News Intel Teases Rocket Lake: Double-Digit IPC Gains, Cypress Cove Architecture

[GetSmart]

Intel Rocket Lake engineering sample on MSI Z590-A PRO-12VO motherboard leak at UserBenchmark https://www.userbenchmark.com/UserRun/34751794

 

[spongiemaster]

Intel Rocket Lake engineering sample on MSI Z590-A PRO-12VO motherboard leak at UserBenchmark https://www.userbenchmark.com/UserRun/34751794

Something to put those scores in context:

Intel Rocket Lake 8 Core & 16 Thread Desktop CPU Benchmarks Leak Out - 4.2 GHz ES Sample Destroys 5.1 GHz Core i7-10700K

The Intel 11th Gen Rocket Lake 8 Core ES Desktop CPU benchmarks have leaked which shows it much faster than its 10th gen predecessor.

wccftech.com

"As you can see in the benchmark table above, the Intel Rocket Lake Desktop CPU is up to 21% faster than its predecessor, the Core i7-10700K which has the same amount of cores & threads within single-core benchmarks. At the same time, the Core i7-10700K has a clock advantage of 5.1GHz which is 21% faster than the 4.2 GHz. Given these numbers, a 5 GHz+ Rocket Lake CPU is going to destroy the Core i7-10700K in single-threaded workloads. The CPU was also 18% faster than the Core i9-10900K which has a clock speed advantage of 26% at 5.3 GHz. "

Color me skeptical on this one. I'm not questioning the score itself, but what are the odds the clock speed of the Rocket Lake CPU is being properly detected by this benchmark? Running at 4.2Ghz, its single core score is 21% faster than a 10700K running at 5.1Ghz? I would think a 21% advantage would be the upper bound vs Comet Lake when running at the same clock speed based on Ice Lake benchmarks. No way it would be that much faster with a 17.5% clock deficit.

 

[TerryLaze]

Color me skeptical on this one. I'm not questioning the score itself, but what are the odds the clock speed of the Rocket Lake CPU is being properly detected by this benchmark? Running at 4.2Ghz, its single core score is 21% faster than a 10700K running at 5.1Ghz? I would think a 21% advantage would be the upper bound vs Comet Lake when running at the same clock speed based on Ice Lake benchmarks. No way it would be that much faster with a 17.5% clock deficit.

It all depends on what the benchmarks are measuring, as said by others artificial benchmarks will give you a theoretical difference that might not be achievable in actual software because actual software is there to do a certain job while artificial benchmarks just fill out everything they can without needing to do any actual work.
RL has 1.5 to 2 times more cache which could be a big part of the difference for benchmarks but might not factor into any daily software.

 

[LordVile]

Advantage of what? Wider architecture? No. You can run the exact same code and the instruction scheduler will shuffle things around to put the extra execution units to use regardless of what the code may have originally been optimized for. It may not be quite as efficient as code optimized specifically for how a specific CPU's instructions are distributed across its 10-wide architecture vs another CPU's 8-wide one but it'll get most of the way there, especially when SMT is being leveraged to give the scheduler a whole extra thread to to help it find something for all execution units to work on.

The whole point of having out-of-order execution in CPUs is to decouple low-level internals that can vary drastically between CPUs from software development so developers don't need dedicated code paths for every CPU model and variant thereof in existence to achieve remotely decent performance.

Just ignore what I said and answer the point you want to then.

 

[shady28]

From Anandtech
(Note that in the single threaded test, the power limits ultimately should not apply because one core should not consume all the power of the chip. For the Tiger Lake processor, because this is a nominal 15 W TDP part with a 50 W turbo, this actually does go above the power limit with one core active, as it scores 554. As a result, the 50 W mode with a 28 W TDP was used and scores 595. This is more akin to a desktop processor anyway.)
This is from their Zen 3 preview and talking about AMD's stated Cinebench score.

Cinebench. That's a single use case, and one that has favored AMD for many years. It’s mot a good measure of ipc. Ipc should be a cross section of many workloads,

But regardless, Tom’s just tested the XP’s 13 with TL. When they ran cinebemch, they stated its average core freq was 2,7Ghz.

People are just now taking off the blinders and, if the benchmarks are right, raw performance of a 5Ghz cpu using a similar core are likely to be more than enough to unseat zen 3.

