Someone stole and reviewed an Intel i7-11700K engineering sample, a.k.a. Rocket Lake

mongeese

Posts: 482   +109
Staff member
Something to look forward to: The 11700K could perform similarly to current flagships, or it could perform better if Intel refines it ahead of release. We don't generally condone reviewing processors ahead of the embargo, as it isn't fair on other publications, but it's pretty funny when it's for revenge. Furthermore, what some of these numbers show, particularly the power draw and temperature tests is that his engineering sample is only partially showing what the final CPU will offer.

Four years ago, Intel closed up shop in Romania. They've continued to sell their processors there but without a local office, their media support has been woeful, says lab501, a local tech publication. Of the 21 Intel processors lab501 has reviewed since being left in the lurch, 5 came from Intel's UK office (all arriving after the review embargo had lifted), 12 came from their friends in the industry, one they purchased themselves, and 3 more came from anonymous industry sources.

Those three chips, an 8700K, an 8600K, and a 10980XE, caused some trouble for Intel. Once lab501 stopped receiving review samples from Intel, they stopped being beholden to the review embargoes. When the chips they've reviewed have been provided by their industry friends, they've respected the embargo for their friends' sake. But when the chips have arrived without Intel's permission, they've published their reviews ahead of time – much to Intel's chagrin.

Now, lab501 has tested an 11700K engineering sample before the processor could be officially revealed. Admittedly, there's not much left for Intel to reveal. It's generally known that the processor will have 8 cores and 16 threads, will use a variant of the Cypress Cove architecture, and have a boost clock of around 5 GHz. But there's been fiery debate surrounding the processor's performance.

On one hand, it's got the same core/thread configuration as the preceding 10700K, and similar clocks. On the other hand, it should have an IPC (instruction per clock) advantage over its predecessor, but that could come at the cost of power efficiency. In lab501's testing, it was found that, on average, everything balanced out roughly how you'd expect.

A precursor to the results of the testing: engineering samples and final release samples are physically different in many ways, so these results aren't indicative of final performance. Future microcode revisions and BIOS updates will also alter performance.

But without further ado:

In synthetic multi-core testing, the 11700K performed better than its predecessor but worse than the 10-core 10900K, as expected. In synthetic single-core testing, it usually outperformed all previous Intel processors to roughly match the newest AMD processors.

In games, it performed similarly to other flagship processors like the 10900K and 5950X. Its performance was very distinct from those other processors– oftentimes, it had oddly high or low minimum framerates, and in some games it struggled unexpectedly. Perhaps explained because this is again, an engineering sample. Overall, though, it wasn't better or worse.

To summarize, the 11700K could perform similarly to current flagships, or it could perform better if Intel refines it ahead of release… Not an exciting conclusion, I'll admit, but what did you expect?

The power draw and temperature tests prove just how unfinished this processor is: under a Prime95 load, it simultaneously consumed an unprecedented amount of power for an octa-core processor while running as cool as a hexa-core. This would seem to indicate that the BIOS and clock table were written too conservatively (in other words, it could've been clock higher).

Basically, we've learned nothing, except that reviewers don't like being screwed over. Stay tuned for our review featuring accurate data and meaningful conclusions.

Permalink to story.

 

Kosmoz

Posts: 137   +222
We already knew this and will be proven again when official reviews come out: RKL will have at best +5% better performance than Zen 3, most likely less on average and also depending on the game, but less efficient on all other counts.

So unless you are a die hard intel fan or you absolutely need those 5-10fps (at 1080p) for that particular game, the performance will be the same as with Zen 3 in games that don't use more than 8c/16t, in everything else Zen will smash the ground with this RKL, or how I like to call it Skippy Lake.

Oh and not before long AMD will bring Zen3+ in some months time, probably before Alder Lake, this year.
 

Irata

Posts: 1,349   +2,159
I am certainly hoping the final CPU does not use even more power.

Temperature wise, that could be due to the larger die (more area = better heat dissipation) but also due to die thinning.

Either way, we‘ll find out soon enough.
 

GeforcerFX

Posts: 1,006   +473
I will hold judgement until final release. I am impressed withy tiger lake chip in my laptop but that's on 10+nm not the backport.

This "reviewer" prob just did a crap ton of damage to his friends that receive review samples. Either they will have a strict no share clause or just get removed from the review sample list.
 

Shadowboxer

Posts: 1,431   +1,037
She’s going to be a power guzzler. But in a desktop CPU that isn’t a big deal for enthusiasts. If the performance is as expected it will be down to deals on the day and motherboard configs to separate AMD and Intel. Temps could be high too but once again, if you have a tower you’ll have more than enough room to get sufficient cooling in it.

One advantage Intel’s Z590 boards seems to have is thunderbolt. Oh and no PCH fan. I really hate that my X570 has a fan on it. I look forward to removing the graphics card and finding a nice big dust bunny in there in a few months.

One thing to note is that if this does match Ryzen 5xxx then it’s bloody impressive. As it’s a (5th gen? I’ve lost count) 14nm part matching a 2nd gen 7nm part. Although that doesn’t mean much in terms of value.
 

Austinturner

Posts: 132   +129
I am certainly hoping the final CPU does not use even more power.

Temperature wise, that could be due to the larger die (more area = better heat dissipation) but also due to die thinning.

Either way, we‘ll find out soon enough.
Why does power consumption under load matter to you out of interest? Is it just because you need a bigger cpu cooler and case fans?

I see power getting mentioned a few times, I just don’t really see it as much of an issue?
 

HardReset

Posts: 1,059   +655
One advantage Intel’s Z590 boards seems to have is thunderbolt. Oh and no PCH fan. I really hate that my X570 has a fan on it. I look forward to removing the graphics card and finding a nice big dust bunny in there in a few months.

One thing to note is that if this does match Ryzen 5xxx then it’s bloody impressive. As it’s a (5th gen? I’ve lost count) 14nm part matching a 2nd gen 7nm part. Although that doesn’t mean much in terms of value.

PCH fan is not needed on x570 boards so that's not advantage. There are several fanless x570 boards available.

Nothing impressive there. For example Nvidia could easily match their 3000-series cards (8nm) with just overclocking 2000-series cards (12nm) and probably making them bigger if just overclocking is not enough. That would be very "impressive". We just have to ignore power consumption...

Why does power consumption under load matter to you out of interest? Is it just because you need a bigger cpu cooler and case fans?

I see power getting mentioned a few times, I just don’t really see it as much of an issue?

You need better (=more expensive) cooler, better (=more expensive) motherboard VRM, better (=more expensive) case cooling, it's much harder to keep system silent...
 

psycros

Posts: 3,384   +3,855
PCH fan is not needed on x570 boards so that's not advantage. There are several fanless x570 boards available.

Nothing impressive there. For example Nvidia could easily match their 3000-series cards (8nm) with just overclocking 2000-series cards (12nm) and probably making them bigger if just overclocking is not enough. That would be very "impressive". We just have to ignore power consumption...



You need better (=more expensive) cooler, better (=more expensive) motherboard VRM, better (=more expensive) case cooling, it's much harder to keep system silent...

Also heats up a small living area which isn't so bad in winter but sucks any other time.
 

Austinturner

Posts: 132   +129
PCH fan is not needed on x570 boards so that's not advantage. There are several fanless x570 boards available.

Nothing impressive there. For example Nvidia could easily match their 3000-series cards (8nm) with just overclocking 2000-series cards (12nm) and probably making them bigger if just overclocking is not enough. That would be very "impressive". We just have to ignore power consumption...



You need better (=more expensive) cooler, better (=more expensive) motherboard VRM, better (=more expensive) case cooling, it's much harder to keep system silent...
Thanks, I’m surprised the difference is that significant, but I can see why you would be unhappy with it.
 
She’s going to be a power guzzler. But in a desktop CPU that isn’t a big deal for enthusiasts. If the performance is as expected it will be down to deals on the day and motherboard configs to separate AMD and Intel. Temps could be high too but once again, if you have a tower you’ll have more than enough room to get sufficient cooling in it.

One advantage Intel’s Z590 boards seems to have is thunderbolt. Oh and no PCH fan. I really hate that my X570 has a fan on it. I look forward to removing the graphics card and finding a nice big dust bunny in there in a few months.

One thing to note is that if this does match Ryzen 5xxx then it’s bloody impressive. As it’s a (5th gen? I’ve lost count) 14nm part matching a 2nd gen 7nm part. Although that doesn’t mean much in terms of value.

Just as an addendum to what you wrote, while it's true that the overwhelming number of X570 boards have an active chipset fan, some do offer a reduced speed or "fan-stop" setting which can keep it from spooling up under many conditions. That said, fan speed and noise can and does vary greatly from model to model, and things like GPU selection and case airflow can play a role in how bad the resulting noise is. One person may have a great outcome while someone else with the same board could have issues.

