Question Intel's current woes and the low end of the desktop CPU market

[inquiss]

They screwed up!

APUs are the name of the game now.
Gorgon-Halo is next.
Then, we have a new stick.

Oh noes, RDNA2 was ramping dang well.

Words words words.
They have to ship.

Words.
You know you can just ship.

There's a halo before medusa halo? What does it have?

 

[regen1]

They screwed up!

But by your logic they should still just ship even if little to no adopters, right ?

APUs are the name of the game now.

Irrelevant to the point discussion also in terms of existing time line and years gone.

They have to ship.

Words.
You know you can just ship.

You are just arguing for sake of it. Just look at discontinuation data and everything else. Shipping turds doesn't work just for the sake of shipping.

 

[poke01]

There's a halo before medusa halo? What does it have?

not RDNA 5 or 4 so who cares.

 

[adroc_thurston]

But by your logic they should still just ship even if little to no adopters, right ?

YES. They they do and yes they will.

Irrelevant to the point discussion also in terms of existing time line and years gone.

Years ain't gone at all.

Shipping turds doesn't work just for the sake of shipping.

It does work.
If Vega20 worked, so could anything else.

There's a halo before medusa halo? What does it have?

It's just an stxH refresh with two new SKUs.

 

[regen1]

YES. They they do and yes they will.

No they didn't, no RDNA4 notebook SKUs. Stop inventing stuff.

Years ain't gone at all.

All these years of Nvidia notebook dGPU domination is plain to see and again no RDNA4 notebook SKUs. AMD didn't agree to your "just ship".

If Vega20 worked, so could anything else.

No, not always, for the umpteenth time. No silly comparisons.
Seems like you had no idea about Rialto perf and other issues but you still have to defend your silly take.

 

[adroc_thurston]

AMD didn't agree to your "just ship".

Yea they did.

No, not always, for the umpteenth time. No silly comparisons.
Seems like you had no idea about Rialto perf and other issues but you still have to defend your silly take.

Rialto had labs as customers.
Now Intel doesn't have labs as customers.
Can I make it simpler for you? Idk, you're struggling to read.

 

[inquiss]

YES. They they do and yes they will.

Years ain't gone at all.

It does work.
If Vega20 worked, so could anything else.

It's just an stxH refresh with two new SKUs.

The whole lineup is refreshed or like just the two new skis make it a refresh? What's it bring bedside rebranding?

Any actual improvements?

 

[511]

The whole lineup is refreshed or like just the two new skis make it a refresh? What's it bring bedside rebranding?

Any actual improvements?

Name +100

 

[regen1]

Yea they did.

In the original context of shipping RDNA4 notebook SKUs, no they didn't yet. Again don't change context.

Yea they did.

Rialto had labs as customers.
Now Intel doesn't have labs as customers.
Can I make it simpler for you? Idk, you're struggling to read.

So did PVC, if you had read. Doesn't make Rialto's position anything near good enough. For whatever reason you seem to ignore all the problems associated with Rialto.
Think I have said and cited enough about PVC and Rialto's struggle
but if you are hel-bent on defending your stance on Rialto just for sake of it let's just agree to disagree.

 

[adroc_thurston]

In the original context of shipping RDNA4 notebook SKUs, no they didn't yet. Again don't change context.

Well CES is next with some goodies you see.

So did PVC, if you had read. Doesn't make Rialto's position anything near good enough. For whatever reason you seem to ignore all the problems associated with Rialto.

This is cope.
If you have customers, you ship. Intel had committed customers for Rialto.

Think I have said and cited enough about PVC and Rialto's struggle

I know the problem with Level Zero enablement on Rialto.
You still ship.

 

[regen1]

Well CES is next with some goodies you see.

You forgot haven't shipped "yet". If they were that confident in having good adoption it would have already launched considering the failure
RDNA 3 notebook SKUs adoptions were. We will see how well it does if it finally launches. Neither Intel and AMD are practically in game with notebook dGPU markets. For iGPU it's very different.

This is cope.
If you have customers, you ship. Intel had committed customers for Rialto.

Only cope here is your "you ship whatever dud" logic. Few customers means nothing. Anything can have that and no luck this time with big contracts(Aurora) like PVC got beforehand. If Rialto had any significant wins like Aurora they would have launched it in however limited capacity.

If Intel brings a hypothetical improved Alchemist+ dGPU with h/w and s/w issues to compete with Rubin/RDNA5 launch it could still have few customers doesn't mean it should be launched.

