News Intel GPUs - Battlemage officially announced, evidently not cancelled

[Asterox]

Those are for lowest possible TDP. High TDP versions have higher clocks.

No, it is very obvious that Mobile ARC GPU-s dont have high GPU clocks.

Intel would very much point out if GPU can work(game clock or similar) at much higher GPU frequencies.

Intel A730M

- GPU 1100mhz

- 80-120W TDP or TGP

Intel® Arc™ A730M Graphics - Product Specifications | Intel

Intel® Arc™ A730M Graphics quick reference with specifications, features, and technologies.

www.intel.com

AMD Radeon RX 6600M

- GPU or Game Clock 2170mhz

- TDP or GPU power 100W

https://www.amd.com/en/products/graphics/amd-radeon-rx-6600m

 

[Shivansps]

That is absurdly low GPU frequency=TDP , if we put in comparison with AMD Mobile 6000 series APU-s.

View attachment 59359

Me: Intel, surely you can't beat AMD and Nvidia consufing naming...
Intel: Hold my silicon...

Someone would care to explain to me how they are going to name the next gen?

 

[jpiniero]

Here's some Rembrandt numbers as a comparison too

And 3dmark has historically been generous to Intel.

Someone would care to explain to me how they are going to name the next gen?

Starting with a B is my guess.

 

[coercitiv]

Someone would care to explain to me how they are going to name the next gen?

Starting with a B is my guess.

Arc, Bend, Curl, Dome, Ellipse, Flex, Gyre, Hyperbola, Iris.

There we go, they have more than enough rebranding names for the next decade. The one at the end is already a trademark and completes the circle. (pun intended)

 

[Glo.]

No, it is very obvious that Mobile ARC GPU-s dont have high GPU clocks.

Intel would very much point out if GPU can work(game clock or similar) at much higher GPU frequencies.

Intel A730M

- GPU 1100mhz

- 80-120W TDP or TGP

Intel® Arc™ A730M Graphics - Product Specifications | Intel

Intel® Arc™ A730M Graphics quick reference with specifications, features, and technologies.

www.intel.com

AMD Radeon RX 6600M

- GPU or Game Clock 2170mhz

- TDP or GPU power 100W

https://www.amd.com/en/products/graphics/amd-radeon-rx-6600m

If what you are saying is correct, then it is absolutely impossible for 1150 MHz 768 ALU GPU, with 64 bit bus to equal performance of 1024 ALU GPU, with 128 bit bus, with 1.2 GHz boost clock.

IMO, Intel is advertising the lowest possible clock speeds, similar to their CPU lineups. Base clocks are clocks that you are going to get at base TDP/power state.

 

[Glo.]

Me: Intel, surely you can't beat AMD and Nvidia consufing naming...
Intel: Hold my silicon...

Someone would care to explain to me how they are going to name the next gen?

Intel Arc B380, B580, B780.
Intel Arc C380, C580, C780.

And so on, and so on. Its actually the best naming scheme we have.

 

[moinmoin]

IMO, Intel is advertising the lowest possible clock speeds, similar to their CPU lineups.

I know what you mean, but shouldn't "the lowest possible clock speeds" technically be 0?

Its actually the best naming scheme we have.

Let's first see if they actually do it that or another way.

 

[Glo.]

I know what you mean, but shouldn't "the lowest possible clock speeds" technically be 0?


Let's first see if they actually do it that or another way.

Technically, yes .

I meant: the clock "floor" which we get at certain thermal envelope.

 

[eek2121]

If what you are saying is correct, then it is absolutely impossible for 1150 MHz 768 ALU GPU, with 64 bit bus to equal performance of 1024 ALU GPU, with 128 bit bus, with 1.2 GHz boost clock.

IMO, Intel is advertising the lowest possible clock speeds, similar to their CPU lineups. Base clocks are clocks that you are going to get at base TDP/power state.

Clocks don’t matter in the grand scheme of things. I know you guys know this. Someone can design a 700 mhz part that beats the fastest out there. The wider the design, the lower the clocks.

