Intel Rocket lake i7-11700 CPU in the Dell Vostro system. Is it good?

[cagoblex]

Today we will be looking at something very special. It is a Dell Vostro SFF PC, but it comes with the 11^th generation 14nm++++++++++++++++++++++ Rocket lake-S Processor!

 

 

This is a Engineering Sample system and the actual system is scheduled to release in March 2021. It has a blue motherboard, which is what Dell uses for ES motherboards. The production units will have green motherboards. This is a pretty simple board, or to say a cost optimized boards. According to my experience with ES Dell motherboard, it will usually contain more stuff than the production units, so I would assume it would be even more cost optimized when they make the production motherboards This particular board has a PCI-E x1 slot and a missing PCI-E x8 slot. The production unit will most likely be equipped with the PCI-E x8 slot. However I have confirmed with my source that this board will be using PCI-E3.0 although the CPU supports PCI-E 4.0. That means the M.2 slot and the PCI-E slots will be running in PCI-E3.0.

 

It has a M.2 2240 slot, which means you won’t be able to install an M.2 2280 SSD into it. For the power supply, it is using the new Intel 12V standard. 12V is the only voltage being fed to the motherboard. It has a 4pin CPU power connector and a 6 pin motherboard connector.

 

It has LGA1200 socket with B560 chipset. A simple aluminum heatsink covers the chipset. The power delivery for the CPU is pretty simple, it has only 4 phases for CPU core power delivery. It supports 65W maximum TDP, which means it won’t support the K SKU. The heatsink looks pretty slim.

 

Coming with it is a ES processor. It is supposed to be an ES for 11700. Here we have the processor on the left, and a 10700 on the right. As you can see, the IHS on the 11700 is taller than the 10700, and it’s more of a rectangular shape. This makes it looks a little bit like the Alder lake 12^th gen processor. Sorry I can’t show you what it looks like right now...but it’s basically a taller version of 11700. On the back, the capacitor is totally different.

 

Here is a CPU-Z screenshot. It is 8 cores and 16 threads, it is a stepping 0 processor, which is 1 stepping before the production units. The production units will be stepping 1. The biggest change for this generation is the added support of AVX512F instruction set. But please note that it’s not the full AVX512 instruction set, it is missing a few instructions. Another big difference is the cache algorithm. It has 12 ways 48KB of L1 DATA cache, compares to 8 way of 32KB on the 10700. It has 512KB 8 way L2 cache, compares to 4 way 256MB of cache on 10700. The L3 cache remains the same. It has a single core boost of 4.4GHz and all core boost of 3.8GHz.

 

 

The current drivers listed on Intel website does not support the B560 chipset or the new iGPU. But luckily I have the driver package for this ES PC. However they are not the latest. I used the drivers that came with the Asus Z590 instead. Yes the Maximus XIII Hero Z590 motherboard has just arrived. I will be publishing a review of the board this weekend. GPU-Z still cannot recognize the iGPU. 

 

 

 

Firstly the CPU performance. The score is very disappointing in CPU-Z Benchmark. It is slower than i7 10700 in both single and multi core testings. It only got 511 points in single single bench and 5120 in multi core bench. 

 

As for Cinebench R15 and R20, things remains the same. It is considerably slower than the 10700. Yeah I know it’s an Engineering Sample with only 1.8GHz clock speed, but Intel processor runs on Turbo boost frequencies anyways, so this shouldn’t be an excuse.

 

Next let’s try some Prim95 pressure test. Here we are doing small FFTs, which is the most stressful test. The CPU frequency drops to 2.2GHz all core, and that raised my attention. At first I thought it’s thermal throttling, however after close inspection that is not the reason. The real problem is the TDP limit. With higher IPC and different cache designs, 11^th rocket lake CPUs are consuming more powers per core at the same frequency, the voltage is also considerably higher than the 10^th gen processor. With that said, the 65W power budget is becoming a problem. Unlike Asus boards that comes with Multi core Enhancement that removes the power limit, the Dell system is strictly following the Intel specs. 65W is nothing for an 8 core chip especially considering it has iGPU as well. Our CPU is being handicapped by the power limit.

