Intel Xe HP "Arctic Sound" 1T and 2T Cards Pictured

[Vayra86]

Nice specs.

STILL No performance metric. We already know for many years that Intel can make overpriced, non-competitive GPU designs. That they abandon shortly after. All I am seeing here is a continuation and 'making it scalable' of their eternal IGP. So yay, you have lots of EUs now and you need pricy memory to feed it. It is nowhere near the level of refinement of the competition. All I smell here is something akin to company XYZ coming out with their own version of an ARM soc.

The fact that Raja straight up jumped on four tiles as a starting point speaks volumes. It smells like that strange junkyard CPU range Intel is trying with four dies glued together. A slight hint of bullshit coupled with smoke from electrical fires and some of Raja's hair. Their 2T is already running on below optimal clocks and 4T will be even worse if they plan on burning less than 1KW per unit.

Next!

 

[zenlaserman]

According to GPU-L this thing puts down 281 GP/s @ 2200MHz, with 563 GT/s. 18 TFLOP FP32, 4.5 TFLOP FP64

I dunno why anyone would hate on that. That is a damn solid base, even if it's allegedly 275W. One could UC/UV this thing and it would still be a beast.

More GPUs is always good! About time, Intel lol

 

[Valantar]

According to GPU-L this thing puts down 281 GP/s @ 2200MHz, with 563 GT/s. 18 TFLOP FP32, 4.5 TFLOP FP64

I dunno why anyone would hate on that. That is a damn solid base, even if it's allegedly 275W. One could UC/UV this thing and it would still be a beast.

More GPUs is always good! About time, Intel lol

That sounds decent, though how TFLOPs translate into gaming performance is always the real kicker. Are they on a level with ... GCN? RDNA(2)? Pascal? Turing? Ampere? Above, below all of these? The current range is pretty wide already, so there's no telling where this will fall. Might be good for compute though.

 

[zenlaserman]

That sounds decent, though how TFLOPs translate into gaming performance is always the real kicker. Are they on a level with ... GCN? RDNA(2)? Pascal? Turing? Ampere? Above, below all of these? The current range is pretty wide already, so there's no telling where this will fall. Might be good for compute though.

It's not too hard to throw a rough guesstimate at the card's general gaming performance by using the GP/s and GT/s figures. Of course it takes proper drivers to get the most out of that (and we all know Intel's track record on drivers for gaming haha). It wasn't until the Radeon VII -> 5700XT that AMD exceeded the 100 GP/s mark reliably, while nVidia's 1070 was hitting that mark. Today, the 3090 needs OCing to hit 200 GP/s, the 6900XT is about 250 GP/s at 2GHz.

Intel's Xe 2T card appears to perform identically clock for clock to the 6900XT, at least when it comes to pixel and tex math. Gaming performance remains to be seen ofc, but the hardware is obviously ready to brute force something. I remember way back when GPUs hit 1GP/s and then 10GP/s, I've done a helluva lot of PC gaming on sub-10GP/s cards lol. 50GP/s today is about medium-high 1080P settings on most games.

 

[regs]

Why 4+4 pin EPS?

Why not use PCI-E 8 pin, its more prevalent on PSUs?

Why on Earth we need have 3 different 12 V connectors? EPS, PCIE, ATX12VO... Why not to make a single type connector for everything?

 

[Valantar]

Why on Earth we need have 3 different 12 V connectors? EPS, PCIE, ATX12VO... Why not to make a single type connector for everything?

ATX12VO is a replacement for the 24-pin ATX connector, and carries a PS_ON signal, 12VSB standby power (like the 5VSB of current PSUs), and a PWR_OK signal. Those are kind of essential, woudln't you say? For things like shutting off your PC without pressing the power button (which you had to do in the old pre-ATX days), allowing the PC to hibernate (or charge devices while off), or not turning on unless the PSU is okay? And you'd agree that your GPU doesn't need those, or your motherboard doesn't need two sets? It makes perfect sense for there to be one standard for "this is all the stuff the PC needs to work", and another for "this is extra power delivery for stuff that needs it".