 

[linkgoron]

Starting with a B is my guess.

IIRC it'll be Battlemage and then Celestial?

 

[moonbogg]

Sir, can you please make up your mind? Will the 4060 be 1000$ or 350$. I cannot stand this emotional roller coaster any more! xD

The 4060 is a weapon and nothing more. If Nvidia aims it at YOU, then it will cost $1000. If Nvidia aims it at Intel, it will cost $350. Like any weapon, it looks different depending on which direction it faces.

"As fast as a 3090Ti for HALF the price!"

 

[Tup3x]

No, it is very obvious that Mobile ARC GPU-s dont have high GPU clocks.

Intel would very much point out if GPU can work(game clock or similar) at much higher GPU frequencies.

Intel A730M

- GPU 1100mhz

- 80-120W TDP or TGP

Intel® Arc™ A730M Graphics - Product Specifications | Intel

Intel® Arc™ A730M Graphics quick reference with specifications, features, and technologies.

www.intel.com

AMD Radeon RX 6600M

- GPU or Game Clock 2170mhz

- TDP or GPU power 100W

https://www.amd.com/en/products/graphics/amd-radeon-rx-6600m

What is obvious is that that you didn't do further reading. Hardware Unboxed guys explained those quoted clock speeds and why they were low. Intel said to them that depending on the TDP config, the clock speeds can be around 2GHz or even higher than that.

The quoted numbers where game clocks when running at the lowest configurable TDP.

 

[Glo.]

The 4060 is a weapon and nothing more. If Nvidia aims it at YOU, then it will cost $1000. If Nvidia aims it at Intel, it will cost $350. Like any weapon, it looks different depending on which direction it faces.

"As fast as a 3090Ti for HALF the price!"

4060 most likely will be based on 106 die, and that has 2304 ALU with 128 bit bus. So we are looking at 3070-3070 Ti performance. At best.

Its absolutely nothing that Intel has to be affraind of, especially considering it will be 220W TDP GPU.

 

[IntelUser2000]

And 3dmark has historically been generous to Intel.

When you compare Vega 8 to Iris Xe numbers, the ~30% gain in Time Spy is roughly the difference in actual performance(games). It's the iGPUs that perform too well compared to dGPUs in 3DMark.

Though dGPUs behave differently. Time Spy I believe is a more demanding benchmark.

This is not bad for a 96EU version of Gen 12.7. Dedicated nature is helping here, which is good. It's essentially 30% faster at the same clock as Iris Xe G7 on it's most optimal configuration.

 

[mikk]

This is not bad for a 96EU version of Gen 12.7. Dedicated nature is helping here, which is good. It's essentially 30% faster at the same clock as Iris Xe G7 on it's most optimal configuration.

We don't know the clock speeds. 1150 Mhz is the base clock for the lowest TDP.

 

[moonbogg]

4060 most likely will be based on 106 die, and that has 2304 ALU with 128 bit bus. So we are looking at 3070-3070 Ti performance. At best.

Its absolutely nothing that Intel has to be affraind of, especially considering it will be 220W TDP GPU.

You think so? A next gen 4060 performing the same as a 3070? They'd probably still charge $600 for the lolz.

 

[IntelUser2000]

We don't know the clock speeds. 1150 Mhz is the base clock for the lowest TDP.

The 2400 TS GPU score is Default mode. It's in performance mode that gets 3100 points. Default being 25W makes lot of sense. Performance is probably 35W. Most high end Iris Xe scores around 1700 points and with 30W TDP settings. They are at least 1.1GHz clock at those settings.

1.15GHz isn't the base clock. They said it's realistic frequency for gaming workloads. Refer back to HWUboxed. "Typical average frequency across a wide range of workloads. You can think of this as Turbo in "average workloads" since Turbo is active in modern CPUs almost all the time in one form or another.

In the Performance mode it's 80%+ faster. With 96EUs clock can't be the only reason for it.

