News Intel GPUs - Falcon Shores cancelled

[DAPUNISHER]

Fortnite has a replay feature that is evidently useful for testing GPUs in a repeatable manner. Daniel Owen uses it. But I don't know if it loads the CPU the way live gameplay does. Fortnite will at times, 100% overload the 9600 Hardware Canucks used for testing.

 

[mikk]

https://twitter.com/x/status/1876845796110295247

FYI he predicted BMG is coming so...
I saw someone on Twitter post it

He removed his postings. It could mean he found out it was wrong or he didn't like the attention he got from these postings. We have to wait for real infos. I think B770 on N4 in H2 2025 could be quite good and worth a try. Maybe they realized it is better than expected even against the new gen from Nvidia/AMD.

 

[511]

He removed his postings. It could mean he found out it was wrong or he didn't like the attention he got from these postings. We have to wait for real infos. I think B770 on N4 in H2 2025 could be quite good and worth a try. Maybe they realized it is better than expected even against the new gen from Nvidia/AMD.

Well it is certainly efficient for AI Inference/training and they can get good margin out of it

 

[Win2012R2]

Maybe they realized it is better than expected even against the new gen from Nvidia/AMD.

They've got real opening now.

 

[DavidC1]

They've got real opening now.

Assuming they can get much improved low overhead drivers by that time. And for loyalty's sake the improvement has to come to Alchemist too. If they de-prioritize driver development significantly on Alchemist which is still a very relevant generation, they'll never get to the point where people like Intel GPUs and ultimately it'll hurt their bottom line.

There are many things in product development where the effects on financials are immediate and where the MBA types don't get. Things like loyalty through long term support, caring about seemingly niche segments but have hardcore user base, addressing problems quickly.

No shortcuts exist.

 

[Win2012R2]

Assuming they can get much improved low overhead drivers by that time.

Not necessarily - if they get bigger chips next year, then they can target higher level of performance tier that is much more reasonable to expect to be paired with faster CPUs.

 

[DigDog]

fyi some Ray-Tracing benchmarks

 

[DAPUNISHER]

Is that a game bug that is says DLSS performance instead of XeSS?

 

[DigDog]

Is that a game bug that is says DLSS performance instead of XeSS?

no idea. that part of the UI is not selectable, the only thing i can change is the resolution, or the setting to Very High.

I note that when i ran the Very High benchmark, a model of a bottle that gets broken by a bullet did not render at all - the bullet was just flying through the air.

 

[DigDog]

i note a clear error in this benchmark. Whether this is a card error or a software error, i do not know.

this is a free bench on Steam https://store.steampowered.com/app/1409670/Bright_Memory_Infinite_Ray_Tracing_Benchmark/

running on High, the second "scene" (a few seconds after start) is a glass bottle that is shattered in slow-mo by a bullet.
on Very High, one run the bottle was invisible, on the second run the bottle was untextured.

 

[igor_kavinski]

An Amazing 4K Virtual Walking Tour of the Titanic

Titanic: Honor and Glory is an incredible game that reconstructs every facet of the ill-fated Titanic using Unreal Engine 5 by Epic Games.

laughingsquid.com

Run that with max settings. Runs at 100+ fps on the RX 6800.

 

[DigDog]

An Amazing 4K Virtual Walking Tour of the Titanic

Titanic: Honor and Glory is an incredible game that reconstructs every facet of the ill-fated Titanic using Unreal Engine 5 by Epic Games.

laughingsquid.com

Run that with max settings. Runs at 100+ fps on the RX 6800.

nope.
i'm headed to the train station and tomorrow night i will be in Rio until the end of the month.
see you later suckas.

 

[DAPUNISHER]

i note a clear error in this benchmark. Whether this is a card error or a software error, i do not know.

this is a free bench on Steam https://store.steampowered.com/app/1409670/Bright_Memory_Infinite_Ray_Tracing_Benchmark/

running on High, the second "scene" (a few seconds after start) is a glass bottle that is shattered in slow-mo by a bullet.
on Very High, one run the bottle was invisible, on the second run the bottle was untextured.

