Monday, September 19th 2022
Intel Arc A770 Overclocks Up to 2.70 GHz on Stock Cooling, with Minimal Effort
In its latest video presentation dealing with the reference board design and overclocking architecture of the Arc A770 Limited Edition graphics card, Intel revealed that the cards should be "monster overclockers," and that they've been able to get their randomly selected card to run at 2.70 GHz (up from 2.10 GHz reference), without the need for custom-cooling, just by using the overclocking controls on the Arc Control software. The cooler has a noise output of up to 39 dBA, and even with the overclocked GPU, Intel claims, the temperatures never crossed the 80-90 °C range. The GPU power was claimed to be around 228 W.
Intel clarified that the "GPU Clock" advertised with the A770 is the guaranteed clock-speed sustained by the GPU at least 50% of the time, even on the "least performing" silicon. The actual clock will vary around this point. This is represented as a bell-curve on top of the voltage-frequency curve of the GPU. There are two ways to go about increasing the performance of the GPU—increasing the voltage, which would increase the clock residency (sustainability of elevated clock-states); and by increasing the frequency itself. Both of these can be accomplished using Arc Control.
Source:
Intel Graphics
Intel clarified that the "GPU Clock" advertised with the A770 is the guaranteed clock-speed sustained by the GPU at least 50% of the time, even on the "least performing" silicon. The actual clock will vary around this point. This is represented as a bell-curve on top of the voltage-frequency curve of the GPU. There are two ways to go about increasing the performance of the GPU—increasing the voltage, which would increase the clock residency (sustainability of elevated clock-states); and by increasing the frequency itself. Both of these can be accomplished using Arc Control.
21 Comments on Intel Arc A770 Overclocks Up to 2.70 GHz on Stock Cooling, with Minimal Effort
My guess is that they will wait until after Nvidia's GeForce 4090 launch (expected this Tuesday, September 20) to reveal a date.
Kudos to team blue for this but the question still remains the same - When it ll be out?
2.7Ghz or not, the arch itself is still weak compared to Nvidia or AMD.
only downside is the higher power, the less efficient and completely new drivers that need ALOT of work. Buy a ARC today and DX 9 ~ 11 games are'nt native.
Writing graphics drivers is very hard. Both the NVIDIA and AMD kernel mode drivers have more lines of code than the Windows kernel itself.
Intel needs time to write a proper graphics stack. Alchemist is excellent hardware for a first generation product, Battlemage should perform significantly better and by Celestial I think they will have largely achieved parity with AMD in the driver performance front, though I expect the Radeon and GeForce cards to still be faster hardware overall, unless Intel greatly increases their research and development budget for GPU division.
Well, the good news is they can't possibly bore us with anymore press releases before an actual card materializes right...RIGHT?
That won't happen tomorrow but shortly. My guess is that the Intel senior management team will review what Nvidia announces tomorrow at their GTC event and finalize the A770 launch schedule and pricing accordingly.
There is a chance that there are also shipping boxes of individual cards sitting in Intel marketing offices around the globe, ready to be shipped out as review samples to PC hardware writers who would be bound under NDA terms to keep their mouths shut until the press embargo is lifted.
Depending on the Intel marketing team's strategy, there may be additional units destined for key social media influencers and streamers as well. Based on the sheer volume of teaser A770 marketing slides, it would not be a surprise if Intel's marketing budget included some cards to hand out as well as the typical schwag (t-shirts, caps, hoodies, commuter cups, etc.).
And then the announcement for Battlemage.
The hardware actually looks to be capable of between 3060 Ti and 3070 performance levels. The drivers just don't have the benefit of 25 years of optimizing like Nvidia / AMD. It's just going to be a real inconsistent experience, where one game runs like a 3050 and another game runs like a 3060 Ti.
I think these cards will age well though. With the crazy power draw on the new Nvidia cards, and the exorbitant pricing on the midrange Ampere cards, this or AMD (6600, 6600XT, 6650XT, 6700XT) seem to be the two best options for midrange buyers given the A750 and A770 are 225W cards.
AD will hit that clock and rDNA 3 definitely will as a base clock by the sounds of it.
I just hope they're priced aggressive verses AMD, and I still hope to buy one but dammmmmmmnnnnmnnnnmmmmn that's not a hype train, It's The hype train.
Now about that release date, wtaf.
remember the Maxwell GTX 460? it needed to be castrated to 675 MHz, in-order to hit 120w tdp! thy all universally hit 800 at stock voltage - the problem here is, its a decade later, and Intel is FINALLY launching with something similarly broken
There is some sense in releasing a product that is at its maximum v/f efficiency curve at some market segments, for example, the midrange that the GTX 460 aimed to service when it was a current generation product. The flagship such as the A770 isn't it, though, given the market conditions and the performance that will be akin to the RTX 3060's, I don't think it is too much of a crime. :oops:
I hope they have till Celestial because also Battlemage will not be very good, the architecture needs redesign and the time gap is not that great, in the same process they have to be max +10% in die size and max 10% less power efficient, while being equivalent performance wise in synthetics and -10% at new DX12/Vulcan games, this is where they become competitive, till then they will just have to sell with very little margins.
If you compare 6nm Navi33 with A770 it falls way short of these prerequisites.
But the industry needs a third GPU player badly imo!