Intel Gemini Lake Refresh Coming This November

[btarunr]

Intel is preparing to debut its next generation Pentium Silver and Celeron "Gemini Lake Refresh" low-power processors in November 2019. The latest company roadmap slide detailing low-power SoC rollout, sourced by FanlessTech, pinned their launch sometime between week 45-47 (November). These are two key variants of this silicon, J and N. The J variant targets low-power desktops and AIOs, while the N variant targets notebooks, tablets, and other portables.

"Gemini Lake Refresh" SoCs are built on Intel's latest 14 nm node, and pack up to four "Goldmont Plus" CPU cores, and the same Intel UHD graphics, but offer significantly higher clock-speeds on both the CPU cores and the iGPU. Leading the pack is the Pentium Silver J5040, clocked at 2.00 GHz with up to 3.20 GHz boost. This chip succeeds the J5005, which ticks at 1.50 GHz with 2.80 GHz boost. The table below details the other J and N series models with the clock-speeds and core-counts.



View at TechPowerUp Main Site

 

[GoldenX]

Yay, more Atoms.

 

[seronx]

Meanwhile, AMD's response will either be the A9-9435/A6-9235 or Raven1(Dali). And, here I am....

 

[Vya Domus]

Those 1.1 Ghz base clocks are criminal, that shouldn't exist in 2019, there are ARM SoCs that will destroy these in sustained performance. What's odd is that at 3.1 Ghz boost on a single core it should consume way more power compared to when it's at the base clock, I don't know how they come up with these designs. I don't think they ever managed to make a good well balanced SoC.

 

[Owen1982]

Those 1.1 Ghz base clocks are criminal, that shouldn't exist in 2019, there are ARM SoCs that will destroy these in sustained performance. What's odd is that at 3.1 Ghz boost on a single core it should consume way more power compared to when it's at the base clock, I don't know how they come up with these designs. I don't think they ever managed to make a good well balanced SoC.

I really don't get those base clock thing anymore - how much time is it actually going to spend at that clock speed? It either downclocks lower that when not busy (~800Mhz?) or finds some middle ground between that and the turbo clock depending on the load. Or am I missing something? I would look at the boost clock and benchmarks to gauge performance of the chip along with the TDP and price to see if a CPU was viable for my needs. I would not look at base clock.

Also, yes the boost clocks are low, but for this segment it's expected and acceptable due to the use case and TDP limitations. I think these would be great for little NASes. Also wonder when these will actually get to manufacturers considering Intel's current supply troubles.

"it should consume way more power compared to when it's at the base clock" - yes great question & point - I think Intel should publish the TDP for the boost clocks so that manufacturers and consumers can better design & build their systems. I think most small 'Ultrabooks' with Intel's 8th Gen U series processors throttle after some hard work because their cooling solution can't handle the sustained heat output. I'm sure OEMs test for these kind of things but maybe knowing the max TDP from the get go would help them to better mitigate this.

 

[Deathy]

I'm sure OEMs test for these kind of things but maybe knowing the max TDP from the get go would help them to better mitigate this.

That's a bit naive. The OEMs know exactly what sort of cooling power is needed to sustain boost clocks. They just think (probably rightly so) that most consumers don't care much, since few do shopping based on sustained clock speeds and actual reviews of the performance of a product and rather shop by spec sheets alone, where the "up to xxx GHz" is eye catching and legal. For them, the few millimetres less dimensions and few grams of weight saved is more important than a few extra hundreds of MHz for sustained performance. And to be frank, for most consumers it probably is a fine choice, since few do actually taxing work that requires minutes of high performance from their laptop.

 

[Darmok N Jalad]

I had an Apollo Lake BRIX for a while, and it spent all of its time at the max boost under any kind of workload. I guess it probably had more cooling, but the chip was only 10W, and the whole system couldn’t even hit 25W under load. It’s about time they pushed Atom into 3.0GHz+ territory. I’d like to see them make an 8-core version with the highest clocks they can manage. It would probably still be under 50W, but that day will likely never come.

 

[Vayra86]

I really don't get those base clock thing anymore - how much time is it actually going to spend at that clock speed? It either downclocks lower that when not busy (~800Mhz?) or finds some middle ground between that and the turbo clock depending on the load. Or am I missing something? I would look at the boost clock and benchmarks to gauge performance of the chip along with the TDP and price to see if a CPU was viable for my needs. I would not look at base clock.

Also, yes the boost clocks are low, but for this segment it's expected and acceptable due to the use case and TDP limitations. I think these would be great for little NASes. Also wonder when these will actually get to manufacturers considering Intel's current supply troubles.

"it should consume way more power compared to when it's at the base clock" - yes great question & point - I think Intel should publish the TDP for the boost clocks so that manufacturers and consumers can better design & build their systems. I think most small 'Ultrabooks' with Intel's 8th Gen U series processors throttle after some hard work because their cooling solution can't handle the sustained heat output. I'm sure OEMs test for these kind of things but maybe knowing the max TDP from the get go would help them to better mitigate this.

The irony of these CPUs is, that only if you do very little on them they are of any use. Do a lot and they become slower. Any sustained load that saturates a core will kill the turbo almost instantly. And worse.

I think most small 'Ultrabooks' with Intel's 8th Gen U series processors throttle after some hard work because their cooling solution can't handle the sustained heat output

They do, when you hear jet engines in your room, you know what time it is... But then they're really not meant for anything more than burst usage; loading pages, files, network traffic, etc. In my experience anything over 30-45% sustained load will raise temps and eventually drop performance.

Its still better than it used to be. I remember very hot CPUs from way back; had an Ivy Bridge lappy that just couldn't do anything without throttling. These newer TDP neutered CPUs with more aggressive boost are still less likely to throttle fast.

 

[CityCultivator]

Those 1.1 Ghz base clocks are criminal, that shouldn't exist in 2019, there are ARM SoCs that will destroy these in sustained performance. What's odd is that at 3.1 Ghz boost on a single core it should consume way more power compared to when it's at the base clock, I don't know how they come up with these designs. I don't think they ever managed to make a good well balanced SoC.

Gemini lake on boost can use up to 15W. This allows for 3.1GHz single boost.
On sustained, the rated tdp is used. 1.1GHz is the rated all-core frequency. On single core, the chip will be higher than 1.1GHz.