 

[GetSmart]

Seems there is a big increase in performance in Intel's newer generation architectures. 🤔

Has anyone seen the recent Intel Jasper Lake leaks? Here is Intel Pentium N6000 at https://browser.geekbench.com/v4/cpu/15844740 and https://browser.geekbench.com/v5/cpu/4447101 The results are more like Intel's mobile Core-series than that of any Intel Atom lineages (including the current Intel Gemini Lake Refesh)

 

[InvalidError]

Seems there is a big increase in performance in Intel's newer generation architectures. 🤔

Well, Intel has been working on their architectural updates for five years, it would be somewhat of a shocker if Skylake's successor (Comet Lake is still based on Skylake) didn't deliver any IPC gains. As Ice Lake has shown on laptops, Sunny Cove has ~18% IPC gains, so it would make sense that Cypress Cove (Sunny for desktop/14nm) delivered something similar.

 

[shady28]

TGL has been benched on Phoronix at 45/65W power limits.

As I expected, the delta vs stock is massive. In some cases it is more than twice as fast - generally 30%+. Keep in mind this is still limited to 4.7ghz and 45w sustained, 65w peak. This bodes well for rocket lake with desktop class power limits.

https://openbenchmarking.org/result/2010296-FI-2010170FI14

 

[jeremyj_83]

Well, in that comparison it doesn't state how long the chip can maintain that boost. The lower power chips won't maintain full turbo on a single core very long , and of course on all core they drop off very quickly.

I guarantee you that TL didn't maintain 4.7Ghz on one core the entire test. Probably more like 4.5-4.6 for 28 seconds. A higher power chip - a desktop chip - will maintain higher boost for longer. Power does matter because you don't actually get or maintain max speeds without it, and a 15w / 28w doesn't have much power.

I think people here are really reaching to explain away a massive difference in power / performance.

In Ananadtech's Ryzen 5000 series review, they have tested the IPC across different CPUs in SPEC and then normalized the performance by GHz. It included Ice Lake and Tiger Lake in the comparison. In SPEC2017 the Ryzen has a 13.2% higher INT IPC and a 7.77% higher FP IPC that the Ice Lake/Tiger Lake CPUs.

 

[shady28]

In Ananadtech's Ryzen 5000 series review, they have tested the IPC across different CPUs in SPEC and then normalized the performance by GHz. It included Ice Lake and Tiger Lake in the comparison. In SPEC2017 the Ryzen has a 13.2% higher INT IPC and a 7.77% higher FP IPC that the Ice Lake/Tiger Lake CPUs.

In Anand's review, they crippled the Intel systems with DDR4-2933 and ran the Zen 3 at DDR4-3200.

 

[jeremyj_83]

In Anand's review, they crippled the Intel systems with DDR4-2933 and ran the Zen 3 at DDR4-3200.

They are not crippling the Intel system by running 2933MHz RAM. That is the official spec for Comet Lake i5 10600K and higher. While you can try faster RAM, you are not guaranteed to have it work. Same with Ryzen being run at its official spec of 3200MHz. If they were to run the Intel at 3200MHz, they would be artificially inflating the score since that is a RAM OC. Ian Cutress has a YouTube Video which explains why he tests CPUs with RAM only at official spec. The numbers for Tiger Lake seem to come from their review from mid September.

 

[shady28]

They are not crippling the Intel system by running 2933MHz RAM. That is the official spec for Comet Lake i5 10600K and higher. While you can try faster RAM, you are not guaranteed to have it work. Same with Ryzen being run at its official spec of 3200MHz. If they were to run the Intel at 3200MHz, they would be artificially inflating the score since that is a RAM OC. Ian Cutress has a YouTube Video which explains why he tests CPUs with RAM only at official spec. The numbers for Tiger Lake seem to come from their review from mid September.

Ryzen is spec'd to run JEDEC DDR4-3200 -> at CL22. If it's at lower than CL22, it's overclocked.

What was the CL on the Ryzen RAM?

 

[jeremyj_83]

Ryzen is spec'd to run JEDEC DDR4-3200 -> at CL22. If it's at lower than CL22, it's overclocked.

What was the CL on the Ryzen RAM?

What is the CL spec for the Intel CPU? If both are running at say CL 16 for the Ryzen and CL 15 for the Intel then they both have the same then they overclocked by the same amount.

 

[shady28]

What is the CL spec for the Intel CPU? If both are running at say CL 16 for the Ryzen and CL 15 for the Intel then they both have the same then they overclocked by the same amount.

You're reaching. Overclocking is overclocking, cherry picking the type of OC you can or can't have and in so doing cripple one platform and help the other is intellectually dishonest

JEDEC spec for DDR4-2933 and DDR4-3200 are below.