There's absolutely an argument to be made for the redundancy of PCIe power and EPS, but sadly those two standards weren't developed concurrently - the PCIe power connector is a PCI SIG standard, while EPS is an Intel ATX standard. EPS is far superior though - 8-pin PCIe for some reason only adds ground wires compared to the 6-pin (why not 1 ground and 1 12V? Who knows?), and EPS connectors and wires are rated for >300W, meaning they need to conform to a higher quality standard for pins and wires.

ATX12VO is actually really smart, as the base 10-pin connector is specced to deliver 216-288W of power (6-8A @12V across three 12V pins, while the 24-pin ATX connector only carries two 12V pins), and there's an optional supplementary 6-pin connector for boards that require lots of power (many PCIe slots, for example) which is identical to 6-pin PCIe (though oddly enough specced for the same 216-288W - PCIe power connectors are clearly specced very, very low, given the 75W rating of the 6-pin (that's just slightly above 2A per pin! There's little doubt the exact same connectors could easily carry 2x the power with a slightly stricter spec). This will go a long way towards reducing cable clutter in future PCs. Most motherboards, particularly those smaller than ATX, can get away with just the 10-pin + EPS, with larger boards gaining much more robust power delivery than today through the extra 6-pin. It's a win-win all around. I don't ever see the PCI-SIG adopting EPS as a replacement for PCIe power connectors though. It's more likely there'll be a new standard like Nvidia's new 12-pin or something similar.

 

[regs]

Custom pins is quite a trivial stuff to deal with. Type C deals with tones of different formats - USB, TB, DP, HDMI etc etc.

 

[Solaris17]

Very nice looking GPU. Half height, low profile, passively cooled. Great for small form factor computers. If this product reaches retail Intel will have a sale from me.

No its not. These are meant for 1 and 2u servers with high static pressure design. If you put these in a consumer case they would bake themselves dead.

 

[Valantar]

Custom pins is quite a trivial stuff to deal with. Type C deals with tones of different formats - USB, TB, DP, HDMI etc etc.

... so you want your power supply to use logic to switch what the various output pins do, and negotiate voltages with receiving devices? Yeah, sorry, that's a terrible idea for something carrying high current, or even for a power-only cable standard. That would necessitate stuffing your PSU full of high-current relays (which aren't that expensive, but they're huge, loud, and wildly impractical) as well as making PC-PSU connections extremely complicated. You'd require logic on each end to negotiate the suddenly entirely custom pinout of the plug, and more than likely identifying logic in the cables to ensure they can handle the current that's being sent through them and not catch fire - like "high current" USB-C has (btw, high current USB-C is 5A@20V, while something like a 650W PSU can typically output 650/12=~55A of power on its 12V rail). The PSU would need that logic for every single output connector. That's a lot of new chips. This would drive up the price of PSUs, motherboards, GPUs and anything else that needs power fed into it quite significantly. And for what benefit? Being able to plug any cable into any connector? Yeah, sorry, I don't see the benefit. I'd rather take a compact, efficient, silent 12VO PSU than some relay-packing, logic-controlled but supposedly idiot-proof "one cable does it all" solution. Not to mention the horror show that would happen when the logic chips inevitably fail, or even some peripheral circuitry causes them to misread the incoming signal ("oh, yeah, a tiny capacitor in my PSU failed, so my PSU reversed the voltages on its outputs and fried every component in my PC. But at least I didn't have to plug in different kinds of cables!").

What a low-power data-first cable standard can do cheaply, easily and safely, especially one with a single bus voltage at any given time, and what a high-current, power-first, multi-voltage cable can do cheaply, easily and safely ... are very, very different things. I would be all for abandoning PCIe power cables entirely and implementing EPS as the new universal standard for anything needing more power than the motherboard can provide, but anyting else is making things way more complicated than what it's worth.