Discussion Intel current and future Lakes & Rapids thread

Page 522 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
SisoftSandra bench, hm memory bandwidth why it is so lower on Alder Lake.

Cause it is not measuring mem bw after all? DDR4 3200 is ~50GB/s affair, so that already tells us this is measuring some other "quantity".
Benchmark is probably old and getting defeated by large L3 caches and since ZEN3 has two L3 cache domains of 32MB, more of that extra "bandwidth" is leaking into score.

Still with DDR5 latency being as horrible as it is, it will impact bandwidth as well. Not in the most synthetic benchmarks like Aida64, but things like Intel's Memory Latency checker will paint different picture - latency x bandwith product is gonna be bad and that will hurt real world ( and get mitigated at same time by DDR5 extra channels and larger L3 cache ).
 

Doug S

Platinum Member
Feb 8, 2020
2,509
4,113
136
DDR generational transitions are curious. It's not like the manufacturing process of memory ICs gets worse.


It happens because people compare enthusiast DDRn with conservatively timed DDRn+1 that gets targeted at server/workstation customers first. The early adopters of a new DDR standard care about system reliability far more than performance, and conservatively timed DDR5 (which is all there is now) is a nice performance bump over conservatively timed DDR4.
 

dullard

Elite Member
May 21, 2001
25,214
3,632
126
Still with DDR5 latency being as horrible as it is, it will impact bandwidth as well. Not in the most synthetic benchmarks like Aida64, but things like Intel's Memory Latency checker will paint different picture - latency x bandwith product is gonna be bad and that will hurt real world ( and get mitigated at same time by DDR5 extra channels and larger L3 cache ).
Isn't the best (the 'A' specification) DDR5 at a ~14 ns latency? That is the same as the best DDR4 latency. The marketing latency number is higher (CL), but the actual time taken is the same since you double the number of clocks (CL is twice as high) but have twice as fast of clocks and it all cancels out.
 

jpiniero

Lifer
Oct 1, 2010
14,845
5,457
136
Isn't the best (the 'A' specification) DDR5 at a ~14 ns latency? That is the same as the best DDR4 latency. The marketing latency number is higher (CL), but the actual time taken is the same since you double the number of clocks (CL is twice as high) but have twice as fast of clocks and it all cancels out.

You can do a lot better than DDR5 JEDEC on CL. Like at 4800, DDR5 JEDEC is 34 for A and even just a quick look on Newegg there's DDR4 under 20. I was kind of assuming that there would be DDR5 available with tighter timings but perhaps not at launch?
 

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
Isn't the best (the 'A' specification) DDR5 at a ~14 ns latency? That is the same as the best DDR4 latency. The marketing latency number is higher (CL), but the actual time taken is the same since you double the number of clocks (CL is twice as high) but have twice as fast of clocks and it all cancels out.

I think rock bottom is 3200CL18 => 11.25ns first word

Is there DDR5 4800CL27 out there to match it? Nope, ~15ns first word stuff with god knows how terrible secondaries and tertiaries.
 

Abwx

Lifer
Apr 2, 2011
11,172
3,869
136
I don't think this is a source, it's a copy and paste from other known sources. Look they don't have more SKU infos, the P/E clock speeds of the K-SKUs were already known, they don't know more than this. The non K core/thread counts might be nonsense there. 12700 listed as 8+4. On Geekbench it's 8+0: https://browser.geekbench.com/v5/cpu/9528015

Clock speed seems off as well.

They could have disabled the E cores in the GB submission.
Also the Sandra test says 32 threads instead of 24, so unless E cores have SMT look like there s still some correct CPUID implementations to be done on softs.
 

dullard

Elite Member
May 21, 2001
25,214
3,632
126
I think rock bottom is 3200CL18 => 11.25ns first word

Is there DDR5 4800CL27 out there to match it? Nope, ~15ns first word stuff with god knows how terrible secondaries and tertiaries.
So, you are comparing the very fastest latency DDR4 (not in JEDEC specification) to the very worst allowable (within JEDEC specification) for DDR5? And then you are concluding that the not yet released DDR5 memory won't eventually match or exceed?
 

mikk

Diamond Member
May 15, 2012
4,175
2,211
136
I think rock bottom is 3200CL18 => 11.25ns first word

Is there DDR5 4800CL27 out there to match it? Nope, ~15ns first word stuff with god knows how terrible secondaries and tertiaries.