 

 

With that discovered, let’s fire up Intel XTU and increase the power limit. Here let’s increase the PL1 to 125W and the Tau to 128 seconds. Let’s run the benchmarks again and see how it goes. The results are pretty self-explanatory. We are seeing a great improvement of 15%-20% in all the benchmarks. With the TDP set to 125W, and upgrading to the new graphics driver that came with the Asus motherboard, the 3D Mark score increased by another 500 points. That makes it about 40% faster than the 10^th gen UHD630 iGPU. This is very impressive and that makes me looking forward to the Intel XE graphics card.

 

 

 

 

 

 

However gaming is not the only improvement. In Geekbench 5 compute tests, the new iGPU is able to achieve 15% increase in both OpenCL and Vulkan.

 

 

 

It is hard to recommend the actual Vostro system to most of the viewers here on the forum. It is cost optimized, and is not able to use a discrete graphics card. It is built for business users not for gamers. But as for the CPU and the platform, it is actually worth upgrading to. Yes I know the performance figures doesn’t look that impressive here in this review, but spoiler alert, it performs much better on the Asus M13H board. I was able to get anywhere from 10-20% better scores with the Asus board. And again remember, this is just an ES processor with 1.8GHz frequency, 10700 is a processor with 2.9GHz frequency and much higher boost clock. So even a tie here would mean higher IPC and better multi core efficiency for the 11700.

 

Please stay tuned for the Asus Maximus XIII Hero Z590 review. I will be comparing it heads to heads with the Asus Maximus XII Hero Z490 motherboards and tell you why you should or shouldn’t buy the new Z590 motherboard. Those are expected to ship in early March so you still got time to sell your current platform if you decide to upgrade. We will be comparing PCI-E performance with 4 Samsung PM9A1 PCI-E 4.0 SSDs in RAID0 mode. You won’t believe the number you are seeing. I will also be doing a lot of gaming benchmarks as well as more 3D rendering tests with an RTX3080. I have a 11900K and a 11900T on the way and I will be testing that as well if it can make it here on time.

 

For more details, you can check out the video version on my Youtube channel:

 

 https://youtu.be/-S23HcQcs4Q

 

Thanks for reading!

 

 

 

[Downkey]

3 minutes ago, cagoblex said:

11^th generation 14nm++++++++++++++++++++++ Rocket lake-S Processor!

 

 

 

you forgot a couple pluses

[cagoblex]

Just now, Downkey said:

you forgot a couple pluses

LOL. It's fair to put a whole page of + for Intel

[porina]

 

10 minutes ago, cagoblex said:

The biggest change for this generation is the added support of AVX512F instruction set. But please note that it’s not the full AVX512 instruction set, it is missing a few instructions.

AVX-512 is a collection of mandatory and optional features. As far as I'm aware, there is no single Intel CPU that supports every possible option, and it isn't supported at all by AMD currently.

 

10 minutes ago, cagoblex said:

 

Yeah I know it’s an Engineering Sample with only 1.8GHz clock speed

It would be much more helpful if you showed what the clock was while benches were running as I gather it wasn't 1.8 GHz, and you had mentioned 2.2 GHz for Prime95.

 

10 minutes ago, cagoblex said:

 Unlike Asus boards that comes with Multi core Enhancement that removes the power limit

Adjusting the power limit and enabling MCE are two totally different and unrelated things. Power limit adjustment is allowed by Intel and not considered overclocking, and it is up to the system builder to set appropriate limits for their thermal design. MCE is an overclock as it intentionally adjusts clocks and voltages directly.

 

 

[cagoblex]

15 minutes ago, porina said:

 

AVX-512 is a collection of mandatory and optional features. As far as I'm aware, there is no single Intel CPU that supports every possible option, and it isn't supported at all by AMD currently.

 

It would be much more helpful if you showed what the clock was while benches were running as I gather it wasn't 1.8 GHz, and you had mentioned 2.2 GHz for Prime95.

 

Adjusting the power limit and enabling MCE are two totally different and unrelated things. Power limit adjustment is allowed by Intel and not considered overclocking, and it is up to the system builder to set appropriate limits for their thermal design. MCE is an overclock as it intentionally adjusts clocks and voltages directly.

 

 

Thanks for your input.

 

Yes I know AVX512 is a set of instruction set but the support on Rocket lake is pretty limited. 

 

I agree. I just cannot upload more pictures because of the 20MB attachment limit. The frequency for Geekbench5 is 3.8GHz all core. It is about 3.3GHz for Cinebench and 2.2GHz for Prime95. So the IPC and multi core efficiency has definitely improved quite a bit. 

 

Yes I know MCE and adjusting the power limit are two different things, but enabling MCE on Z590 does automatically set the PL1 and PL2 to 4095W for 448 seconds. So I guess that's just a shortcut to do it fast. I didn't get much time to play with the Z590 yet so that's the first thing I did. The clock and voltage curve is really interesting for the CPU. It reaches 1.45V at once the frequency goes beyond 4.0GHz. I'm not sure if it's the issue with the particular ES chip I have but from what I heard, Rocket lake generally requires much higher voltage at the same frequency compares to Comet lake. 

[porina]

2 minutes ago, cagoblex said:

Yes I know MCE and adjusting the power limit are two different things, but enabling MCE on Z590 does automatically set the PL1 and PL2 to 4095W for 448 seconds. So I guess that's just a shortcut to do it fast. I didn't get much time to play with the Z590 yet so that's the first thing I did. The clock and voltage curve is really interesting for the CPU. It reaches 1.45V at once the frequency goes beyond 4.0GHz. I'm not sure if it's the issue with the particular ES chip I have but from what I heard, Rocket lake generally requires much higher voltage at the same frequency compares to Comet lake. 

Was the testing done with MCE on or off? If on, this is an overclocked condition and working outside Intel's intended operation. For maximum non-overclocked performance and power testing, I'd set MCE off, but manually adjust the power limits to effectively unlimited (4095W?). The time doesn't matter if PL1=PL2.

[cagoblex]

1 minute ago, porina said:

Was the testing done with MCE on or off? If on, this is an overclocked condition and working outside Intel's intended operation. For maximum non-overclocked performance and power testing, I'd set MCE off, but manually adjust the power limits to effectively unlimited (4095W?). The time doesn't matter if PL1=PL2.

These testing are made on the Dell Vostro system. You know how these computers are. They won't offer any kind overclock options. So these scores you are seeing are completely stock. For the MCE I'm talking about the Asus board review I am working on. Yeah the Tau doesn't really matter when PL1=PL2 but Asus would set it to maximum anyways. For the testing, I will do both with MCE and w/o MCE. I am testing it against a 10700 anyways, which has the same 65W TDP. As long as the testing conditions are the same for the two, it's still a fair comparison.

When it really matters is the 11900K. MCE will make a big difference with it. However I doubt if 11900K would make it here on time with all the delays in shipping. So I might be doing the Z590 motherboard review with the same ES chip in this review.  

[porina]

51 minutes ago, cagoblex said:

These testing are made on the Dell Vostro system. You know how these computers are. They won't offer any kind overclock options. So these scores you are seeing are completely stock. For the MCE I'm talking about the Asus board review I am working on.

Gotcha, I got confused due to the mentioning of Asus board forgetting it was the Dell system.

 

Quote

For the testing, I will do both with MCE and w/o MCE. I am testing it against a 10700 anyways, which has the same 65W TDP. As long as the testing conditions are the same for the two, it's still a fair comparison.

Sounds good, look forward to it.

 

 

If possible, any chance of a custom Prime95 bench on the ES? Select benchmark in the menu, Throughput Benchmark, set min and max FFT size to 128, can leave HT deselected, and you can reduce time to run to minimum of 5 seconds, but run it a few times as a check there isn't a bad run. Please check the clock while it is running (is it still 2.2 GHz or something else?), and results can be found both on screen and also in file results.bench.txt in p95 folder.

 

I'm curious in two parts: 1, if prime95 will recognise and use AVX512, and 2, if performance in this area has improved over Skylake. Zen 2 was the first AMD CPU to pass Skylake IPC running about 4% higher. Zen 3 is about 10% higher IPC than Zen 2. 

