Intel Core i7-1065G7 Gaming Test: Integrated Graphics Performance

[Julio Franco]

After exploring the capabilities of the new Ice Lake Core i7-1065G7 on the productivity side, we're coming back to test the CPU for ultrabook gaming using the G11 Iris Plus integrated graphics.

Read the full article here.

https://www.techspot.com/review/1951-intel-core-i7-1065g7-iris-plus-graphics/

 

[Vulcanproject]

Well far as I can see it's not really much faster than the nearly 2 year old 2500U (the 3500U is a tad faster than that), and that's when you're testing with monstrous LPDDR4X-3733 RAM which the 2500U won't be paired with. What was the RAM in the 2500U machine if I may ask? Standard 2400MHz?

Only because this makes a world of difference on many games, when integrated graphics pull from the same pool of memory and bandwidth as the CPU.

Although I would say that drivers matter an awful lot too. AMD's are now refined for Vega integrated and Intel's need quite a bit more work, but they are improving slowly.

It's a decent improvement on previous generations, however the G7 configuration probably won't appear in many machines unfortunately. They will have the lower configs (G1 or G4) for budget machines, and then above that manufacturers will just jump straight to discrete parts.

This is Intel's fault again, for pricing out the better graphics cores on their chips only at the higher end. The Ryzen 5 2500U and now the 3500U find their way into budget machines, and offer fantastic price performance at that level.

 

[Verrm]

I cant wait for Zen 2 based APUs. I need in my life APUs that can run most titles 1080p on lowest settings 45fps+ and esport titles 60fps+. Only then id say you got a good discrete card alternative. Otherwise you should forget APUs exist and get cheapest discrete. Unless you only play really old titles or one specific game that can runs on any PC.

 

[QuantumPhysics]

I wouldn't be relying on integrated graphics for anything beyond a ultraportable designed solely for "work".

Beyond that, I'm buying a $1500 or less "gaming spec" laptop with at least a 2060.

Two of my coworkers came to me for advice on buying a laptop and I steered them towards Black Friday sale HP Omen's with i7+2060+16GB+ SSD

These games simply demand too much nowadays to rely on integrated graphics.

It would be nice, however, to see Intel come up with a competitor GPU card.

 

[MarkHughes]

"But as we know, for years Intel’s integrated GPUs haven’t been powerful enough to run popular titles like Fortnite or even CS:Go at reasonable but the lowest settings."

I think you have a bit missing there...

 

[fps4ever]

I wonder who at Intel decided 15 watt cpu's should be the norm in laptops. That is the real travesty. The 25-35 watt cpu's are so much better. I mean look at the base clocks from previous generations. I guess you can't have a laptop weighing 8 more ounces...

 

[Theinsanegamer]

I wonder who at Intel decided 15 watt cpu's should be the norm in laptops. That is the real travesty. The 25-35 watt cpu's are so much better. I mean look at the base clocks from previous generations. I guess you can't have a laptop weighing 8 more ounces...

I miss the old laptops with 90+ wh batteries, 1 inch thick bodies, and 35 watt CPUs that could do actual work. Throw one of these G7s in there, with a 35 watt TDP and hgiher clocks, or a ryzen zen 2 APU with the same or even a 45 watt TDP. Preferably with linux.

that's all I've been looking for for years. We were so close, then the thinness wars began. A good, powerful 14 inch laptop that is built well is still on my wish list, but I''ve given up on it ever being made.

 

[Lew Zealand]

I wonder who at Intel decided 15 watt cpu's should be the norm in laptops. That is the real travesty. The 25-35 watt cpu's are so much better. I mean look at the base clocks from previous generations. I guess you can't have a laptop weighing 8 more ounces...

There's nothing wrong with the base clocks, they go up every year. When you keep the same number of cores.

The base clocks are going down in new gens because Intel is cramming in 4 and now 6 cores into the same thermal envelope. What matters for everyday performance is the Turbo clocks when using the same number of cores, and that number increases every year.

 

[opckieran]

Well far as I can see it's not really much faster than the nearly 2 year old 2500U (the 3500U is a tad faster than that), and that's when you're testing with monstrous LPDDR4X-3733 RAM which the 2500U won't be paired with. What was the RAM in the 2500U machine if I may ask? Standard 2400MHz?

I wouldn't fault Intel for supporting faster RAM. It's not their fault Ryzen doesn't support it (yet).

 

[fps4ever]

There's nothing wrong with the base clocks, they go up every year. When you keep the same number of cores.

The base clocks are going down in new gens because Intel is cramming in 4 and now 6 cores into the same thermal envelope. What matters for everyday performance is the Turbo clocks when using the same number of cores, and that number increases every year.