DDR4-3200 CL22 is the JEDEC standard and the rock bottom, almost all of Intels TGL-H and TGL-U notebooks with DDR4-3200 using the JEDEC standard with CL22.
 

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
DDR4-3200 CL22 is the JEDEC standard and the rock bottom, almost all of Intels TGL-H and TGL-U notebooks with DDR4-3200 using the JEDEC standard with CL22.


On desktop, at 3200 speed the slowest i can buy is CL18, that is not enthusiast, junk grade memory without cooling.
Enthusiast stuff is 8ns first word.

Rest assured Anandtech will test JEDEC speeds and timings and those slow 15ns timings will hurt Alder Lake hard. Intel has warped back to 2015 with memory latency and they will pay for it in "stock" tests
 

AMDK11

Senior member
Jul 15, 2019
347
243
116
Granted, no one knows what app this slide is based on but ADL-S is ~17% ahead of Zen 3 and slightly less than that against RKL-S in single core uplift in this particular scenario, Maybe we can trace the app by the difference between RKL-S and Zen 3. GB5? CBxx?

View attachment 49878
Not! In this test, Single-thread GoldenCove is 18.3% faster than CypressCove and 21.2% faster than Zen3.
CypressCove AVX512
GoldenCove + Gracemont AVX2
Zen3 AVX2
 
Last edited:

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Could the next-gen mont core be seeing action already?

Edit: Seems it is: "SSHPmont."
 
Last edited:

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Not! In this test, Single-thread GoldenCove is 18.3% faster than CypressCove and 21.5% faster than Zen3.
CypressCove AVX512
GoldenCove + Gracemont AVX2
Zen3 AVX2
I'm eyeballing the graph, and both Zen3 and Cypress Cove are past the 4/5ths mark (1600), so where are you getting your figures from?
 

moonbogg

Lifer
Jan 8, 2011
10,637
3,095
136
That i7 model is looking really damn MOONBOGGY right now! That's the Intel Moonbogg edition right there. If these slides are accurate, then Intel 12th gen has me pretty damn excited. If true, then this feels like a return to glory for Intel. It's exciting and awesome and it's exactly what I want in my rig. $430 for the i7 even sounds just about perfect for such a chip. Everything looks amazing! I still have my X99 platform plus my 10700K platform and I could sell them both and get 12th gen for a pretty low cost!
Will I buy it though? Sadly not. Why? No GPUs. /thread.
 
Reactions: lightmanek

Hulk

Diamond Member
Oct 9, 1999
4,377
2,256
136
A while back in thread I wrote my prediction that ADL 12900K will perform somewhere between 5900X and 5950X. Having a look at early benches I think that looks like it may be the case. If they price 12900K at 5900X levels they will have competitive parts. Of course we still need to see actual power measurements but even if they are high that won't affect desktop too much. For mobile on the other hand it could be a big deal.
 

mikk

Diamond Member
May 15, 2012
4,175
2,211
136
On desktop, at 3200 speed the slowest i can buy is CL18, that is not enthusiast, junk grade memory without cooling.
Enthusiast stuff is 8ns first word.

Rest assured Anandtech will test JEDEC speeds and timings and those slow 15ns timings will hurt Alder Lake hard. Intel has warped back to 2015 with memory latency and they will pay for it in "stock" tests


You are wrong. DDR4-3200 CL22 is a JEDEC standard and in fact it is the second biggest CL group in the desktop market after CL16: https://geizhals.eu/?cat=ramddr3&xf=5015_3200~5828_DDR4

CL18 is not a JEDEC standard unlike CL22 (or CL40 on DDR5-4800). Of course nobody is using DDR4-3200 CL22 in a deskop test, this is a different matter. That's why comparing DDR4 OC against DDR5 JEDEC might not be the best idea.
 