 

For results context, I estimate something around 12250 iter/s for 8 core, 8 workers at 2.2 GHz, 128k FFT using AVX2. If it is a LOT higher than that (>50%), it is probably using AVX-512 with two unit hardware implementation. If it a lot lower than that, it is probably using AVX-512 on a one unit implementation and Prime95 has had a performance problem with that, and it might be more interesting to re-run with AVX-512 support disabled in P95.

 

 

Edit: I'm part way through the video where it is possible to gain a little more information. I note at the start of the Prime95 torture test, the "disable AVX-512" option was available and not selected. So it does look like it has detected and is using AVX-512. It was also interesting to see the clock reported by CPU-Z to bounce between 2.2 - 2.3 GHz, which isn't unexpected when running at a power limit, but it means it would need to be taken into consideration.

[cagoblex]

9 hours ago, porina said:

Gotcha, I got confused due to the mentioning of Asus board forgetting it was the Dell system.

 

Sounds good, look forward to it.

 

 

If possible, any chance of a custom Prime95 bench on the ES? Select benchmark in the menu, Throughput Benchmark, set min and max FFT size to 128, can leave HT deselected, and you can reduce time to run to minimum of 5 seconds, but run it a few times as a check there isn't a bad run. Please check the clock while it is running (is it still 2.2 GHz or something else?), and results can be found both on screen and also in file results.bench.txt in p95 folder.

 

 

9 hours ago, porina said:

 

I'm curious in two parts: 1, if prime95 will recognise and use AVX512, and 2, if performance in this area has improved over Skylake. Zen 2 was the first AMD CPU to pass Skylake IPC running about 4% higher. Zen 3 is about 10% higher IPC than Zen 2. 

 

For results context, I estimate something around 12250 iter/s for 8 core, 8 workers at 2.2 GHz, 128k FFT using AVX2. If it is a LOT higher than that (>50%), it is probably using AVX-512 with two unit hardware implementation. If it a lot lower than that, it is probably using AVX-512 on a one unit implementation and Prime95 has had a performance problem with that, and it might be more interesting to re-run with AVX-512 support disabled in P95.

 

 

Edit: I'm part way through the video where it is possible to gain a little more information. I note at the start of the Prime95 torture test, the "disable AVX-512" option was available and not selected. So it does look like it has detected and is using AVX-512. It was also interesting to see the clock reported by CPU-Z to bounce between 2.2 - 2.3 GHz, which isn't unexpected when running at a power limit, but it means it would need to be taken into consideration.

Yes I will give that a try. If I run it completely stock with the 65W TDP then 2.2GHz is the highest it would go for Prime95. I actually didn't take AVX512 into consideration when drawing the conclusion. Given the frequency, I would say prime95 picked up AVX512 because that's the only reason it's consuming 65W at 2.2GHz all core. This is similar to the power consumption I'm seeing on LGA3647 Skylake-SP processors running AVX512. But I will retest it in the way you suggested and include the result in the Z590 review. I will also test in some other AVX512 workload and see how it works. Still hope I can get the 11900K in on time coz this ES chip is one stepping before retail, and things might be different. 

 

 

[WhitetailAni]

Dell's been using the 12v only PSU thing since at least 4th gen Intel.

[cagoblex]

18 minutes ago, ragnarok0273 said:

Dell's been using the 12v only PSU thing since at least 4th gen Intel.

Yes I This particular PSU is compatible with 8th gen and up. 

[porina]

34 minutes ago, cagoblex said:

I will also test in some other AVX512 workload and see how it works.

What others do you know of? The only other one I run other than Prime95 (and close relatives) is y-cruncher. It is well optimised for various architectures, but it is also quite memory bandwidth bound so caution may be needed in interpreting results.

[cagoblex]

48 minutes ago, porina said:

What others do you know of? The only other one I run other than Prime95 (and close relatives) is y-cruncher. It is well optimised for various architectures, but it is also quite memory bandwidth bound so caution may be needed in interpreting results.

Yeah Y-cruncher is one. AIDA 64 actually uses AVX 512 in system stability tests. I will look into other benchmarks too.