Z590 having no chipset fan is definitely a good thing, especially for users building a "set and forget" style system. Do note that some high end Z590 boards WILL include a fan, or fans, on the VRM stage. The Z590 Taichi, for example, has a 40mm fan on the VRM block.

One other thing to remember is that Z590 is functionally closer to AMD's B550 platform, as the chipset itself is still only PCI Express 3.0. Those B550 boards are also fanless, and provide the same CPU attached PCI Express 4.0 connection to the primary PCI Express slot (or slots, for a few boards supporting x8/x8 or x8/x4/x4 split), along with the primary NVMe M.2 drive. Z590 does have an advantage over B550 for the DMI link (CPU<->Chipset), due to the wider x8 lane (8GB/s) interconnect, vs x4 (4GB/s).

Unambiguous Thunderbolt support is absolutely a strength on Intel platforms, and it's frustrating to see few vendors on the AM4 side dip their toes into it. There are ways to get connectivity with an add-in card, but otherwise it's essentially only ASRock and Gigabyte making any moves in that space.

Rocket Lake doesn't really bring anything especially exciting to the table in terms of a single standout feature, but it looks like a really well rounded cap to the platform as they transition into Alder Lake and DDR5/Gen5+. If they price it right, it'll do well. The extra connectivity for the DMI link as well as dedicated NVMe M.2 connection are genuinely useful quality of life updates that go a long way to making it a more viable long term platform for things like content creation, even if they're not really a feature that's easily quantified or looked at in most reviews.

Competition is good. :)
 

Tom Yum

Posts: 85   +205
"The power draw and temperature tests prove just how unfinished this processor is: under a Prime95 load, it simultaneously consumed an unprecedented amount of power for an octa-core processor while running as cool as a hexa-core. This would seem to indicate that the BIOS and clock table were written too conservatively (in other words, it could've been clock higher)"

Given the power draw is measured separately (ie wall draw) but the CPU temperature is self-reported by the motherboard, it is more likely that the motherboard isn't reporting the CPU temperature correctly and it is hotter than indicated. Not sure why you would think the clockspeed is lower than it could be, the power draw shows what is actually happening (if it is clocked too low, then imagine how bad power draw would be at full speed).
 
We already knew this and will be proven again when official reviews come out: RKL will have at best +5% better performance than Zen 3, most likely less on average and also depending on the game, but less efficient on all other counts.

So unless you are a die hard intel fan or you absolutely need those 5-10fps (at 1080p) for that particular game, the performance will be the same as with Zen 3 in games that don't use more than 8c/16t, in everything else Zen will smash the ground with this RKL, or how I like to call it Skippy Lake.

Oh and not before long AMD will bring Zen3+ in some months time, probably before Alder Lake, this year.

What universe are you living in? What evidence do you have?
 

Irata

Posts: 1,349   +2,159
Why does power consumption under load matter to you out of interest? Is it just because you need a bigger cpu cooler and case fans?

I see power getting mentioned a few times, I just don’t really see it as much of an issue?
It matters for several reasons:

- inefficiency: if a product needs more power to do the same - or in the case of mt workloads less - it‘s an inefficient product which I do not like. Now if said product can do something the other can‘t that might be a redeeming factor
- heat generated: more power used means more heat energy that needs to be evacuated from the case. It does not matter if the CPU itself runs cooler due to having a larger surface area - the heat still needs to be evacuated from the case and is affecting all other components in it
- if the CPU needs more power and produces more heat, so will the mainboard, PSU and cooling system (there are always conversion losses, I.e. neither the mainboard circuitry nor the PSU run at 100% efficiency)
- higher cost: this is linked to all of the above points. If you need a beefier PSU, mainboard, cooling system and case fans (or more of them), this adds to cost.

If I get noticeably higher performance for the higher power consumption, this is a trade off I would be willing to accept. But if it‘s essentially the same or in some cases less, than no.
 

tancabean

Posts: 23   +31
Why does power consumption under load matter to you out of interest? Is it just because you need a bigger cpu cooler and case fans?

I see power getting mentioned a few times, I just don’t really see it as much of an issue?

Noise. I disabled half the cores on my 5950x and the cooler is still annoyingly loud at times.
 

Lionvibez

Posts: 2,256   +1,746
One thing to note is that if this does match Ryzen 5xxx then it’s bloody impressive. As it’s a (5th gen? I’ve lost count) 14nm part matching a 2nd gen 7nm part. Although that doesn’t mean much in terms of value.