Last edited:
Reactions: Tlh97 and ryan20fun

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
CL18 is not a JEDEC standard unlike CL22 (or CL40 on DDR5-4800). Of course nobody is using DDR4-3200 CL22 in a deskop test, this is a different matter. That's why comparing DDR4 OC against DDR5 JEDEC might not be the best idea.

I am not wrong about being able to buy 3200CL18 as slowest on my country market, but You are also right that JEDEC is CL20 and CL22 for 3200 speed.

What i am not wrong tho, is that Anandtechs of the world will test JEDEC and that sweet 15ns first word latency will bite Alder Lake performance. Add some unknowns about gear modes and penalty for early DDR5 will be high.
50% more bandwidth "stock" is nice, but only few workloads are even starved by it, mostly Linpack style calculations. While latency plays a role in every access coming from/to memory.


I think the question for desktop enthusiasts will be: will there be decent Z690 DDR4 overclocking motherboards, or will those be limited to DDR5?
 

DrMrLordX

Lifer
Apr 27, 2000
21,813
11,168
136
Weren't there some hiccups testing Rocket Lake-S early on due to the funky stock memory controller settings?
 

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
Weren't there some hiccups testing Rocket Lake-S early on due to the funky stock memory controller settings?

11700K as previewed has GEAR2 mode as default that costs extra latency + it was on funky prerelease uCode. Still initial DDR5 4800 JEDEC is gonna have horrible latency and that will show versus DDR4 opponents. Similar to Skylake DDR4 2133 release where it was baaad in gaming, even if same core served for years and had amazing memory scaling.

My statement is - DDR4 will provide said "scaling" for Alder Lake from day 1, while having comparable bandwith and better performance and price*

* as long as Intel does not shoot their feet with segmentation marketing machine gun this time.
 

eek2121

Diamond Member
Aug 2, 2005
3,053
4,281
136
I am not wrong about being able to buy 3200CL18 as slowest on my country market, but You are also right that JEDEC is CL20 and CL22 for 3200 speed.

What i am not wrong tho, is that Anandtechs of the world will test JEDEC and that sweet 15ns first word latency will bite Alder Lake performance. Add some unknowns about gear modes and penalty for early DDR5 will be high.
50% more bandwidth "stock" is nice, but only few workloads are even starved by it, mostly Linpack style calculations. While latency plays a role in every access coming from/to memory.


I think the question for desktop enthusiasts will be: will there be decent Z690 DDR4 overclocking motherboards, or will those be limited to DDR5?

Technically when you run CL16 or CL18, or any other non-JEDEC setting, you are violating warranty.

There will be better speeds and timings when DDR5 releases, and at least one company has a DDR5 8400 SKU planned with timings on par woth the best DDR4 modules.
Weren't there some hiccups testing Rocket Lake-S early on due to the funky stock memory controller settings?
Yes, and ADL-S supports gear 2 AND gear 4. I expect we will see better performance once the chip is properly tested.
 

lobz

Platinum Member
Feb 10, 2017
2,057
2,856
136
There will be better speeds and timings when DDR5 releases, and at least one company has a DDR5 8400 SKU planned with timings on par woth the best DDR4 modules.
Hopefully the decent reviewers will count the price of such modules in the price-performance conclusions as well.
 
Reactions: Joe NYC

Bouowmx

Golden Member
Nov 13, 2016
1,142
550
146
(Stale news) but from the list of upcoming ASRock 600 series boards
The Z690-C/D5 model may indicate DDR5 will be put on low-end Z boards too. The current Z590-C is an unbranded version of the Phantom Gaming 4.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |