is Intel integrated graphics (non F cpu) worth it ?

[spacef]

Hi all !

I am budgeting a new PC to upgrade from win7, and I see people saying that intel iGpu + graphic card GPUs can work together ? (on youtube videos)
There are people saying they do and you may gain 5% performance in windows, videos, or "content creator" (i think it is about encoding/decoding) ?
Does it mean it shares tasks with the graphic card ?

The CPU I am looking at is a 12700 K , but the KF (F= no on-cpu GPus) is much cheaper.
If it was only for music, I would go for an F version (no cpu graphics), but I also use phostoshop, after effects or video software, why not streaming in the future

Did anyone knows if there is the slightest difference between graphic card only and graphic card+cpu graphics ? (my graphic card is an nividia GTX 960)

Thanks !


NB: potential rig
- i7 12700K
- 32 GB ddr5 4800 / CL 40
- Z790 Motherboard gygabite UD AX (because it is the only one with more than 4 sata connectors)
- Thermalright Phantom Spirit Air cooler (cheap, kills noctua or be quiet, but not sure it is durable)
may be an M2 pci-e 4 drive for OS.

The rest comes from the current PC (psu, case, drives, etc).

[nebelfuerst]

From my personal experience, I would not use two different GPU within one system. I own an acer laptop which has intel + radeon and it allows to switch them for each app. Sounds cool, but if it comes to opengl, a mess begins.
The GTX 960 is quite weak, compared to rtx 3xxx or rtx 4xxx. Using GPU for computing makes sense for the later GPUs with 8GB or more memory. ( some tasks mainly benefit from the faster GPU memory, not the GPU processing unit)
I own the 3080 & 3090, which are impressive accelerators, but you'll typically need a new PSU beyond 1000W to get a stable machine.

M2 NVME work well these days (they were problematic 10 years ago)
pci4 or pci3 doesn't matter, as the limitation of rate is on the OS and the onboard M2 controller.

[spacef]

So you mean they are not meant to "work together" as some people say ? There is no "dispatch" of tasks ?

And yes the graphic part is old but it will wait, I don't play games much and I am not sure that what I use would really be boosted.
I am also waiting to see how IA offers evolve for desktop, to run IA models locally in a year or two (not big brother AI, more like your own local GPTs..).

[fra77x2]

If i understand correctly

my laptops have integrated gpu. They are old but i don't play games that suck anyway. For my usage i find these good
my i3 core has hd3000 and the i5 has hd4000 both are 10 years old but i get 120 fps for my synthesizers with the hd3000. I think spending spoils the character of companies so i try not to. they should build a better character with some austerity

Now that i use shaders and modern opengl. I remember a friend of mine that has invested on orange apple shit when he saw my analyzers he felt a bit more stupid.
They were very fast in relation to what he had seen and i have programmed my analyzers with the old opengl that wasn't so fast

I updated my analyzers now with shaders and i get 120 fps on my i3 laptop. They are great. Also I don't have to watch advertisments while i use them because they run on my precious no genuine windows 7 tablet and i have programmed them with my own hands.

Intel is a nice company it doesn't oblige you to pray on its president everyday and you don't have to wear a hat that says intel on it.
You don't have to play in the google garden, you watch simple films not apple films and you can grow and become a mature guy instead of remaining a kid for the rest of your life as the corporate shit would prefer you to do by brainwashing you with every possible mean

Haven't tested with ai models because what i do on ai is not related to mass crunching of data and i do not use gpu for this

Also gpt makes you an idiot so if you are not already stay away

[valis]

I have no problems using AMD or Nvidia along with an Intel IGP. Older laptops are certainly more of an issue, but this is because many of those implement the chipset in a way that means you can't use the off the shelf (AMD/Nvidia) drivers and rely on the manufacturer for all driver updates--meaning you need to keep the software tightly in sync and it is often quickly out of date.

Don't worry about it on the desktop, and if you don't choose a top SKU Intel chip then the combined heat of using the CPU under heavy load plus an IGP isn't a big deal. Ie, you can rely on the onboard Intel iGP for Audio tasks under modern OSes just fine, and the later model CPUs have plenty of horsepower for low tier gaming and any intensive GUI for audio plugins.

