News New Rocket Lake Chip Matches Zen 3 in Single Core Performance

[Admin]

A new Core i9 Rocket Lake chip was spotted on Geekbench with just a 35W TDP, and manages to pull out some amazing single-core performance numbers.

New Rocket Lake Chip Matches Zen 3 in Single Core Performance : Read more

 

[ginthegit]

Nice, Just synthetic benchmarks that are based on Imaginary facts. Of course we must take ones word for it.

This kind of report just makes the Intel fanboi wait longer for a product they will be ultimately be disappointed with. If the TDP goes down and the Multi processor benchmarks suffer, then usually it equates (in a smaller way) to a lesser loss in single thread too.

 

[CerianK]

I am anxious for a truly competitive product to take some of the pressure off of the AMD CPU supply chain.
Hopefully Rocket Lake will help... but, I guess, Alder Lake will not.

 

[thGe17]

That's illogical. Alder Lake with its eight big Golden Cove cores will have even more performance than Rocket Lake (and therefore also Zen3). The only remaining factor is scaling, because as far as it seems: ADL will provide a maximum of eight big cores at best (sufficient for the general/mainstream market including gaming, but it will not compete with the 16 core CPUs, but that's still a niche).

*) Its architecture is two gens ahead and the clock will be also high. Already Tiger Lake H parts will reach a 5,0 GHz boost in 10nm++, therefore this shouldn't be a problem for ADL, which is going to use 10nm+++ (or in Intel's marketing speech 10nm Enhanced SuperFin).

 

[jimmysmitty]

Nice, Just synthetic benchmarks that are based on Imaginary facts. Of course we must take ones word for it.

This kind of report just makes the Intel fanboi wait longer for a product they will be ultimately be disappointed with. If the TDP goes down and the Multi processor benchmarks suffer, then usually it equates (in a smaller way) to a lesser loss in single thread too.

Not always. The lower TDP will limit multi core more than single core by far as it takes more power to keep more cores clocked higher than one core clocked really high.

I can see this performance metric pretty easily. Process is important but so is the uArch. Tweaks and design changes can still boost performance even on an older process.

 

[CerianK]

That's illogical. Alder Lake with its eight big Golden Cove cores will have even more performance than Rocket Lake (and therefore also Zen3).

Sure, but any AL variant with mixed big/bigger cores will need to have OS and/or Application support for choosing best performing cores, unless Intel has some hardware-level optimization baked-in (which I have not heard of). Bigger-only variants should be well accepted in the desktop space, which should help even out supply, but pricing will decide.

 

[Makaveli]

Here are my 5800X Geekbench scores.

 

[ginthegit]

Not always. The lower TDP will limit multi core more than single core by far as it takes more power to keep more cores clocked higher than one core clocked really high.

I can see this performance metric pretty easily. Process is important but so is the uArch. Tweaks and design changes can still boost performance even on an older process.

Then it would make the point in having this Chip as a multi core chip pointless. Why not just make a Chip that is Smaller, only two core and clock the hell out of them. Nope they won't Why!?

1. Because Clock speed may Improve the IPC, but also causes other negative impacts as Thermals are not the only thing that impact the Operation on the clocks of the Chip. This is evident in the Memory clocks on RAM with RAS CAS and TAS. If you run a component over a certain speed, it takes longer to stabilize and wait states are used to allow this settling time. Its one of the reasons they stopped Raising clock speeds over the 5GigHertz. Quantum and Nano Physics starts to take over the dynamics of the Chips, which is part of the reason that Intel struggled with the 10nm node.

2. It usually takes a change in materials and interconnects on the refresh, without changing the Die to allow the Stability on the increased speeds. This is why the Rare Earth materials are so important.
AMD changed its interconnects and memory issues using this (including the solder type) and resolved the INTEL advatage in IPC over them in Single thread operations.

3. It also depends on Software optimizations that were so evident in the 2000's to propel sales for Intel over AMD.

4. Often the faster refresh involves lesser or more simplified OPcodes on the Chip . This usually increases the speed of the chip at the price of the security elements (hence Spectre bugs et al.

5. Companies also often are filled with AMD or INTEL fanboi's that optimise codes for benchmarks. Though Intel is less likely to do it, I image it is now considering that AMD wont care too much considering its Own IPC advantages at stock.

 

[TechyInAZ]

Sure, but any AL variant with mixed big/bigger cores will need to have OS and/or Application support for choosing best performing cores, unless Intel has some hardware-level optimization baked-in (which I have not heard of). Bigger-only variants should be well accepted in the desktop space, which should help even out supply, but pricing will decide.

We'll have to see how this plays out, Intel could go the easy route and give both the big Alder Lake cores and the smaller Atom (skylake IPC) Gracemount cores full access to Windows 10 scheduler.

Then windows would treat both units like a NUMA node or a dual-socket system, albeit with some modifications to make Windows 10 insure high workloads happen on the Alder cores and not the Atoms.

 

[cryoburner]

The real question is whether the chip is actually drawing 35 watts during that test. Considering Intel bases their TDP on base clocks these days, not accounting for boost, the answer is almost assuredly "no". I doubt it will draw much less power than their current CPUs when running at its boost clocks, and it could easily draw more.

For example, the existing i9-10900T and i7-10700T are also supposedly "35 watt" processors, but both have 123 watt power limits when boosting...

