Discussion Nvidia Blackwell in Q1-2025

[DavidC1]

Yes I do think companies should be fined large amounts for deceptive marketing.

Here's a radical new idea: Rather than being fined, which results in accountants basically firing average people, individuals responsible should serve prison time or fined proportionately to their revenue.

My head was starting to spin when most of Volkswagen's problems were fined at a company level. What do you think they will do a quarter or two? Layoff employees!

So for an average employee getting paid $100K, fine them $100K, which will hurt. But a CEO earning $10 million being fined $10 million isn't enough, because he has other investments and backups. So in this case he should be fined $30 million. Make it hurt, otherwise how will they learn?

 

[SiliconFly]

Oh my! Just see the 5090 specs... it's just wow! No wonder AMD postponed RDNA4 launch. They just didn't have a choice.

CES 2025 | NVIDIA Officially Reveals the Blackwell GeForce RTX 50 Series: RTX 5090, RTX 5080, RTX 5070 Ti and RTX 5070 | Hardware Times

Las Vegas – January 6, 2025 – After many months of speculations and rumours leading up to its release, the next generation of GeForce RTX consumer desktop and laptop GPUs have been officially announced. These include the RTX 5090, RTX 5080, RTX 5070 Ti and the RTX 5070, powered by the incredible...

hardwaretimes.com

 

[poke01]

Pipe down champ, Max has twice the membw.

True, but hey AMD compared the iGPU and CPU of 288V to 16Core/40CU Strix halo. i consider this be bad too, Apple pulled this crap as well during the M4 keynote.

Marketing just doesn’t care about fair comparisons.

 

[SiliconFly]

True, but hey AMD compared the iGPU and CPU of 288V to 16Core/40CU Strix halo. i consider this be bad too, Apple pulled this crap as well during the M4 keynote.

Marketing just doesn’t care about fair comparisons.

Another AMD fake stunt!

 

[coercitiv]

Just look at the M4 Pro vs Strix halo marketing from AMD

Another AMD fake stunt!

Let's all bow down to the god of Whataboutism!

In case any of you need to reach me, you can find me in the RDNA4 thread talking about Nvidia lies and in the Qualcomm CPU thread tallking about Intel misleading benchmarks.

 

[poke01]

Let's all bow down to the god of Whataboutism!

In case any of you need to reach me, you can find me in the RDNA4 thread talking about Nvidia lies and in the Qualcomm CPU thread tallking about Intel misleading benchmarks.

oh so when we talk about AMDs deception, that’s where the buck stops for you?

You didn’t care to quote the person who started the comparison in the first place.

EDIT: I’m trying to make a point when the whole industry does it it’s hard for the DoJ to actually pursue an antitrust case against deceptive marketing. It’s also hard when the upcoming administration is itself full of deceptive marketing and lies to the people

 

[coercitiv]

oh so when we talk about AMDs deception, that’s where the buck stops for you?

And when whataboutism gets called out, go for the personal attack. It's even more telling considering what you were saying a page ago:

bit dramatic, these are companies their job is sell products/services.

I have a history on this forum calling out AMD for all their stupid and deceptive claims. Simply imagining things about me because it's more convenient won't make them true.

 

[poke01]

you can find me in the RDNA4 thread talking about Nvidia lies

I found you in the RDNA4 thread talking about Larrabee 😛

 

[poke01]

I have a history on this forum calling out AMD for all their stupid and deceptive claims. Simply imagining things about me because it's more convenient won't make them true

Part of that is convincing to sell is where a company does deception.

And when whataboutism gets called out, go for the personal attack.

it was personal when you singled me out but not the person who started talking about Intel Arrow lake in a Blackwell thread.

 

[poke01]

Anyway I’ll stop here and I’m sorry for derailing the thread

 

[SiliconFly]

oh so when we talk about AMDs deception, that’s where the buck stops for you?

You didn’t care to quote the person who started the comparison in the first place.

EDIT: I’m trying to make a point when the whole industry does it it’s hard for the DoJ to actually pursue an antitrust case against deceptive marketing. It’s also hard when the upcoming administration is itself full of deceptive marketing and lies to the people

That’s so true in many forums. It’s okay to trash talk about Intel & Nvidia. But it’s never okay to say the same about AMD. People might get shipped off to CIA black site prison for doing that.

 

[SiliconFly]

Just saw this… not sure about the settings, but still the latency is nowhere near what was predicted before.



If they can replicate the same across multiple titles, DLSS4 should actually be exciting.

 

[Keller_TT]

If they can replicate the same across multiple titles, DLSS4 should actually be exciting.

DLSS4 and equally Reflex 2 can make Blackwell quite potent because the nearly double AI TOPS help not just for MFG, but for typical 2x FG, RT Denoising, and DLAA. Blackwell also has much higher bandwidth with GDDR7. It has come at a big expense to wattage but 5070 anyways is positioned to co-exist with ADA Super tier.
If they have done a decent job with SMs, the 5070 should slot in place of 4070Ti, plus these other toys.

 

[CastleBravo]

Just saw this… not sure about the settings, but still the latency is nowhere near what was predicted before.

View attachment 114910

If they can replicate the same across multiple titles, DLSS4 should actually be exciting.

I would rather play CP2077 at 140fps 31ms than at 249fps 36ms. Better yet, I would rather turn down a couple settings and play it with ~15ms latency. I have zero interest in MFG. The transformer model DLSS IQ improvement looks promising though.

 

[PJVol]

nearly double AI TOPS help not just for MFG, but for typical 2x FG, RT Denoising, and DLAA.

Not only that, but apparently AI in general helps people to stop thinking critically or judge things objectively.
Not quite on topic, but just saying...

 

[steen2]

...nearly double AI TOPS...

Isn't that just FP4 vs FP8?

 

[MrTeal]

SOLID series looks nice

I'd sure hope even MSRP cards don't look like cheap plastic POSes when MSRP is $2k.

 

[Keller_TT]

Not only that, but apparently AI in general helps people to stop thinking critically or judge things objectively.
Not quite on topic, but just saying...

Not sure if this sarcasm is for me, but anyways just to be clear, I'm not saying RT denoising or DLAA is going to depend on deep learning in a game setting, but more DLSS4 models can be deployed to be used in parallel for improving image quality substantially. Not long before we see such games. DLSS4 on Blackwell is designed to use those TOPS.

 

[jpiniero]

Isn't that just FP4 vs FP8?

Yes

 

[Keller_TT]

Isn't that just FP4 vs FP8?

It's a bit more than 2x like for like vs ADA. INT8 and FP16. That's what I remember from the launch. It also adds FP4 precision boost now but the stated 2x TOPS was for INT8 and FP16.

 

[Heartbreaker]

Just saw this… not sure about the settings, but still the latency is nowhere near what was predicted before.

View attachment 114910

If they can replicate the same across multiple titles, DLSS4 should actually be exciting.

I'd bet no third party will replicate these misleading BS marketing numbers, even in the same title, let alone multiple ones.

The first trick NVidia pulls in all these latency tests, is to disable reflex in the cases where they FG/MFG are off, to inflate the base latency numbers, to hide the added latency.

Even accounting for that, the numbers looks sketchy.

Expect reliable testers to use Reflex across the board, for a fair comparison, and the extra latency hit for FG and MFG will be clear.

Image quality remains to be seen. I was NOT impressed with the DF footage. Ghosting improved in areas they were highlighting, but I saw more annoying artifacts in other areas.

 

[Elfear]

Oh my! Just see the 5090 specs... it's just wow! No wonder AMD postponed RDNA4 launch. They just didn't have a choice.

CES 2025 | NVIDIA Officially Reveals the Blackwell GeForce RTX 50 Series: RTX 5090, RTX 5080, RTX 5070 Ti and RTX 5070 | Hardware Times

Las Vegas – January 6, 2025 – After many months of speculations and rumours leading up to its release, the next generation of GeForce RTX consumer desktop and laptop GPUs have been officially announced. These include the RTX 5090, RTX 5080, RTX 5070 Ti and the RTX 5070, powered by the incredible...

hardwaretimes.com

Is there something new in that article that hasn't been discussed over the last week?

Also, did you think AMD was targeting 5090 performance with their RDNA4 launch? Rumors for the last several months have indicated a much lower target.

 

[Meteor Late]

Just saw this… not sure about the settings, but still the latency is nowhere near what was predicted before.

View attachment 114910

If they can replicate the same across multiple titles, DLSS4 should actually be exciting.

You need to be able to navigate through marketing and deception BS.
-When you are GPU limited and maxed out (so 98-99%), latency increases a lot compared to when you are not GPU maxed out.
-DLSS 3 and above, when FG is enabled, automatically activate Nvidia Reflex, which is a technology that fixes this huge latency increase when GPU maxed out by very slightly reducing framerate and technically not being maxed out anymore.

So what's happening in this instance is that DLSS off and DLSS 2 are showing latency numbers without Reflex enabled, whereas DLSS 3.5 and DLSS 4 are numbers with Reflex on. But you can also enable Reflex on DLSS off and DLSS 2 instances, and then the latency would be lower than on DLSS 3.5 and DLSS 4.

The reality is, the way this technology works, makes it inevitable to have around an extra frame of latency, because it holds the future frame along with the previous frame to be able to generate the extra frames between the two.

 

[SiliconFly]

You need to be able to navigate through marketing and deception BS.
-When you are GPU limited and maxed out (so 98-99%), latency increases a lot compared to when you are not GPU maxed out.
-DLSS 3 and above, when FG is enabled, automatically activate Nvidia Reflex, which is a technology that fixes this huge latency increase when GPU maxed out by very slightly reducing framerate and technically not being maxed out anymore.

So what's happening in this instance is that DLSS off and DLSS 2 are showing latency numbers without Reflex enabled, whereas DLSS 3.5 and DLSS 4 are numbers with Reflex on. But you can also enable Reflex on DLSS off and DLSS 2 instances, and then the latency would be lower than on DLSS 3.5 and DLSS 4.

The reality is, the way this technology works, makes it inevitable to have around an extra frame of latency, because it holds the future frame along with the previous frame to be able to generate the extra frames between the two.

Naturally. But considering the generational uplift, DLSS4 on 50 series shouldn't have more latency than DLSS3 on 40 series. In fact, it should be a bit less, considering the 50 series has better raster.

 

[CP5670]

I'm still having a hard time getting excited about the 5090. The UE5 games just don't look that good to justify their poor performance on current cards. 35% faster than the 4090 is nice but not amazing. However I still see the 4090 selling for $1800+ on ebay and FB marketplace, so I might be able to get close to what I paid for it.

Also an AIO model is a must with the 575W power. My current AIO 4090 is great and I don't want to "downgrade" to an air card that dumps so much heat into the case. The AIO models will be easily $2700-3000 street and be even harder to buy. It's odd how CPU AIOs are so common but are not really needed, while GPU AIOs are much rarer.