Intel Iris Xe DG1 GPU from ASUS Tested

[Uskompuf]

Intel has been working with its partners to release the Intel Iris Xe DG1 GPU for use in prebuilt computers. ASUS is one of these partners along with GUNNIR who have both announced versions of the card. The ASUS DG1 was the first of these cards to release as part of the "Gamer Xtreme Gaming Desktop" from CyberPowerPC in May. The ASUS DG1 found in that system has recently been tested by ETA PRIME in various games and synthetic benchmarks. The DG1 is only compatible with select motherboards so the card was tested with the Intel Core i5-11400F and ASUS PRIME H410M-A/CSM motherboard included in the computer.

The ASUS DG1 features 80 Execution Units for 640 shading units running at a speed of 1.5 GHz, along with 4 GB of LPDDR4X memory running at 4266 MHz. The card only draws 30 W so features a fully passive design with no additional power connectors. The ASUS DG1 scored 1630 points in the 3DMark Time Spy test, 5837 points in Fires Strike, and 21313 points in Night Raid. The card was also tested in a variety of games including Forza Horizon 4, Injustice, Overwatch where it managed 1080p low 60 FPS performance. These performance numbers are comparable to a Radeon HD 7870 or a GTX 760 while drawing significantly less power. You can view the complete video from ETA PRIME testing the card below.





Video

View at TechPowerUp Main Site

 

[tabascosauz]

ETA Prime claims he does a lot of APU benching and that 1630 graphics score is "on par" with the 5700G with 2100MHz Vega 8.........if that's really the case, that just reflects poorly on his ability to extract performance out of Renoir/Cezanne? lmao

According to him, 4.5GHz 5700G / 2400MHz Vega 8 / 4400CL19 Viper Steel..............loses to a stock 4650G / 2100MHz Vega 7 / 4400CL19 Vipers @ 4200CL16??

Edit: 5082 vs. 5580 in Fire Strike and 21029 vs. 22729 in Night Raid as well, yeah okay, can we have someone other than this guy bench these cards? Clearly this guy has no idea wtf he's doing.



So I come out on top in Time Spy, lose to the DG1 in Fire Strike, and win in Night raid - I take it that DG1 is basically supposed to be a GT1030 competitor that's slightly faster, then? I hope Intel gets their act together in providing decent drivers, this could be an interesting card for memory benching or for SFF if it comes in low-profile form.

 

[Sempron Guy]

ETA Prime claims he does a lot of APU benching and that 1630 graphics score is "on par" with the 5700G with 2100MHz Vega 8.........if that's really the case, that just reflects poorly on his ability to extract performance out of Renoir/Cezanne? lmao

According to him, 4.5GHz 5700G / 2400MHz Vega 8 / 4400CL19 Viper Steel..............loses to a stock 4650G / 2100MHz Vega 7 / 4400CL19 Vipers @ 4200CL16??

Edit: 5082 vs. 5580 in Fire Strike as well, yeah okay, can we have someone other than this guy bench these cards? Clearly this guy has no idea wtf he's doing.

View attachment 204407 View attachment 204408

I take it that DG1 is basically supposed to be a GT1030 competitor that's slightly faster, then? I hope Intel gets their act together in providing decent drivers.

APUs as always are mem bandwidth starved. It's a no brainer why your set-up with a 4200 cl16 mem kit outperforms his set-up. And btw he buys his stuff to test it's not like it's being sent to him. Maybe he is not into mem tweaking and so as most of the mainstream consumers.

 

[tabascosauz]

APUs as always are mem bandwidth starved. It's a no brainer why your set-up with a 4200 cl16 mem kit outperforms his set-up. And btw he buys his stuff to test it's not like it's being sent to him. Maybe he is not into mem tweaking and so as most of the mainstream consumers.

Did you even read what you quoted? We have the same sticks, timings are more or less irrelevant for graphics tests when he's got a +200MT/s mem freq advantage, a 64SP bigger GPU core, and a +125MHz core clock. Even at XMP 4400CL19, the bandwidth advantage is clear and DRAM latency is not that far behind.

I can forgive lower scores on a stock APU with stock CPU/GPU clocks and a 3200/3600 kit that most people would buy, but that 5700G is anything but stock. Curious to see if those disappointingly low scores carry over to the DG1 if anyone else can get their hands on that card.

 

[owen10578]

Soo just buy a 5300G/5600G/5700G and use the Vega graphics instead? That's pretty disappointing, I was expecting at least GTX 1050Ti performance.