Official Intel 10th Gen Comet Lake-S CPU Power Limits Revealed - PC Perspective

Official Intel 10th Gen Comet Lake-S CPU Power Limits Revealed With any processor, and Intel CPUs in particular, power limits are a critical aspect to

pcper.com

They can maintain that boost up to 28 seconds, and on most motherboards will ignore that time limit by default. So it's more accurate to consider those processors as "123 watt" parts, and that will likely apply to this processor as well. And that makes the entire "power efficiency" side of this story wrong. If the chip were configured to actually adhere to a 35 watt power limit, it would be getting a fraction of that performance.

 

[waltc3]

Kind of interesting that Cinebench 20 single-threaded bench scores--along with many, many other synthetics--are obviously avoided... 'nuff said. At least Cinebench is reflective of real-world rendering work--with the Cinebench renderer.

 

[VforV]

Amazing performance from Intel - for a whopping 10 second burst of power, before it throttles down like a champ!

 

[ctgunner]

I am anxious for a truly competitive product to take some of the pressure off of the AMD CPU supply chain.
Hopefully Rocket Lake will help... but, I guess, Alder Lake will not.

You might check your local Microcenter if you have one, I live in a relatively affluent area in the Midwest and they've had 5000 series in stock for over a week now.

 

[CerianK]

You might check your local Microcenter if you have one, I live in a relatively affluent area in the Midwest and they've had 5000 series in stock for over a week now.

The closest to me is a 2 hour drive... it might be 6 months, or more, before I find myself in that direction.

 

[Jesse_20]

Tom's NY Resolution: To use "Take this with a pinch of salt" in EVERY article published this year.

 

[thGe17]

Sure, but any AL variant with mixed big/bigger cores will need to have OS and/or Application support for choosing best performing cores, unless Intel has some hardware-level optimization baked-in (which I have not heard of). Bigger-only variants should be well accepted in the desktop space, which should help even out supply, but pricing will decide.

Doesn't really matter because
a) Intel and Microsoft are working on this for some time (since Lakefield).
b) Microsoft already knows this topic from Windows X.
c) For example there will also be 8+0 ADL CPUs (which still will easily outperform RKL).
Obviously both companies are working on this and are going to make this "mainstream", therefore I see no problem with this, especially when considering that special adjustments also had to be implemented for the Zen architecture.

 

[cryoburner]

At least Cinebench is reflective of real-world rendering work--with the Cinebench renderer.

The program is actually called Cinema 4D though.

You might check your local Microcenter if you have one, I live in a relatively affluent area in the Midwest and they've had 5000 series in stock for over a week now.

Most people's "local Microcenter" tends to be several hours away though, assuming they even live in the US. They only have 25 locations, after all. Of course, if they had lots of locations, their limited stock would be spread thinner, and they would likely be sold out, just like other retailers. >_>

 

[ingtar33]

quoting geekbench?

yikes. hardly trustworthy.

 

[watzupken]

The article says "Power efficient rockets!". Why so? The single core performance is surely better, but that does not mean that it is more power efficient. The high end chips are still burning through at least 250W, and I am not sure if Rocket Lake will be significantly faster than the 10 cores Comet Lake.

 

[dalek1234]

The real question is whether the chip is actually drawing 35 watts during that test. Considering Intel bases their TDP on base clocks these days, not accounting for boost, the answer is almost assuredly "no". ......

Exactly. With this total BS benchmark, Intel just his the lowest of lows.

 

[mitch074]

If one considers the way Intel calculates TDP, a 35W Intel chip should compete with a 65W AMD chip.
Wait - that's what the 5800X is. And it gets pwned.

 

[Makaveli]

If one considers the way Intel calculates TDP, a 35W Intel chip should compete with a 65W AMD chip.
Wait - that's what the 5800X is. And it gets pwned.

The 125 Watt version of Rocket Lake S will offer about 5% more ST and be equal in MT. And not compete with the 5900X or 5950X so I guess these two will "pwn"the 8 core rocket lake chip as you put it.

Even writing that felt weird as i'm not use to speaking in "Fan boy"

 

[mitch074]

The 125 Watt version of Rocket Lake S will offer about 5% more ST and be equal in MT. And not compete with the 5900X or 5950X so I guess these two will "pwn"the 8 core rocket lake chip as you put it.

Even writing that felt weird as i'm not use to speaking in "Fan boy"

125W Intel means 220-250W 'real' TDP... AMD usually announce TDP that is closer to reality.
It also means that Intel needs twice the electricity AMD needs for the same amount of computing power, so when one takes that into account it means that a binned low power top of the range chip from Intel's future product line barely matches an AMD mid-high tier product that has been out for months now while using twice as much power.
Ouch.

 

[TerryLaze]

125W Intel means 220-250W 'real' TDP... AMD usually announce TDP that is closer to reality.
It also means that Intel needs twice the electricity AMD needs for the same amount of computing power, so when one takes that into account it means that a binned low power top of the range chip from Intel's future product line barely matches an AMD mid-high tier product that has been out for months now while using twice as much power.
Ouch.

The stock 10900 uses the same amount of power to do the same amount of work compared to the 5800x, the stock 11900 will probably have a ~20% better IPC in cinebench, it will also draw more power since it will have more transistors so we will have to see how that pans out.
If you want higher clocks than that then yes you are going to loose efficiency.
The top highest possible power draw, while a real number, has nothing to do with how much power any normal software draws.
https://www.techpowerup.com/review/amd-ryzen-9-5900x/19.html

 

[cryoburner]

Exactly. With this total BS benchmark, Intel just his the lowest of lows.

I don't think this was a benchmark that Intel was advertising, just something that appeared in an online benchmark database from someone who tested one. They do advertise the chip as having a far lower TDP than its actual power use under load, but that's been the case with their processors for years. The "low-power" spin this article did was the only thing really wrong here, as the writer should know better than to take Intel's TDPs at face value.