The main reason I opt for BOTH personally (Nvidia + Intel iGP) is because Intel's codecs for encoding video are superior to Nvidia's, so using Intel quicksync to encode in Adobe & other encoders is lightning fast and gives good results. Nvidia is good for other tasks that use the gpu, and so I keep that in machines that will need it. Older Quadro cards are actually quite good, use less power than their gaming brethren, and have drivers that are optimized to avoid the DPC Latency spikes gaming GPU's give. So, when in doubt that's a good option (low tier quadro from the used market).

[Bud Weiser]

My Lenovo W540 and 541 workstation laptops,- running Win10 and 11 Pro x64,- both use iGPU of Intel i7 mobile quad (Haswell) and a dedicated graphics card in harmony.
No issues up to now.



Bud

[spacef]

Thanks a lot for all the feedback.
I think I will go for iGpu, for all you said and also because there is one situation it might be helpful... if the graphic card breaks it may give a few extra days to select a replacement, ie, not be in absolute emergency... And if it also helps with encoding/decoding that's a bonus.

ok that part is done. Now I am evaluating the need for a cpu bracket replacement to avoid bending the cpu or motherboard and keep the air cooler tight.... (eg thermalright "BCF" (Bending correct frame or "contact frame")....

ah, those geek videos about PCs... they certainly know how to make you panic and interested in stuff you might not need at all.... :-/( however, with the weight of current air cooler, hard to know. for sure)

And I agree GPTs are dumb, a bit hyped, but it will quickly evolve into stuff that are actually useful. I use Chat GPT a lot at work to create blog content. There is human work to add for quality and meaning, but I type much much less, and that's a benefit for me. MidJourney, Stable diffusion, video, and music are the stuff I am interested in (imagine feeding your whole song creations of the last xxx years, and it could suggest variations and arrangements ro existing tracks, or make whole tracks, based on your own style... I would use that, especially in moments when inspiration is hard to come, or when I am lost in selecting an instrument or another). Also interested in AI coding to translate plugins into native/vst etc but I think we are very far (you need to understand audio and sound, not just code). Now it looks more work correcrting mistakes than learning how to code in a clean manner. Guys like me who don't have the time to start such venture are very interested.

[valis]

I thought he was referring to General Partition Tree (GUID/GPT partition instead of MBR).

[fra77x2]

I have used gpt once to see what the hype was about, but i am prejusticed to not believe in it.

So I i typed something like socrates is a man. What is Socrates? And i got a reply that showed absense of any serious
programming work. No memory. No meaning retreival.

I spend some time with pasting questions that suposed to provide certain answers and to my surprise the responce was different
each time with a degrading slope on quality

So I consider gpt a glorified copy paste algorithm with some minor tuning for sentence creation. Most probably there are tailored results
so to raise the expectations and interest so to secure funds. There is potential but i am quite sure such systems are unmanagable
and out of my interests.

As helpful tool for blog writing well personaly i take somewhat serious the writing process as communicating with others. I would never miss a chance to be rude myself instead of using prebaked fake gpt responces to prove what?

Soon Ill finish my voice/ sound recognition system. It is based on convolutional networks and fft analysis. I am making it in Greek first but it is able to recognise sounds in general... birds, dogs the bike passing down the street.

For voice it is very simple:

first when the sound passes a threshold it gates it and starts analyzing
First it looks for vowels. Vowels can be very easily distinguised from their ffts. So for any vowels there is a responce where are these vowels positioned.
And then it looks behind the vowels for consonants, because a vowel may be preceded by a consonant or it may not.
So you get the consonants and the word is almost recognised.

For other sounds i have made a cute "memory" of my own invention that stores sounds somewhat how the human mind stores them
with repetition, importance, usage, correlation. The trick is to not search for identical patterns but make the patterns trigger their aknowledgement.
Its crazy really.

Do you know the little song:

Oh Daisy Daisy give me your answer, do
I am half crazy
oh for the love of you

[spacef]

I quite agree and disagree: I use it in a legal context (text analysis) and it is still very random sometimes, so it cannot be trusted for legal work. Also, it is true that replies degrade over iterations, hence the importance of making a good prompt from the start.
The problem is that all these AI are censored and politically correct, making them shallow and uninsteresting, with little critical mind. So humans are not replaced yet, but they may benefit from a tool to accelerate work and may be get different perspectives on various subjects.

-------
just for fun, I asked Chat GPT to make a reply to your comment

Here is what it has to say (at least it seems good at replying in a diplomatic tone - or trying its best to be "likable" , which is good and bad, depending on the situation).