 

[watzupken]

To be honest, low end cards like this have very little point to exist when APUs are getting increasingly powerful and have the ability to support up to 3 displays at one go. Unless one requires even more display support by the GPU, I am not sure why pay extra for very little benefits. Correct me if I am wrong, but the worst part about this GPU is that it needs to go with an Intel based system.

 

[Legacy-ZA]

Soo just buy a 5300G/5600G/5700G and use the Vega graphics instead? That's pretty disappointing, I was expecting at least GTX 1050Ti performance.

Well, let's consider; my GTX660Ti is about 10% slower than a GTX1050Ti, and the 660Ti uses around 125W of power, then, you look at this card from Intel that performs around the same performance as the 660Ti. Is that 10% performance extra really going to mean that much to you? I am sure the Intel GPU will be more power-efficient though.

Anyways, I think the iGPU from the 5300G/5600G/5700G/ is going to be the better option, assuming you pair it with high clocked RAM with tight timings.

I am also really excited to see what the 3DV tech will do for iGPUs on the next-gen 5nm process node (AM5) and integrated RDNA2 instead of VEGA.

 

[rutra80]

What I find very important, is that drivers do seem to work decently - everything looks smooth and stable. Scores are decent, it's passively cooled, what else do you want? If DG2 will be like RX590 and DG3 like 3070 with availability and decent prices, I'm sold.

 

[ExcuseMeWtf]

But will it mine as poorly as it should?

 

[Vya Domus]

I wasn't expecting much but damn if this is true this thing is rubbish.

 

[DeathReborn]

A 7870 paired with a 4770K at stock gets a graphics score of 1687 in Time Spy. I am also fairly sure if you shrunk a 7870 to 7nm you could get to 30W power consumption, at higher clocks to boot.

 

[lexluthermiester]

ETA Prime claims he does a lot of APU benching and that 1630 graphics score is "on par" with the 5700G with 2100MHz Vega 8.........if that's really the case, that just reflects poorly on his ability to extract performance out of Renoir/Cezanne? lmao

According to him, 4.5GHz 5700G / 2400MHz Vega 8 / 4400CL19 Viper Steel..............loses to a stock 4650G / 2100MHz Vega 7 / 4400CL19 Vipers @ 4200CL16??

Edit: 5082 vs. 5580 in Fire Strike and 21029 vs. 22729 in Night Raid as well, yeah okay, can we have someone other than this guy bench these cards? Clearly this guy has no idea wtf he's doing.

View attachment 204407 View attachment 204408

So I come out on top in Time Spy, lose to the DG1 in Fire Strike, and win in Night raid - I take it that DG1 is basically supposed to be a GT1030 competitor that's slightly faster, then? I hope Intel gets their act together in providing decent drivers, this could be an interesting card for memory benching or for SFF if it comes in low-profile form.

You're being a tad-bit harsh. While ETA makes the occasional error in methodology, he general gets testing right and in this situation the testing done was reasonable and in line with general testing standards. The issue I have with that review was not his methods but his conclusion, which I feel was completely off the mark. He seemed to miss the point of that card existing.

The DG1 is an excellent first entry in the budget market. It hit a sweet-spot in performance/price. While there is room for improvement, this sample is only an OEM model. My guess is that the retail models will be a bit more fleshed out and will likely have better specs/performance. Keep in mind the DG2 is inbound as is the DG3. This is Intel getting started, not finishing up.

 

[Fourstaff]

Hope this becomes a good budget card and relieve some pressure off TSMC.

 

[RedBear]

To be honest, low end cards like this have very little point to exist when APUs are getting increasingly powerful and have the ability to support up to 3 displays at one go. Unless one requires even more display support by the GPU, I am not sure why pay extra for very little benefits. Correct me if I am wrong, but the worst part about this GPU is that it needs to go with an Intel based system.

It's even worse than that, it only goes with selected Intel systems and it's available only to OEMs, ASUS has confirmed only two motherboards compatible with this GPU, the PRIME H410M-A/CSM (used in the system tested by ETA Prime) and the PRO B460M-C/CSM. I guess that the point with this GPU is probably providing a basic graphic solution for OEMs that need/have to use processors without integrated graphics, like the i5 11400f used in the tested system, because of supply issues or whatnot.

 

[chrcoluk]

Well remember low end cards seem to have been abandoned by Nvidia now, so this gives something to to those with weak systems, or with very low budgets.

I dont remember seeing a nvidia 2030 or 3030 product been released.

 

[lexluthermiester]

I dont remember seeing a nvidia 2030 or 3030 product been released.

That's because there isn't. The GTX 1650/1660 are NVidia's budget level cards.