Look at the sustained turbo clocks not the instant turbo boost for a minute or so. And where are you seeing the base clocks going up on the newer cpu's? They are not keeping same number of cores so you are contradicting yourself about keeping the same number of cores. Now spread twice as many cores performance over 15 watts and there you have it. And how many general use laptop programs are going to use more than two or four cores?

 

[Lew Zealand]

Look at the sustained turbo clocks not the instant turbo boost for a minute or so. And where are you seeing the base clocks going up on the newer cpu's? They are not keeping same number of cores so you are contradicting yourself about keeping the same number of cores. Now spread twice as many cores performance over 15 watts and there you have it. And how many general use laptop programs are going to use more than two or four cores?

Sustained Turbo clocks are higher for each successive generation, save the Ice Lake one as it's using the immature 10nm node. Yes, they are lower in that case thanks to Intel's well publicized problems at 10nm. Improvements in sustained Turbo are most easily viewed in Cinebench R15 single core results:

Note the single core progression of Core i7s from 7500U to 8550U to 8565U to 10710U (all 14nm Skylake cores so same IPC). Too bad Skylake and earlier weren't present in this graph, the progression would continue.

I clearly stated that base clocks continue to rise when you keep the core count (OK, and rated CPU wattage) the same. Have a look at Intel's product stack for reference. Again 10nm Ice Lake fails to continue this thanks to the immature process node, but apparently makes up for it with higher IPC. However this year 14nm Comet Lake does keep the same base clocks so Intel may have finally found the limits of 14nm+++, but at least the boost clocks offer slightly higher sustained performance.

 

[lipe123]

Really the article just shows two things to me,

1. Even brand new igpu's are still really really bad.
2. Naming conventions are getting worse as time goes on.

How did we go from a geforce 2 mx 200 in 2001 to a geforce mx 250 in 2019?
How is the 1065xx later/better than 107xx cpu?!

Is it THAT hard to pick model numbers that make sense so the consumers can easily tell lower is worse and higher is better?

 

[Lew Zealand]

Your original point about the lack of mid power processors is spot on. The last processors released in the 28-45W range are from Coffee Lake a year and a half ago. Apparently this market gets "older" chips or nothing at all.

 

[Vulcanproject]

I wouldn't fault Intel for supporting faster RAM. It's not their fault Ryzen doesn't support it (yet).

It was just an observation. Most machines are not going to be paired with this extreme memory speed, is what I'm thinking here.

The Ryzen part is two years old so was never going to support that speed of RAM. My point was the spec used here is going to to give this Intel iGPU a leg up, but it still doesn't really beat an old Ryzen part running at vastly inferior memory speed!

Which means in an average memory configuration you're more likely to come across the Intel part is only going to be slower than the tests show here. It's absolute best case scenario, needing mega RAM just to match up to an old Ryzen APU in 3D applications.

I think what you're likely to see is Zen 2 based Ryzen APUs knocking this into a cocked hat. Bridging the gap better to discrete GPUs like the MX250, and a lot cheaper too.

 

[systemBuilder22]

What matters is that Intel saved a TON of money by replacing GPU designers with marketing guys who did five years of heavy lifting by thinking up new HD GPU numbers from HD 4000 - HD 650 and putting together a bunch of slick slide presentations. All during that time there was zero performance improvement, overall (Iris 5200 still is king). As I'm sure everybody is aware the same thing also has happened on the CPU design side, when the 3rd rate people took executive control of the company in 2012 ...

 

[Lew Zealand]

I'm looking forward to a 7nm 6C12T Ryzen 5 4500 with Navi 12 or 14 Gfx, hopefully sometime next year. We'll see. I assume there'll be a 8C16T Ryzen 7 4700 with Navi 16 or so but I'm not (too) greedy, I'll settle for the less expensive Ryzen 5.

 

[systemBuilder22]

I'm looking forward to a 7nm 6C12T Ryzen 5 4500 with Navi 12 or 14 Gfx, hopefully sometime next year. We'll see. I assume there'll be a 8C16T Ryzen 7 4700 with Navi 16 or so but I'm not (too) greedy, I'll settle for the less expensive Ryzen 5.

That will be a 35w part ...

 

[Lew Zealand]

What matters is that Intel saved a TON of money by replacing GPU designers with marketing guys who did the heavy lifting of think up new HD GPU numbers from HD 4000 - HD 650 and putting together a bunch of slick slide presentations. As I'm sure everybody is aware the same thing also has happened on the CPU design side, when the 3rd rate people took executive control of the company ...

Intel made more than just the GT2 iGPUs mentioned here as they've had better GT3 Iris Plus and Pro iGPUs since the Haswell days. However almost nobody outside of Apple and Intel put CPUs with those (somewhat) capable iGPUs in any devices.

Seeing as these Ice Lake GPUs are also branded Iris Plus, it does mean a compelling comparison is also to those previous gen Iris Plus GPUs which have 48 EUs (like the G4 variants listed in the table) and 2400MHz RAM +128MB of L4/dedicated cache.