It is the perfect linkedin corporate buttercoated bullcrap style of reply... machine clones of human clones ... I have nothing against that, it exposes what corporate mindset is...
---

Thank you for sharing your thoughts and experiences with GPT. It's understandable to approach new technologies with a degree of skepticism, especially when they're as hyped as AI models like GPT. Your analogy of GPT as a "glorified copy-paste algorithm" is an interesting perspective, though it might not fully capture the complexities and capabilities of these models. GPT and similar AI systems are designed to understand and generate human-like text by learning from vast amounts of data. While they're not perfect and can indeed produce varying results, the technology behind them involves more than just simple sentence creation or tailored results. It's about understanding patterns, contexts, and sometimes the nuances of human language, which is a monumental task.

Your point about the importance of genuine communication in blog writing is well-taken. Authenticity and personal touch in writing are invaluable, and AI should not replace these elements but rather serve as a tool that can assist in the creative process, perhaps by offering ideas, drafting, or editing support.

Your work on a voice/sound recognition system sounds fascinating and quite innovative, especially the approach to recognizing vowels and consonants through FFT analysis and creating a "memory" for sound recognition. It's impressive how you're applying convolutional networks to not just voice but also environmental sounds, indicating a broad understanding and application of AI technologies beyond text-based systems.

The project you're working on, especially with its focus on recognizing a variety of sounds and its potential applications, highlights the diverse possibilities of AI and machine learning technologies. It's a reminder that while tools like GPT have their limitations and areas for improvement, the field of AI itself is vast and filled with opportunities for creative and practical applications.

In the spirit of exploration and innovation that your project embodies, it might be worth considering the potential of AI technologies like GPT not as endpoints or solutions in themselves but as components of a larger toolkit. These tools can be refined, improved, and combined with other technologies (like your voice/sound recognition system) to create more comprehensive and effective solutions.

Again, thank you for sharing your insights and your exciting work. It's conversations like these that push the boundaries of what's possible and encourage a deeper understanding of both the capabilities and limitations of emerging technologies.

[fra77x2]

It is impressive isn't it?

Well not so much. Please let me deconstruct the responce a bit so to get an idea who is really this diplomatic interlocutor.

The trick is to throw ashes on the air so to cover the illusion. The main element is anthropomorphism. Mainly we see in the responces what we like to see, basically ourselves. But this is an illusion its like thinking that clouds in the form of dogs or something else is created by the artistic tendency of the cloud formation process instead of looking back to our minds and see that we only observe these forms because we have a strong pattern recognition process fitted inside our perception ability. And a malleable one. Its called fantasy.

Let's see what this thinking machine really understood from what I wrote. One thing I talked about is gpt itself and one of the tricks is that most dialogues with gpt are about exactly that. How real it is and how much it really understands. When I was 7 I played with Eliza one of the first chat bots that only used keywords to select responces and had a clever way to rearrange the input text so to mimick understanding. Some people experienced antropomorphism even with that. But it is obvious from the responces that this tool is tuned to advertise itself and has a vast amount of preconfigured sentences that it can present in any order without influencing the overall impression.

Also it is tuned on discussions about ai because its programmers are training it on their own conversations. It is a self fullfilled prophecy. It knows the questions knows the answers.

Do you know what I see here: the desperate need of its designers to succeed in their careers and get their true desire: money

They should only fool the politicians and the journalists and how difficult is that. Well not so hard when you are working on innovative technology.

In general the interesting part of the algorithm is the context retreival. What google does. Context is very important and can convey meaning but building a real ai means sophisticated design of ontological structures, forming and auto arranging of symbolic structures, etc. and other stuff that are related to philosophy, sociology, psychology and more. A single sentence can encompass so much infinity such systems have no chance. It is like waiting your coffee maker to become concious one day. It is harder than this.

[smits]

I have no issues combining AMD or Nvidia GPUs with an Intel Integrated Graphics Processor (IGP).

While older laptops pose compatibility challenges due to proprietary driver implementations, desktop setups handle this seamlessly.

With lower-tier Intel CPUs, the combined heat isn't concerning. I choose both for their strengths: Intel's superior video encoding and Nvidia's GPU tasks. Older Quadro cards offer efficient performance and optimized drivers, making them a reliable option from the used market when in doubt. gb whatsapp pro v17.00 download

[spacef]

Thanks for the input smits. Since last time, I've seen several benchmark and the version with iGPU outperforms the F version by a few percents, with Adobe software. And all what people said here confirms it.
The improvement is sometimes only 1 or 2%, but it represents more than 3 to 6 days per year
I take the version with iGPU