 

[uzzi38]

Seems a bit uncalled for but AMD decided to take potshots at Intel for some reason or another.

https://twitter.com/x/status/1509673301081309192

 

[coercitiv]

Seems a bit uncalled for but AMD decided to take potshots at Intel for some reason or another.

Is that encoding speed? /s

PS: Lol, I didn't bother to check Twitter comments.

 

[Glo.]

You think so? A next gen 4060 performing the same as a 3070? They'd probably still charge $600 for the lolz.

Well, considering that Lovelace is going to be effectively Ampere architecture but with higher ALU count and higher clocks, I would not expect miracles.

We have seen the leaked configs, its easy now to reverse theorize( ) the performance of 106 die.

36 SMs, 2304 ALUs, 128 bit bus, with 32 MB L2, expected clocks: 2300-2400 MHz during gaming. ALU count is actually below the ALU count of 3060 Ti, but has higher clocks. Somewhere around 3070 is where we should expect full 106 die to land.

 

[moinmoin]

Seems a bit uncalled for but AMD decided to take potshots at Intel for some reason or another.

https://twitter.com/x/status/1509673301081309192

Definitely uncharacteristic as of late, but why not completely randomly liven up things...

 

[uzzi38]

When you compare Vega 8 to Iris Xe numbers, the ~30% gain in Time Spy is roughly the difference in actual performance(games). It's the iGPUs that perform too well compared to dGPUs in 3DMark.

Though dGPUs behave differently. Time Spy I believe is a more demanding benchmark.

This is not bad for a 96EU version of Gen 12.7. Dedicated nature is helping here, which is good. It's essentially 30% faster at the same clock as Iris Xe G7 on it's most optimal configuration.

It's funny you say that, because in a blog post 2 days ago Intel talked about how the latest set of drivers include additional optimisations for Intel ARC in the 3DMark suite, which nets them an extra 15% performance in Time Spy as one example.

Engineering Arc - 3/30/2022

by Lisa Pearce, Vice President and General Manager for the Visual Compute Group Excited that launch day is here! Hope everyone is tuning in for our launch event today to check out our overview of the Intel® Arc™ graphics mobile product capabilities and key demos. Just like last time, I want to...

community.intel.com

 

[IntelUser2000]

@mikk

[QUOTE="uzzi38, post: 40733022, member: 448427"

https://twitter.com/x/status/1509673301081309192

[/QUOTE]

This is what I mean when I say Intel architecture is still behind incumbents. I wouldn't be surprised if the A370M in that comparison is set at 35W and thus equalize when both are set at 50W.

However it doesn't change the fact that it requires 40% more transistors and die space to achieve the result.

Driver optimizations might help in some corner case scenarios where ARC is abnormally weak, but that's only going to be true if driver is indeed the bottleneck. One cannot expect one architecture to perform equally worse in every game or equally better in every game - different architectures have different strength. If there are no differences, then they must have identical uarch.

@uzzi38 They were talking about 2x differences in Metro Exodus using the A370M over the Iris Xe, and the Time Spy results for the lower A350M seems to be in line with that after you account for EU count differences.

I don't know, but it sounds like you are implying majority of the gains are due to driver optimizations. A dGPU of the same architecture is always noticeably faster than an iGPU period. Sharing has penalties.

 

[Aapje]

The latest rumors are that AMD is going to clock a lot higher, at least at the top end, near 3000 Mhz.

As for AMD taking potshots at Intel, but not Nvidia, this is a common strategy that I've seen many companies use and that makes a lot of sense: compare yourself with competitors that do more poorly than yourself, but not with those that are doing better. I expect AMD to start going after Nvidia if the rumors are true and they will beat Nvidia for the next gen.

 

[AtenRa]

However it doesn't change the fact that it requires 40% more transistors and die space to achieve the result.

Intel Xe-core also has Matrix ALUs (XMX) which AMDs NAVY 24 (6500M) doesn't. Also Intel has Video encoders and 8x lanes of PCIe Gen 4 vs 4x lanes on the AMD NAVY 24.
So comparing die sizes for gaming performance is not apples to apples here.