 

[jeremyj_83]

You're reaching. Overclocking is overclocking, cherry picking the type of OC you can or can't have and in so doing cripple one platform and help the other is intellectually dishonest

JEDEC spec for DDR4-2933 and DDR4-3200 are below.

"As per our processor testing policy, we take a premium category motherboard suitable for the socket, and equip the system with a suitable amount of memory running at the manufacturer's maximum supported frequency. This is also typically run at JEDEC subtimings where possible." They didn't put in the CL timings in there, however, at this point in time CL 16 is standard on 3200MHz and CL15 on 2933MHz. It is also possible they were run at looser timings as well.

Your argument is it is alright to OC one CPU to give it better performance but keep the other running at official spec. While having an impact on performance, CL doesn't do nearly as much for performance now as it did back in 1st or 2nd Gen DDR.

 

[shady28]

"As per our processor testing policy, we take a premium category motherboard suitable for the socket, and equip the system with a suitable amount of memory running at the manufacturer's maximum supported frequency. This is also typically run at JEDEC subtimings where possible." They didn't put in the CL timings in there, however, at this point in time CL 16 is standard on 3200MHz and CL15 on 2933MHz. It is also possible they were run at looser timings as well.

Your argument is it is alright to OC one CPU to give it better performance but keep the other running at official spec. While having an impact on performance, CL doesn't do nearly as much for performance now as it did back in 1st or 2nd Gen DDR.

Memory speed has a massive effect on performance. They aren't being transparent. I prefer the sites that are transparent.

 

[spongiemaster]

Memory speed has a massive effect on performance. They aren't being transparent. I prefer the sites that are transparent.

It doesn't for Intel. On Intel, it's more important to run dual channel than the speed of the memory. Dropping from 3200 to 2933 will make almost no difference.

 

[shady28]

It doesn't for Intel. On Intel, it's more important to run dual channel than the speed of the memory. Dropping from 3200 to 2933 will make almost no difference.

You should read this article, what you just said is completely false. Use Google Translate, or just go look at the graph on the 2nd page.

Core i9-9900K vs. Ryzen 9 3900X mit RAM-OC in Spielen

Wie hoch ist die Spieleleistung wirklich? Core i9-9900K und Ryzen 9 3900X treten mit übertakteten Samsung-B-Dies im CPU-Limit in Spielen an.

www.computerbase.de

 

[spongiemaster]

You should read this article, what you just said is completely false. Use Google Translate, or just go look at the graph on the 2nd page.

Core i9-9900K vs. Ryzen 9 3900X mit RAM-OC in Spielen

Wie hoch ist die Spieleleistung wirklich? Core i9-9900K und Ryzen 9 3900X treten mit übertakteten Samsung-B-Dies im CPU-Limit in Spielen an.

www.computerbase.de

Not going to bother. Almost every major site has done a memory speed comparison article at some point. The only reason to pick a non-English article is because that's all you could find to promote your narrative. For things like office productivity and gaming ram speed means little. Some applications like encoding or some rendering applications will benefit from faster memory, but those aren't common home user applications. Even in those scenarios, the difference between 2933 and 3200 will be minimal.

 

[TJ Hooker]

Not going to bother. Almost every major site has done a memory speed comparison article at some point. The only reason to pick a non-English article is because that's all you could find to promote your narrative. For things like office productivity and gaming ram speed means little. Some applications like encoding or some rendering applications will benefit from faster memory, but those aren't common home user applications. Even in those scenarios, the difference between 2933 and 3200 will be minimal.

Coffee lake memory scaling: https://www.tomshardware.com/reviews/best-ram-speed,5951-6.html
For comparison, Pinnacle Ridge memory scaling: https://www.tomshardware.com/review...le-ridge,6064-6.html?region-switch=1604949757 and Matisse: https://www.tomshardware.com/reviews/amd-ryzen-3000-best-memory-timings,6310.html

Looks like Intel benefits as much (often more) from going from 2800 to 3200 MHz (and faster memory in general) as AMD does. Although in either case the improvement going from 2933 to 3200 obviously isn't going to be huge.

 

[TJ Hooker]

Ryzen is spec'd to run JEDEC DDR4-3200 -> at CL22. If it's at lower than CL22, it's overclocked.

Where do you see Ryzen explicitly specced for JEDEC 3200 MHz?

