Discussion i7-11700K preliminary results

[GunsMadeAmericaFree]

early benchmarks

more benchmarks

Just thought I would provide these so folks could draw their own conclusions. Evidently the processors have been getting sold for a few days now in Germany.

 

[Timorous]

Performance results of this CPU are all over the place. What are people doing wrong?

Latency sensitivity. It seems a good upgrade over skylake in non latency sensitive workloads but is a regression in latency sensitive tasks like gaming. Maybe other testers are using tuned memory which helps but then you really should test tuned ram for all the platforms. I am sure GN / HUB will do some memory tuning testing and like for like comparisons around lunch day.

 

[Shmee]

Wow this is not looking good at all. Now it may be too early to tell performance for sure, but based on prelimary tests, Intel should have just launched Comet lake with PCIe 4, and stuck with that, or do a comet lake refresh with PCIe 4. This change in architecture and drop in cores is no bueno.

I was tempted on the 10900k, but decided to stick with 3700X and X570 at the time due to PCIe 4. Glad I did now with Zen 3 having great performance, even though I don't have a 5900X yet, I at least have a 5800X. My advice now for anyone looking to a new Intel platform, would be to wait for Alder lake. Now if you don't need newest, you could always look at the used market. Or you could try to buy/build a Zen 2 or 3 based system...

 

[thesmokingman]

It's shocking that it uses 50% more peak power in avx2 loads than the equivalent 5800x. When we put that into context on AMD side, the 5800x is the single cpu in AMD's stack with the highest TDP per core. Considering that it has the same TDP as the 5900x and 5950x but only has 8 cores to spread that TDP around. These Intel cpus make the 5800x look like a power miser lmao in comparison!

 

[Zucker2k]

What is going to change in 3 weeks tho?

The 11900k will be faster than the 11700k and it will use abit more power.

The 11900k won't clock faster than the 11700k in AVX-512 code, especially when both chips are throttling hard at 104c temps. That's what's patently false with your assertion.

It's shocking that it uses 50% more peak power in avx2 loads than the equivalent 5800x. When we put that into context on AMD side, the 5800x is the single cpu in AMD's stack with the highest TDP per core. Considering that it has the same TDP as the 5900x and 5950x but only has 8 cores to spread that TDP around. These Intel cpus make the 5800x look like a power miser lmao in comparison!

TSMC 7nm is going to be at least 50%+ more efficient than Intel 14nm by default at peak performance. I don't know why people act so surprised about this. What's interesting is that Intel could push these chips way above their efficiency zones and still remain in the ballpark. Maybe it says something about how much AMD is also pushing TSMC 7nm.

 

[TheELF]

Wow, basically DOA. Yikes.

Intel Core i7-11700K Review: Blasting Off with Rocket Lake

www.anandtech.com

TL/DR - Slower at games than 9th and 10th gen parts on the regular. Uses more power.

Kudos to Anandtech for a rare scoop.

Eh, uses stock settings for gaming and all other benches where the 9900k has 10% higher clocks
while it uses " ‘infinite turbo’ strategy " to produce as high a power draw as possible under avx.
No way to tell at what settings the ryzen CPU is running at.

We will have to wait for release to get some overclock results for gaming before being able to judge if people will buy them en mass or not.

In rendering and encoding it already gets some wins in against the 5800x and they are not even using the iGPU yet.
And that iGPU is going to be one of the main selling points to a lot of people, it's either getting intel and being able to use the system even for some gaming or getting AMD and searching for a heavily overpriced GPU.

 

[Kocicak]

Wow, Ian got his own CPU, tested it, cooked it and published everything. Bless him!

 

[uzzi38]

Eh, uses stock settings for gaming and all other benches where the 9900k has 10% higher clocks
while it uses " ‘infinite turbo’ strategy " to produce as high a power draw as possible under avx.
No way to tell at what settings the ryzen CPU is running at.

We will have to wait for release to get some overclock results for gaming before being able to judge if people will buy them en mass or not.

