News Intel's Iris Xe Max GPU Barely Beats Nvidia's GeForce MX330 In Latest Benchmark

[Admin]

Someone has tested Intel's looming Iris Xe Max discrete GPU in Geekbench 5.

Intel's Iris Xe Max GPU Barely Beats Nvidia's GeForce MX330 In Latest Benchmark : Read more

 

[pacdrum_88]

What a disaster. I mean, the RX 550 is $60 right now. Just bought one the other day. And, it's old. If this is the best Intel has, what all of their R&D has been going into, future doesn't look bright. AMD crushing on CPU and GPU front. And, Intel sold off their NAND fab if I recall... What's their plan going forward?

 

[evdjj3j]

Is anyone surprised?

 

[hotaru.hino]

I don't think Intel is targeting DIYers and people who want add-in cards. Where this will shine is the system builder market where Intel can provide a complete package for those who want something with more GPU power. i.e., it gives Intel clout on clamping down their system builds even further to only their systems.

This will likely put NVIDIA in a bind on the low-end but... eh. I'm sure they make bookoo bucks there /s.

 

[escksu]

What a disaster. I mean, the RX 550 is $60 right now. Just bought one the other day. And, it's old. If this is the best Intel has, what all of their R&D has been going into, future doesn't look bright. AMD crushing on CPU and GPU front. And, Intel sold off their NAND fab if I recall... What's their plan going forward?

If the power consumption is 10w it will be very good.

 

[artk2219]

If the power consumption is 10w it will be very good.

Thats the thing, as far as i know they didnt claim this was a crazy high end chip, honestly its decent enough to play a ton of games (not necessarily at amazing settings) so long as their driver support doesnt suck. If the price is right, the performance is decent, the power consumption is in check, and theyre readily available, they would be a welcome option in the mainstream graphics segment.

 

[i-am-i-u-r-u]

Now I know why AMD let a number of people go who were hired by Intel.

 

[nofanneeded]

This is good result for intel , if it can beat nvidia MX330 at the same TDP .

 

[shady28]

The Xe DG1 is the same as the iGPU part.

DG2 is the one that is supposed to be around 3070 levels of performance. That one hasn't yet been seen.

So basically the iGPU has performance like an MX330. And when they put it in a discrete package, it performs like an MX330.

There's nothing new here.

Congrats on turning "nothing new" into yet another clickbait headline.

 

[SM2262EN]

I mean you also need to remember that this is being paired with slow LPDDR4X memory, rather than the much faster GDDR5/6. When Nvidia did this with the GT1030, the performance often just about halved, as shown here: https://www.gamersnexus.net/hwrevie...1030-gddr5-benchmark-worst-graphics-card-2018

 

[zodiacfml]

What an ultra safe, low volume move. Who still puts mx330 performance level on their laptops which can be replaced by iGPU which I remember correctly 70 to 80% performance of that card.

 

[Rdslw]

If the power consumption is 10w it will be very good.

Exacly my point its ~mx330 performance, ok. AT WHAT POWER ?
Comparison cards use:
20 Watt (rated 25) & 10 Watt (rated 25) so if it's a sub 10W card, like 7W or 5W. Not just on paper but in real usage, then its a good start.


might not be bad for ultrathin devices that can into skyrim for weekend players.

 

[shady28]

What an ultra safe, low volume move. Who still puts mx330 performance level on their laptops which can be replaced by iGPU which I remember correctly 70 to 80% performance of that card.

Only if you get a Tiger Lake laptop will you get that level of iGPU performance though. It's quite a bit better than anything else on the market as far as iGPU.

I think this is just a way for Intel to ease into the dGPU market. They do this for 6-9 months to iron out any issues, then release some heavy iron in the dGPU space maybe late Q3 or early Q4 next year to make a splash in the 2021 holiday season. A solid strategy IMO.

 

[watzupken]

I think the information is quite confusing. On one hand, there are reviews out there saying that the Xe graphics on Tiger Lake is beating the likes of a Nvidia MX350, and yet the dedicated solution is slower than a MX330?

 

[nofanneeded]

I think the information is quite confusing. On one hand, there are reviews out there saying that the Xe graphics on Tiger Lake is beating the likes of a Nvidia MX350, and yet the dedicated solution is slower than a MX330?

Maybe at different watt settings

 

[Soaptrail]

The Xe DG1 is the same as the iGPU part.

DG2 is the one that is supposed to be around 3070 levels of performance. That one hasn't yet been seen.

So basically the iGPU has performance like an MX330. And when they put it in a discrete package, it performs like an MX330.

There's nothing new here.

Congrats on turning "nothing new" into yet another clickbait headline.

Is DG2 a future gen part, because if the Iris Xe Max is DG1 and DG2 is the same gen would they have to call it a Xe SuperMax?

 

[shady28]

Is DG2 a future gen part, because if the Iris Xe Max is DG1 and DG2 is the same gen would they have to call it a Xe SuperMax?

Lol! Dunno about their stupid naming convention. 'Max' is going to backfire on them I think, though if I go by what's written on Nvidia / AMD low end GPU boxes it's par for the course.

If they just called it Xe 1 or Xe DG1 or Xe Mini or Xe LP (Low Power) it would make more sense.

I would bet that DG2 is going to be an upper midranger, like a 1660 Super / 2060 / 3060. I think the media has it wrong talking about 3070 performance, that to me is high end and I think Intel is going from bottom to top.

Bottom line though is the DG2 would be in the same generation as DG1, but bigger in the same way that a 2080 is a bigger version of a 2060.

 

[JayNor]

Bottom line though is the DG2 would be in the same generation as DG1, but bigger in the same way that a 2080 is a bigger version of a 2060.

DG2 is reportedly a monolithic Xe-HPG, built with tsmc n6 process. It is their only GPU with ray tracing.

There will be a 4 tile version of Xe-LP/DG1 called SG1.

They demoed a 4 tile version of Xe-HP processing at 42 TFlop FP32.

David Blythe is listed as the Chief Architect of the Xe GPUs, as well as on the gen11 GPU on Ice Lake. He came over from the MSFT directx group, not from AMD.

 

[watzupken]

Maybe at different watt settings

Perhaps since its not mentioned, but I think it is still very silly of Intel if they programmed their iGPU to have more power then their dedicated GPU. So I still feel the results are very odd.

 

[shady28]

Perhaps since its not mentioned, but I think it is still very silly of Intel if they programmed their iGPU to have more power then their dedicated GPU. So I still feel the results are very odd.

This is a Geekbench OpenCL score. That has to do with using the GPU as a compute processor, and little to do with how it performs as far as FPS in a game. e.g. OpenCL is also used to program FPGAs and DSPs - which (can) have nothing to do with graphics. For most folks, the OpenCL score is irrelevant. It might be useful enough to g ive you 'ballpark order of magnitude' type of performance on a GPU for graphics - but an order of magnitude is +/- 1,000%.

The Xe iGPU is already known to match the MX350 in games. I haven't looked it up but I would bet that the OpenCL score of the iGPU is similar to their DG1 dGPU. Again, it is an irrelevant score for 99.9% of users and is pretty disconnected from gaming performance in terms of avg FPS / 1% lows and so on.

So basically this article is more clickbait for the unwary.