I'm under the impression that finding 3200 MHz sticks that actually have a JEDEC 3200 profile is rare (first ones apparently only came out a little over a year ago, while 3200 MHz DDR4 has obviously been ubiquitous for much longer), I would guess the situation is similar for JEDEC 2933. If 2933/3200 MHz kits actually running JEDEC timings for their rated speeds are the exception rather than the norm, why use them to benchmark?

Sticking to the rated frequency without necessarily minding the official JEDEC timings is also relevant because Intel non-Z boards limit frequency but not timings. I.e you could run 2933 MHz CL 15 on an H70, but not 3200 MHz CL 20. So it would seem that Intel considers exceeding the frequency to be an OC, but not reducing timings below those of JEDEC. Motherboard manufacturers also only denote "OC" based on memory speed, not timings. So it would seem that neither the DIMM manufacturer, the mobo manufacturer, or the CPU manufacturer necessarily consider it an OC just because the timings may be tighter than JEDEC standards.

Running both CPUs at their rated RAM speed, and both with comparable timings (that are similarly lower than JEDEC timings) seems fair to me.

 

[shady28]

Where do you see Ryzen explicitly specced for JEDEC 3200 MHz?

I'm under the impression that finding 3200 MHz sticks that actually have a JEDEC 3200 profile is rare (first ones apparently only came out a little over a year ago, while 3200 MHz DDR4 has obviously been ubiquitous for much longer), I would guess the situation is similar for JEDEC 2933.

Sticking to the rated frequency without necessarily minding the official JEDEC timings is also relevant because Intel non-Z boards limit frequency but not timings. I.e you could run 2933 MHz CL 15 on an H70, but not 3200 MHz CL 20. So it would seem that Intel considers exceeding the frequency to be an OC, but not reducing timings below those of JEDEC.

Running both CPUs at their rated RAM speed, and both with comparable timings (that are similarly lower than JEDEC timings) seems fair to me.

People who think the JEDEC 3200 memory specification is whatever you want it to be, do not know what a standard is.

Or what we all really know, that people want to put their finger on the scales in reviews - which means this forum and all these reviews are complete and absolute rubbish and a waste of time.

If a sites says it is going to use JEDEC standards, use the freaking standard. Not "Oh, I'll use the clock speed but not the CL, the tRAS, not the tCAS, not the voltage spec, etc etc etc". That's just plain stupid, and clearly deceptive.

When you look at a board, or RAM, it usually says certified to run at JEDEC speeds. Most boards will default to that spec. Anything else is overclocked. The B450 for example is not certified to run anything outside of DDR4-2667 JEDEC max.

The 5XX series was the first that allowed DDR4-3200 - JEDEC DDR4-3200.

 

[InvalidError]

If a sites says it is going to use JEDEC standards, use the freaking standard.

The only thing JEDEC preescribes is the WORST CASE for memory rated for a given MT/s class. Any memory that has equal or lower latency for a given transfer rate while also meeting all other specs is still within specs.

It makes no sense to run 3200-22 when 3200-16-16-28 is budget-friendly and plentiful.

 

[jeremyj_83]

The only thing JEDEC preescribes is the WORST CASE for memory rated for a given MT/s class. Any memory that has equal or lower latency for a given transfer rate while also meeting all other specs is still within specs.

It makes no sense to run 3200-22 when 3200-16-16-28 is budget-friendly and plentiful.

It isn't worth it to argue with Shady28. S/he cannot accept the fact that on and even playing field there is evidence that shows even Intel's newest (Tiger Lake) cannot beat Zen 3 in IPC. I would assume the Intel reference laptop is going to use RAM that will put the Tiger Lake in the best light so therefore it had LPDDR4-4266MHz RAM. What is ironic is Shady28 was fighting for Comet Lake's numbers to be higher even though we know Comet Lake has a lower IPC than Tiger Lake.

 

[shady28]

The only thing JEDEC preescribes is the WORST CASE for memory rated for a given MT/s class. Any memory that has equal or lower latency for a given transfer rate while also meeting all other specs is still within specs.

It makes no sense to run 3200-22 when 3200-16-16-28 is budget-friendly and plentiful.

"While meeting all other specs is still within specs" - that's what I'm talking about. It is not in spec, because it's not in spec, period. No amount of fanboy help in this forum will change that fact. That 3200-16-16-28 is "budget friendly and plentiful" is irrelevant to the spec. If you want to use "budget friendly and plentiful" as a "standard" for parts in a review then do it on all platforms.