In rendering and encoding it already gets some wins in against the 5800x and they are not even using the iGPU yet.
And that iGPU is going to be one of the main selling points to a lot of people, it's either getting intel and being able to use the system even for some gaming or getting AMD and searching for a heavily overpriced GPU.

?

They use the same settings for both game and application testing. Whatever their specific motherboard ran at stock. The same goes for their Ryzen CPUs. Bone stock.

And lol at the idea of using overclocking results to see if people will buy them en masse or not. This really isn't a difficult concept: most people don't overclock.

Further lol at the idea of people buying the 11700K for gaming on the iGPU. It's a third of TGL-U's iGPU. You're looking at significantly lower graphics throughput as a 3200G at best. More in line with 3000G (yes, a $50 CPU) that has the iGPU overclocked to ~1600MHz. Something that is actually very achievable in practice.

The selling points for the iGPU are the media decode and QuickSync, the actual performance isn't even worth talking about.

 

[Insert_Nickname]

Worst performance than last-gen and crazy power consumption.. Does this remind anyone of Bulldozer?

220W FX9590? Certainly.

Prescott too. Or Presshot as it was known. Both single and dual core Smithfield/Pressler versions. That went so far Intel introduced the BTX form factor to aid cooling. It never did take off though.

Although compared to this 290W monster they where positively benign in comparison.

 

[tamz_msc]

No way to tell at what settings the ryzen CPU is running at.

All 105W Ryzen CPUs at stock use no more than 142W, regardless of workload because that's the PPT limit.

We will have to wait for release to get some overclock results for gaming before being able to judge if people will buy them en mass or not.

If past experience is any indication the i7s don't overclock as well as the i9, and even among the i9 the 10900K is a much better bin than the 10850K. Nobody is going to overclock these things on normal cooling setups - like AIOs and good air coolers.

In rendering and encoding it already gets some wins in against the 5800x and they are not even using the iGPU yet.

Since when did rendering become important in discussing Intel CPU performance? And Rocket Lake is actually slower than Zen 3 when encoding with AVX2 is involved as seen in the HEVC test. This has been observed with Tiger Lake as well in the forum thread for 4K Handbrake benchmark.

And that iGPU is going to be one of the main selling points to a lot of people, it's either getting intel and being able to use the system even for some gaming or getting AMD and searching for a heavily overpriced GPU.

Its iGPU is a third of what you get with top-of-the-line Tiger Lake, and nobody is buying a $450 CPU to use its iGPU for gaming. Besides many games don't even work with Iris Xe.

 

[moonbogg]

This thing is a pile of crap.

 

[TheELF]

They use the same settings for both game and application testing. Whatever their specific motherboard ran at stock.

Yeah but infinite turbo doesn't add any performance since it sticks with normal clocks but causes a lot higher power draw.

And lol at the idea of using overclocking results to see if people will buy them en masse or not. This really isn't a difficult concept: most people don't overclock.

Then most people won't buy the K version right?! This review is about the K version though right?!

Further lol at the idea of people buying the 11700K for gaming on the iGPU. It's a third of TGL-U's iGPU. You're looking at significantly lower graphics throughput as a 3200G at best. More in line with 3000G (yes, a $50 CPU) that has the iGPU overclocked to ~1600MHz. Something that is actually very achievable in practice.

That's not at all what I said.
The iGPU is going to be usable UNTILL you can find a good GPU at a good price.

If past experience is any indication the i7s don't overclock as well as the i9, and even among the i9 the 10900K is a much better bin than the 10850K. Nobody is going to overclock these things on normal cooling setups - like AIOs and good air coolers.

Sure but still O/C to 5.1 - 5.2 if it's going to be feasible, and for games it very well should be, it's going to change numbers by quite a bit.

Since when did rendering become important in discussing Intel CPU performance? And Rocket Lake is actually slower than Zen 3 when encoding with AVX2 is involved as seen in the HEVC test. This has been observed with Tiger Lake as well in the forum thread for 4K Handbrake benchmark.