ARC driver issue. It's fine on AMD.

Bench is funny. Using Shotokan effectively on a heavily armored opponent.

 

[igor_kavinski]

Bench is funny. Using Shotokan effectively on a heavily armored opponent.

People like him make me feel like a total loser: https://80.lv/articles/bright-memory-infinite-solo-indie-game-development-with-reallusion/

 

[DigDog]

ARC driver issue. It's fine on AMD.

Bench is funny. Using Shotokan effectively on a heavily armored opponent.

ok so Intel still has a lot of work ahead; who knew.

 

[DAPUNISHER]

ok so Intel still has a lot of work ahead; who knew.

🤣 ARC has definitely come a long way already. It will get there if they are allowed to keep at it.

 

[511]

Well ARC IP is getting more polished over the years and it will continue to do so there is no doubt about it Xe2 was 70% perf/XE core and Exists says Celestial is even bigger like 100% perf/XE?

 

[Heartbreaker]

HWUB retests in a variety of conditions with 5600 vs 9800X3d:

 

[511]

HWUB retests in a variety of conditions with 5600 vs 9800X3d:

Still no Intel Mid range CPU

 

[Meteor Late]

Yeah the value from this GPU has just vanished, well not entirely because of the VRAM advantage, but from a performance standpoint it is slower with older CPUs, also he pointed out that even with 9800X3D, when testing 50 games, the average performance turns quite a bit worse than with the initial game selection he used, from 14% better to just 5% better.
I mean, that's why Intel priced it as low as they did, I am sure without this driver overhead issue they would've priced it at idk if 300$ but maybe 269-279$ or something like that.

 

[DAPUNISHER]

Still no Intel Mid range CPU

Are you wondering if there is some secret sauce with all team blue test systems?

 

[DavidC1]

Not necessarily - if they get bigger chips next year, then they can target higher level of performance tier that is much more reasonable to expect to be paired with faster CPUs.

Yea, and its rumored to be 70% higher performance.

It shows noticeable overhead using 9800X3D, which all the big initial reviews were based on. How would a potential 70% improvement help in that regard? 70% improvement would mean the bottleneck would shift from mostly 1080p to 1440p and even 4K!

For bottleneck to be considered "resolved" you would need the original pie-in-the-sky claims of Intel's Royal Core having 2x the perf/clock over the cores in Alderlake.

Are you wondering if there is some secret sauce with all team blue test systems?

Yea I think people are making this argument based on things like E cores existing, and considering limited Intel CPU benchmarks which show poor performance as well, and that the drivers only really care about fast single thread CPUs it would not care one bit about E cores.

And Intel platforms are even more expensive to upgrade for budget conscious folk because you have to buy CPU and motherboard. No matter what AMD CPUs are saving Intel's GPUs right now.

I mean, that's why Intel priced it as low as they did, I am sure without this driver overhead issue they would've priced it at idk if 300$ but maybe 269-279$ or something like that.

Maybe they would have had B770 out right now without the overhead issue. In that hypothetical scenario, the B580 could have been priced at $229 because the B770 is a higher performance tier and being able to play in the $400 market is higher margins and profits.

So you get the double whammy benefit of B580 stealing marketshare with a really attractive price while B770 is an icing on the cake that actually brings in real profits.

 

[DavidC1]

Bionic_Squash on X says the driver versions for Battlemage is little behind Alchemist hence the separate drivers.

Intel ARC reddit shows based on drawcall test the B580 has more overhead in DX12 and Vulkan compared to Alchemist. I don't know if this is true, but what Bionic has said could be related to it? Don't know if couple of month difference is enough to have that much of an impact.

 

[DrMrLordX]

Do we have any data on how B580 performs with extra cores? For example, 5900X vs 5600X?

 

[Thunder 57]

Do we have any data on how B580 performs with extra cores? For example, 5900X vs 5600X?

Sounds like a question for @DAPUNISHER . My guess would be it is more of an IPC than a thread thing but I have no idea.