift away from planar bulk processes at lower nodes.
 

naukkis

Senior member
Jun 5, 2002
895
773
136
Far as I remember, it was the other way round. AMD apparently had trouble with the IGP being built on a CPU prioritizing process, and using SOI. But I can't remember where I read that, so I may be wrong.

Liano was build on same 32nm SOI-process as Bulldozer and Piledriver. Up to 5GHz dozers, barely above 3GHz K10.5. I have had both, A6-3670 Liano and Athlon 750K Trinity - and Trinity was nearly double as fast. Liano's L2 wasn't reason for low clocks - 45nm Athlon2-chips have same size L2 and clocked fine. Old design was just not suitable for newer process.
 
  • Like
Reactions: lightmanek

Mopetar

Diamond Member
Jan 31, 2011
8,104
6,737
136
I think there was some agreement that there were issues with the 32 nm GF process. Even Bulldozer (which was pretty much Prescott 2.0) was banking on being able to achieve higher clocks.

The original Bulldozer CPUs capped out slightly above 4 GHz. Sandy Bridge could get close to 5 GHz despite not being designed to target such speeds. The 32nm Intel node was just that much better.

AMD eventually managed to beat Intel on clock speed once GF got a few of the kinks out, but the MHz wars were pretty much over by that point and AMD still couldn't get anywhere near the clock speed they would have needed to be competitive.
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,693
136
I never said that Prescott didn't sucked. What I said is how the tech industry justified Prescott power issues due to the high leakage of the 90nm process, then arguing about how it was going to become even worse in lower nodes (I don't recall all those issues materializing at all, it was business as usual). Dothan and Winchester showed that 90nm itself worked fine, leaving the previous high leakage statements as excuses to try to justify Prescott being horrible.

Then I misread you. I think we're talking about the same thing after all.

Thus, what I said is that I don't really believe than a 32nm K10 has any reason to clock lower than a 45nm one, and that Llano is a bad example of that due to the compromises made to have the GPU on the same die.
Liano was build on same 32nm SOI-process as Bulldozer and Piledriver. Up to 5GHz dozers, barely above 3GHz K10.5. I have had both, A6-3670 Liano and Athlon 750K Trinity - and Trinity was nearly double as fast. Liano's L2 wasn't reason for low clocks - 45nm Athlon2-chips have same size L2 and clocked fine. Old design was just not suitable for newer process.

The 32nm K10.5 core used in Llano had double the L2 cache compared to 45nm K10 Thuban/Deneb (1024KB vs 512KB per core), and apparently it didn't scale very well past 3GHz.

BD could do 5GHz, but power consumption skyrocketed. I imagine something similar happening with K10.5, just at lower frequency if they pushed it that hard. Mind the 32nm SOI process itself was very good, my A12-6800K does 4.4GHz absolutely reliable to this day.

I do agree a 6C 32nm K10.5 Thuban successor with L3 cache would have been a decent performer for the time. Way better then 1st gen Bulldozer. Though that's a very low bar to get over. Problem was Intel hit it out of the park with Sandy, so AMD would only have been in a slightly better position then what actually happened. They'd have needed a Zen-like design eventually anyway.
 
  • Like
Reactions: Magic Carpet

zir_blazer

Golden Member
Jun 6, 2013
1,192
486
136
Oh my. From AnandTech today review:

Platform Stability: Not Complete
It is worth noting that in our testing we had some issues with platform stability with our Core i9 processor. Personally, across two boards and several BIOS revisions, I would experience BSODs in high memory use cases. Gavin, our motherboard editor, was seeing lockups during game tests with his Core i9 on one motherboard, but it worked perfectly with a second. We’ve heard about issues of other press seeing lockups, with one person going through three motherboards to find stability. Conversations with an OEM showcased they had a number of instability issues running at default settings with their Core i9 processors.

The exact nature of these issues is unknown. One of my systems refused to post with 4x32 GB of memory, only with 2x32 GB of memory. Some of our peers that we’ve spoken to have had zero problems with any of their systems. For us, our Core i7 and Core i5 were absolutely fine. I have a second Core i9 processor here which is going through stability tests as this review goes live, and it seems to be working so far, which might point that it is a silicon/BIOS issue, not a memory issue.

Edit: As I was writing this, the second Core i9 crashed and restarted to desktop.

We spoke to Intel about the problem, and they acknowledged our information, stating:

We are aware of these reports and actively trying to reproduce these issues for further debugging.

Some motherboard vendors are only today putting out updated BIOSes for Intel’s new turbo technology, indicating that (as with most launches) there’s a variety of capability out there. Seeing some of the comments from other press in their reviews today, we’re sure this isn’t an isolated incident; however we do expect this issue to be solved.

There goes the only market advantage that Intel usually has, launching on Day One with platforms that are usually more mature than AMD counterparts. Top bins Rocket Lake reminds me of the Pentium 3 Coppermine 1.13 GHz.

I actually expect than Intel could end up dialing back a bit the Turbo speeds and maybe actually capping power consumption since I don't trust long term stability of a piece of silicon that when doing AVX-512 sucks 300W and thermal throttles. It is living at the edge.
 

Makaveli

Diamond Member
Feb 8, 2002
4,804
1,268
136
Oh my. From AnandTech today review:



There goes the only market advantage that Intel usually has, launching on Day One with platforms that are usually more mature than AMD counterparts. Top bins Rocket Lake reminds me of the Pentium 3 Coppermine 1.13 GHz.

I actually expect than Intel could end up dialing back a bit the Turbo speeds and maybe actually capping power consumption since I don't trust long term stability of a piece of silicon that when doing AVX-512 sucks 300W and thermal throttles. It is living at the edge.

Thats was because they have been launching the same thing since 2015 just refined. Now with RL being a new Arch they don't have that luxury anymore.
 
  • Like
Reactions: Tlh97 and scannall