Rendering is important to many people as a point of reference for IPC, anybody converting HEVC with CPU power alone has missed the train anyway but it's interesting as points of data at least.

Its iGPU is a third of what you get with top-of-the-line Tiger Lake, and nobody is buying a $450 CPU to use its iGPU for gaming. Besides many games don't even work with Iris Xe.

If it can play a few games it's still better than zero iGPU...and it's leagues better than previous intel iGPUs so it's a bigger deal than you think.

 

[uzzi38]

Yeah but infinite turbo doesn't add any performance since it sticks with normal clocks but causes a lot higher power draw.

Then most people won't buy the K version right?! This review is about the K version though right?!

That's not at all what I said.
The iGPU is going to be usable UNTILL you can find a good GPU at a good price.

Sure but still O/C to 5.1 - 5.2 if it's going to be feasible, and for games it very well should be, it's going to change numbers by quite a bit.

Rendering is important to many people as a point of reference for IPC, anybody converting HEVC with CPU power alone has missed the train anyway but it's interesting as points of data at least.

If it can play a few games it's still better than zero iGPU...and it's leagues better than previous intel iGPUs so it's a bigger deal than you think.

Infinite turbo is used to SUSTAIN the same clocks. You don't magically flip a switch and suddenly your system performs the same whilst clocking exactly the same but pulls vastly more power.

Yes, people buy the -K SKUs and don't overclock. Rather, that's what the MAJORITY of people buying K SKUs do - run them stock with XMP.

OC to 5.1-5.2GHz should be feasible based on what exactly? Have you got a chip on hand such that you can make such a call?

 

[lobz]

Anyway, a little over 3 more weeks to go. Even a user on a Z490 board did better than this lol

Eh, uses stock settings for gaming and all other benches where the 9900k has 10% higher clocks
while it uses " ‘infinite turbo’ strategy " to produce as high a power draw as possible under avx.
No way to tell at what settings the ryzen CPU is running at.

We will have to wait for release to get some overclock results for gaming before being able to judge if people will buy them en mass or not.

In rendering and encoding it already gets some wins in against the 5800x and they are not even using the iGPU yet.
And that iGPU is going to be one of the main selling points to a lot of people, it's either getting intel and being able to use the system even for some gaming or getting AMD and searching for a heavily overpriced GPU.

Eyes wide shut, eh? I'm shocked.

 

[lobz]

Infinite turbo is used to SUSTAIN the same clocks. You don't magically flip a switch and suddenly your system performs the same whilst clocking exactly the same but pulls vastly more power.

Yes, people buy the -K SKUs and don't overclock. Rather, that's what the MAJORITY of people buying K SKUs do - run them stock with XMP.

OC to 5.1-5.2GHz should be feasible based on what exactly? Have you got a chip on hand such that you can make such a call?

Of course, since K SKUs clock higher at stock as well. So yes, you're right.

 

[tamz_msc]

Sure but still O/C to 5.1 - 5.2 if it's going to be feasible, and for games it very well should be, it's going to change numbers by quite a bit.

Quite a bit eh? Do you know what the 10700K scores in gaming when overclocked to 5.1GHz compared to stock? A mere 1.3% over stock performance. Overclocking has been irrelevant for gaming ever since Intel started pushing 5 GHz ST clocks with 9th gen. You're delusional if you think Rocket Lake is going to change anything.

Rendering is important to many people as a point of reference for IPC, anybody converting HEVC with CPU power alone has missed the train anyway but it's interesting as points of data at least.

Rendering with a CPU is a completely irrelevant test. If you HAVE to render with a CPU you use Threadripper, or else use a GPU and upscale using AI like it is done in the industry. I'll say it again CPU-based rendering is a completely irrelevant workload for the vast majority of consumers.

If it can play a few games it's still better than zero iGPU...and it's leagues better than previous intel iGPUs so it's a bigger deal than you think.

Nice coping.