I know the problem with Level Zero enablement on Rialto.
You still ship.

Software-stack issues weren't the only thing, just one of many. Shipping duds in a row(delayed) can cause more damage than just canceling the second one.

 

[adroc_thurston]

Few customers means nothing.

It means you can continue with the roadmap while iterating your s/w stack.

If Intel brings a hypothetical improved Alchemist+ dGPU with h/w and s/w issues to compete with Rubin/RDNA5 launch it could still have few customers doesn't mean it should be launched.

Client ain't DC. In client you sell the product, not the roadmap.

Software-stack issues weren't the only thing, just one of many.

It was the primary issue after losing the entire LZ team to Nvidia. Hardware-wise it was a simpler PVC with more oomph.

 

[DavidC1]

Intel's foremost problem is not sticking to a product. If you aren't going to stick to one, why release a lineup in the first place?

And AMD doesn't have GPU marketshare because they keep avoiding halo lineup for mediocre strategies like mid-end dies rather than aiming for Nvidia on the top end. Halo is what is on the mindshare for everyone. And of course you won't steal marketshare from Nvidia in one generation. You do that for years, and then people will start to pay attention to you. It has all to do with trust, execute for years, and you build trust. Won't happen with a middling part.

And both AMD/Intel copy Nvidia strategies for graphics, and be a fast follower at the best for things like DLSS and frame gen, rather than making halo and creating a feature that they can lead with. 1% and 10% marketshare is not a coincidence. Nvidia earned trust by executing year after year after year. You want to win? Then be a winner. How long before AMD took significant marketshare from Intel? Didn't happen in Zen 1, 2, or even 3. But it's happening now and kept accelerating every generation.

 

[DrMrLordX]

And AMD doesn't have GPU marketshare because they keep avoiding halo lineup for mediocre strategies like mid-end dies rather than aiming for Nvidia on the top end.

That was only true for the 9000-series (9070XT) of Radeon. 7900XTX was meant to compete as a halo offering and it didn't win as a top performer.

 

[poke01]

That was only true for the 9000-series (9070XT) of Radeon. 7900XTX was meant to compete as a halo offering and it didn't win as a top performer.

And the RX 5000 series

 

[DrMrLordX]

And the RX 5000 series

Vega and Vega20 were meant to be their halo GPUs in that timeframe (they also failed to top the charts).

 

[Josh128]

Wow.

 

[Thunder 57]

Intel's foremost problem is not sticking to a product. If you aren't going to stick to one, why release a lineup in the first place?

And AMD doesn't have GPU marketshare because they keep avoiding halo lineup for mediocre strategies like mid-end dies rather than aiming for Nvidia on the top end. Halo is what is on the mindshare for everyone. And of course you won't steal marketshare from Nvidia in one generation. You do that for years, and then people will start to pay attention to you. It has all to do with trust, execute for years, and you build trust. Won't happen with a middling part.

And both AMD/Intel copy Nvidia strategies for graphics, and be a fast follower at the best for things like DLSS and frame gen, rather than making halo and creating a feature that they can lead with. 1% and 10% marketshare is not a coincidence. Nvidia earned trust by executing year after year after year. You want to win? Then be a winner. How long before AMD took significant marketshare from Intel? Didn't happen in Zen 1, 2, or even 3. But it's happening now and kept accelerating every generation.

There is a problem with mindshare though. Outside of the 5090 and 5070 Ti the 5000 series is largely a poor choice. Newegg had a 9060 XT 8GB for $200 a few weeks ago. Their Nvidia equivalent will be a 5060 8GB (non Ti) at $200 in a few days. Here's the performance numbers:

At this level Nvidia's feature advantage is less important than raw performance. Yet we still have people like this:



That's not to say it isn't AMD's fault. If they had the mindshare, the sales would follow.

 

[mikegg]

There is a problem with mindshare though. Outside of the 5090 and 5070 Ti the 5000 series is largely a poor choice. Newegg had a 9060 XT 8GB for $200 a few weeks ago. Their Nvidia equivalent will be a 5060 8GB (non Ti) at $200 in a few days. Here's the performance numbers:



At this level Nvidia's feature advantage is less important than raw performance. Yet we still have people like this:

View attachment 134369

That's not to say it isn't AMD's fault. If they had the mindshare, the sales would follow.

Reasons why people still buy Nvidia's part:

1. DLSS tech is simply better than anything AMD has. Once you factor this in, the performance charts are a lot closer or even favors Nvidia. Add in Ray tracing and performance usually always favors Nvidia.