Its not really that impressive this is new arch that was back ported.

If they were doing this still on skylake then sure. I would expect better performance from a new arch.

And the chipset fan thing is overated.

I've been on x570 since Dec 2019 and I've yet to hear that fan.
 

Mr Majestyk

Posts: 668   +579
An utterly useless product that wasted resources that should have been spent on Alder Lake and the GPU group. Wow, who gives a flying toss if it scores a few more fps in 1080 laming without a care in the world for power efficiency.

Bring on Alder Lake and let's forget all about 14nm++++++++++++++++++++. Maybe then we'll see real competition again for AMD.
 

Wereweeb

Posts: 14   +33
"It's really impressive that Intel increased performance by using larger cores that both cost more silicon and draw much more energy"

Think about that notion for a few seconds.

Zen was built from the ground up to be as efficient as possible in multi-threaded applications, while Intel always excelled at single-threaded workloads like games.

At launch (I.e. prior to software optimization), Ryzen already was not just more efficient and generally competitive, but also actually performed better than Intel in a few HEDT tasks, even on the same core counts and while using GloFo's 14nm. See Anandtechs first benchmarks of Zen.

Intel being tied with AMD on the one good thing Intel always excelled at isn't impressive. Quite the contrary.

Intel is only really impressive in applications which they built task-specific accelerators for. AMD is still in the stone age in that regard, since they're still working on their general compute.

There's no way to spin this positively, all in all they're behind. Inferior. They will have to use all that dosh, IP and personnel that they've been collecting in order to catch up.

Xe is a much better example of those efforts, surpassing AMD in iGPU performance and wishing an architecture and ecossystem into being could be considered impressive.

And as for "the quality of their 14nm process", that is just a symptom of years spent refining the same process, and the same architecture for that process. Watch the situation reverse when they get to 7nm and have to compete with a fresh node not only against a mature TSMC 5nm node, but also the 5nm+, 4nm, etc...

If AMD doesn't introduce a new platform and the rumoured 7nm I/O die with Warhol (I.e. USB4, future DDR5 compatibility), Intel will partially surpass them with Alder Lake, and stay on top until AM5/Zen 4. But I don't see them being competitive in HEDT, and for servers it always depends on the task.

Leaving the cores enabled when no software takes advantage of them would be a waste. I’ll enable them when that changes.

Please don't tell me you're a g*mer who bought a 16 core processor for "future-proofing".
 
I am still not convinced by the arguments deriding processors that produce high temperatures.

Inefficiency - sure, it's not the preferred option, but if I get a Lambo, I don't care much that it guzzles gas.

Needing other components to be more expensive? If I'm buying an expensive processor I don't intend to skimp with cheaper components elsewhere anyway.

The power draw is not really going to be higher than processors in past 'eras' were either, from what I can see.

At the end of the day, plenty of us are overclocking our processors anyway, introducing inefficiency and costs to satisfy the same desire of better performance.

I accept that it's not 'impressive' in technological terms, but in practical terms, that may not matter. If I can take the performance crown in its price bracket, that's progress.
 

danwat1234

Posts: 17   +2
Why can't we have any actual facts in the article? What is the unprecedented power consumption? Temperature? Scores?
 

Theinsanegamer

Posts: 2,380   +3,458
It lines up with the rumors sites like wccftech have been posting, in that RKL is better with IPC but is the same or slightly slower then comet lake due to lower clock speeds and high power usage.

I remember saying during the 10900k launch that intel had pushed 14nm too far, there was nowhere left to go and while it was impressive they managed to squeeze yet more performance out of their 14nm node, it was going to turn into a netburst style "damn the consequences fire the reactors" trade-off soon. RKL would seem, from all the leaks so far, to be proving this theory true.

Rumor also has it that l2 cache was halved to make it work on 14nm, which means we may still not be getting tiger lake level performance either.
Leaving the cores enabled when no software takes advantage of them would be a waste. I’ll enable them when that changes.
ERM, power gating and dynamic clocking have been things for 10+ years. Those extra cores won't pull a noticeable amount of power if nothing takes advantage of them. You might save 3-5 watts, at most, and that will make no real descernable impact.

What you need is either a better fan or a better heatsink. Yes the stock one is loud, they always are. Get a proper third party heatsink with 120mm fans. I mean the 5950x pulls a whopping 120-130 watt in stress tests, you can cool that off with a hyper 212 evo