 

[Lew Zealand]

That will be a 35w part ...

True, maybe a 15-25W Ryzen 5 4C8T Navi 10? Pretty please?

 

[fps4ever]

Sustained Turbo clocks are higher for each successive generation, save the Ice Lake one as it's using the immature 10nm node. Yes, they are lower in that case thanks to Intel's well publicized problems at 10nm. Improvements in sustained Turbo are most easily viewed in Cinebench R15 single core results:

Note the single core progression of Core i7s from 7500U to 8550U to 8565U to 10710U (all 14nm Skylake cores so same IPC). Too bad Skylake and earlier weren't present in this graph, the progression would continue.

I clearly stated that base clocks continue to rise when you keep the core count (OK, and rated CPU wattage) the same. Have a look at Intel's product stack for reference. Again 10nm Ice Lake fails to continue this thanks to the immature process node, but apparently makes up for it with higher IPC. However this year 14nm Comet Lake does keep the same base clocks so Intel may have finally found the limits of 14nm+++, but at least the boost clocks offer slightly higher sustained performance.

What do you mean by sustained turbo? That is not shown in your charts. Those are just peak numbers. Nobody is talking about IPC being lower. A benchmark is not real world performance. And you are talking about 25 watt cpu's (not the one in pic), not the average configured tdp of 15 watt cpus they are going to throw in most laptops is what I was talking about? You continue to say base clocks rise and yet I see no proof of that after the 7500U?

 

[Lew Zealand]

I'm referring to sustained CPU GHz when it is running at defined Wattage (say, 15W). This will almost always be above the rated base clock, so will be somewhere in the CPU's turbo range, but rarely will it be at the top of that turbo range (and prob. never at 15W). An example from my desktop: Core i5-8400 65W chip 2.8GHz base clock, 3.8GHz all core turbo. Running Handbrake h.265 will full utilization it uses at 72W at 3.8 all core turbo and clocks down to about 3.6 all core to stay at 65W usage, still well above base clock. That's sustained turbo. The same thing applies even to 15W chips.

Thanks to process refinements, newer generation chips at the same process node (say, 14nm) dissipate less heat at the same GHz speed than prev. gen parts, and therefore can clock up a bit more to sustain a higher turbo speed while operating at the same power limit. It doesn't matter whether they are restricted to 15W or 25W, you'll still get higher GHz from a newer process refinement at the same wattage.

 

[Mr Majestyk]

Basically iGPU is still pathetic that even 1080p at low setting is unplayable on many games going into 2020. That it can hardly beat 2500U is not great. I’m hoping for a laptop next year with low powered Zen3 and RX5300M or APU with Navi.

 

[Lew Zealand]

What do you mean by sustained turbo? That is not shown in your charts. Those are just peak numbers. Nobody is talking about IPC being lower. A benchmark is not real world performance. And you are talking about 25 watt cpu's (not the one in pic), not the average configured tdp of 15 watt cpus they are going to throw in most laptops is what I was talking about? You continue to say base clocks rise and yet I see no proof of that after the 7500U?

Please reread both my comments above about base clocks. Once again, I clearly say that base clocks **With The SAME NUMBER Of Cores** continue to go up. Look at the core count in the chart you posted, the core counts are changing after Kaby Lake, forcing the base clock lower to keep power usage with all-cores active at 15W. Have a look at the entirety of Intel's product stack and you see that base clocks rise each generation when you keep the core count the same.

Anyway the base clock being lower doesn't make it slower, the i7-8550U is a more powerful CPU than the i7-7500u in every way. I doubt you can find a benchmark or real world use case where the 7500u is faster than the 8550u.

 

[hapkiman]

Makes absolutely no sense to buy a processor like this and then try and game on it...and then be disapointed. You could easily buy a low end Ryzen or Intel proc and pair it with a regular discrete GPU to game on for the same or even less money. If you need a laptop for work or school - great, but why expect to game on something like this?

 

[Lew Zealand]

Makes absolutely no sense to buy a processor like this and then try and game on it...and then be disapointed. You could easily buy a low end Ryzen or Intel proc and pair it with a regular discrete GPU to game on for the same or even less money. If you need a laptop for work or school - great, but why expect to game on something like this?

I agree. I started PC gaming on ultra low end devices mostly because that's what I had, gaming wasn't the original intent. And I spent a lot of time looking for something small and/or integrated which competes with a decent low end discrete GPU laptop. It doesn't exist.

If you are going into a build or looking for a form factor like a laptop or mini PC, you can get something more capable for roughly the same price as these iGPU options. For instance a 4C8T Core i5-9300h and GTX 1650 for ~$600 (recent sale price on an MSI laptop).

Good luck finding an iGPU laptop with otherwise similar specs for ~$200 less and even then, just save up the money and get the MSI as it's 2-3x better for gaming.