2. Nvidia GPUs are much more useful outside of gaming. Far more compatible with AI workloads. Learn CUDA programming. GPU rendering. Etc.

3. Nvidia tends to support their GPUs for much longer than AMD. For example, CUDA still works on 10 year old GPUs. Often ROCm doesn't support new AMD GPUs on day 1 and support for old GPUs drops fast. Another example is AMD recently dropping game optimization support for RDNA1 and RDNA2.

4. Nvidia GPUs tend to have better resale value, making the original purchase price worth it in the long term.

So yea, it is not as simple as raster performance per dollar. Nvidia GPU buyers aren't as stupid as AMD fans make them out to be. People buying AMD because they want to support the underdog are dumber than people buying Nvidia GPUs for brand name.

 

[Vikv1918]

Reasons why people still buy Nvidia's part:

1. DLSS tech is simply better than anything AMD has. Once you factor this in, the performance charts are a lot closer or even favors Nvidia. Add in Ray tracing and performance usually always favors Nvidia.

2. Nvidia GPUs are much more useful outside of gaming. Far more compatible with AI workloads. Learn CUDA programming. GPU rendering. Etc.

3. Nvidia tends to support their GPUs for much longer than AMD. For example, CUDA still works on 10 year old GPUs. Often ROCm doesn't support new AMD GPUs on day 1 and support for old GPUs drops fast. Another example is AMD recently dropping game optimization support for RDNA1 and RDNA2.

4. Nvidia GPUs tend to have better resale value, making the original purchase price worth it in the long term.

So yea, it is not as simple as raster performance per dollar. Nvidia GPU buyers aren't as stupid as AMD fans make them out to be. People buying AMD because they want to support the underdog are dumber than people buying Nvidia GPUs for brand name.

Can you show with any hard data what user % actually use their GPU for use-cases beyond gaming?

 

[511]

Can you show with any hard data what user % actually use their GPU for use-cases beyond gaming?

People doing blender/CAD or Adobe

 

[poke01]

Can you show with any hard data what user % actually use their GPU for use-cases beyond gaming?

Blender/3D rendering is next huge one beyond gaming

 

[mikegg]

Can you show with any hard data what user % actually use their GPU for use-cases beyond gaming?

No I don't. A lot of gaming nerds like to tinker and find things to do with their hardware. Vast majority of these gamers have no need for a 9950x or 9900x but they still buy them.

However, I personally would buy an Nvidia card to run local LLMs. They're far more compatible, get day one support from LLM vendors, run faster.

 

[adroc_thurston]

Blender/3D rendering is next huge one beyond gaming

Microscopic market.

 

[Thunder 57]

Reasons why people still buy Nvidia's part:

1. DLSS tech is simply better than anything AMD has. Once you factor this in, the performance charts are a lot closer or even favors Nvidia. Add in Ray tracing and performance usually always favors Nvidia.

2. Nvidia GPUs are much more useful outside of gaming. Far more compatible with AI workloads. Learn CUDA programming. GPU rendering. Etc.

3. Nvidia tends to support their GPUs for much longer than AMD. For example, CUDA still works on 10 year old GPUs. Often ROCm doesn't support new AMD GPUs on day 1 and support for old GPUs drops fast. Another example is AMD recently dropping game optimization support for RDNA1 and RDNA2.

4. Nvidia GPUs tend to have better resale value, making the original purchase price worth it in the long term.

So yea, it is not as simple as raster performance per dollar. Nvidia GPU buyers aren't as stupid as AMD fans make them out to be. People buying AMD because they want to support the underdog are dumber than people buying Nvidia GPUs for brand name.

You must've missed the part where I said:

At this level Nvidia's feature advantage is less important than raw performance.

1. DLSS is nice in that is has more support. Nobody is doing raytracing on a 5060.

2. A tiny market who already know what they need, and it isn't a 5060.

3. Reading too much into the "announcement" where AMD let an intern make the post. There are still drivers for GCN cards.

4. I would hope so since Nvidia costs more at launch.

People who make money with their GPU are going to be using an RTX 6000 or mayve a 5090. People who care about ray tracing are going to be using a 5070 Ti / 5080. A 5070 is probably fine for now but that memory limitation will bite it in the ass long term. I'd take a 9070 with 16GB. This is where they win the mindshare where people think it's an Nvidia only world and buy low end Nvidia even though all the do is game not realizing they maybe making a poor choice.