 

[Racan]

Welp, I didn't expect a latency regression, I wonder how it will fare in gaming with higher clocked ram and tighter timings. I hope AMD will make bank at Intel's expense though, they need that money to fund future R&D and Intel will be fine in the long run, they have plenty of resources unless they continue to screw things up further.

 

[Det0x]

At least Rocket lake is winning in the userbenchmark ranking

 

[exquisitechar]

Quite a bit eh? Do you know what the 10700K scores in gaming when overclocked to 5.1GHz compared to stock? A mere 1.3% over stock performance. Overclocking has been irrelevant for gaming ever since Intel started pushing 5 GHz ST clocks with 9th gen. You're delusional if you think Rocket Lake is going to change anything.

Yes, memory OC matters much more for gaming, and Rocket can’t go above 3733 in 1:1 mode. The latency is worse than Comet Lake even then, let alone if you take into consideration the insane OCs and low latency that CML can further achieve. Not to mention that the L3 cache is messed up too...this is beyond fixing by some magic microcode update that some people expect.

For gaming, Rocket Lake is just a disaster and not worth it over heavily discounted Comet Lake. And if the recent indications of pricing at launch mean anything, the same will be true in most other cases as well. Quite disappointing.

 

[uzzi38]

Welp, I didn't expect a latency regression, I wonder how it will fare in gaming with higher clocked ram and tighter timings. I hope AMD will make bank at Intel's expense though, they need that money to fund future R&D and Intel will be fine in the long run, they have plenty of resources unless they continue to screw things up further.

Higher clocked RAM is a no-go with RKL's IMC. Existing samples require a 1:2 mode after 3733MHz, which brings about a ~10ns latency penalty.

 

[Leeea]

avx 512 is just a mill weight around intel's neck.

They just need to drop that already. Those engineers likely could do amazing things with the freed up die space.

 

[Racan]

Higher clocked RAM is a no-go with RKL's IMC. Existing samples require a 1:2 mode after 3733MHz, which brings about a ~10ns latency penalty.

Yep, I know, still want to see the lowest latency 3733Mhz ram results though.

 

[JoeRambo]

Welp, I didn't expect a latency regression, I wonder how it will fare in gaming with higher clocked ram and tighter timings. I hope AMD will make bank at Intel's expense though, they need that money to fund future R&D and Intel will be fine in the long run, they have plenty of resources unless they continue to screw things up further.

Can't really undo the damage of incompetent L3 cache setup. They have around the same memory latency now as AMD does with memory controller that is in different chip. A milestone. RIP Intel as we knew it.

 

[Racan]

Can't really undo the damage of incompetent L3 cache setup. They have around the same memory latency now as AMD does with memory controller that is in different chip. A milestone. RIP Intel as we knew it.

Kind of surprising, no? Intel used to have the best cache design, correct me if I'm wrong.

 

[JoeRambo]

Kind of surprising, no? Intel used to have the best cache design, correct me if I'm wrong.

Completely surprising. Even more so when considering the fact that this core has kept the same inclusive L3 setup. Even the size is the good old 2MB per slice. They have really dropped the ball here, uncore seems to be clocked at 4Ghz, so no excuses here either.

There is still some hope they can turn around things with uCode update, as it seem 2 cores per chip have ~proper inter core latency. Still it takes being special stupid to loose 8-10ns when accessing memory with completely the same memory hierachy. That is like 20-25% penalty on each and every memory miss.

Oh, and you know things are BAD, when people have to check if they haven't opened 5800x review instead.

 

[SAAA]

Maybe people are being dumb on purpose beating the dead horse, OK it has far worse figures than expected, but saying like "104°C" and "290W!" misses this situation where it's happening:

Hint: that's a little more than your average 19% IPC.

As for gaming definitely big disappointment, the latency hit that hard, maybe some much faster memory even in 1:2 mode will manage something... but at this point it's worth it over Comet only if you get it for a decent price. Or not at all, with Alder coming out in a few months.