Video Card Rumours

[EightyDuce]

This guy has his ear to the ground on a lot of PC/tech topics and put this video out a little while back. Basically, getting an Ampere at a reasonable price isn't going to be a thing anytime soon.

 

 

[LucShep]

https://videocardz.com/newz/amd-intel-and-nvidia-all-to-release-new-desktop-graphics-cards-in-the-next-few-months

New desktop graphics are coming

For the first time in nearly 24 years, we will be seeing GPU releases from three companies in the next few months. A summary of those releases has been put together by 3DCenter.

Intel ARC by end of Q2 2022

Intel is finally expected to launch its Arc Alchemist graphics cards for laptops by the end of this month and then desktop GPUs by the end of June (second quarter). While the company did not confirm the names of its upcoming desktop products, it is commonly believed that the flagship model will adopt a name of Arc A780, featuring the full DG2-512EU GPU and 16 GB of GDDR6 memory. It is expected to compete in the upper mid-range segment with Radeon RX 6700 and GeForce RTX 3070 series.
Probably a bit sooner, Intel might release its low-end DG2-128EU based Arc A300 GPUs (A350/A380), those cards have a chance to compete with Radeon RX 6500 and GeForce RTX 3050 series in the entry-level space. All those cards are officially coming by the end of the second quarter (see article).

NVIDIA RTX 3090 Ti in March 

NVIDIA will be releasing its flagship RTX 3090 Ti on March 29th, claims the rumor from Chiphell. This model has been on a schedule for a few weeks now, after NVIDIA failed to deliver an update on this SKU since the CES 2022 announcement. What is important to note is that the March release date has not been confirmed yet. However, at this moment it is the only NVIDIA RTX 30 card now on a launch schedule, after alleged cancellation of RTX 3070 Ti 16 GB SKU (see article).


AMD Radeon RX 6_50XT refresh coming in April 

AMD will be refreshing its RDNA2 desktop models led by Radeon RX 6950XT, featuring faster 18 Gbps memory and increased TDP. While this model is considered a direct response to NVIDIA RTX 3090 Ti, the RX 6750XT and RX 6650XT should offer competitive performance against Intel Arc GPUs. The RX 6950XT should launch in April, but the launch date of the mid-range XT SKUs is currently unknown (see article).


AMD, NVIDIA and Intel will definitely give us a lot of new hardware to discuss in the next couple of weeks, and that’s right before we switch to next-gen GPU architectures in the third and fourth quarter of this year.
 

  Source: 3DCenter

Edited March 22, 2022 by LucShep

[KoN]

Intel might of made a few mistakes . 

No HDMI - I .4 . 

Intel only card XSS 4k rendering . ??

[LucShep]

Source:  https://videocardz.com/newz/intel-teases-2250-mhz-boost-clock-and-175w-gpu-power-for-its-flagship-desktop-arc-gpu

Did Intel reveal the specs of its flagship Arc A780 desktop GPU?

Interesting detail on Intel discrete desktop Arc GPU has been discovered by 3DCenter.

What 3DCenter noticed is that in Intel’s official Arc Control presentation video, specs of a discrete desktop GPU might have been shown for a split second. The Arc Control is Intel’s new GPU monitoring software for Arc Alchemist and Xe Iris discrete GPUs. The presentation shows various features of the new software, including its new Performance Tab for GPU monitoring and some basic overclocking.

Data such as GPU power or core and memory frequency will be shown either through the Arc Control software or a built-in screen overlay that can be enabled in games. Intel is showing this tab repeatedly in the video, but only for a second, we can see the specs that do not match any mobile Arc GPU that was announced 2 days ago.

According to those two screenshots, the GPU clock would reach 2200-2250 MHz with over 99% GPU utilization. It is the fore safe to assume that this is a boost clock for Intel’s new Xe-HPG GPU architecture.

Potential specs of Intel Arc desktop GPU (Source: Intel):

 

A 2.2 GHz GPU clock shouldn’t really be exclusive to desktop GPUs, though. The Arc A350M Geekbench data shows that the GPU can reach similar frequency during OpenCL test. Officially the core clock for this GPU is 1150 MHz, therefore that would be the base clock, while 2.2 GHz is boost.

Intel Arc A350M (mobile) reaching 2.2 GHz clock (Source: Geekbench):

 

So how do we know the Arc Control video is showing desktop GPU? First give away is the GPU Power of 175W. None of the mobile parts presented by Intel have such a high power draw. Only Arc A770M has a maximum GPU power of 150W, but theoretically it could reach higher thanks to Intel Deep Link technology (which unlocks unused CPU power for the GPU like NVIDIA Dynamic Boost).

Since this is GPU power only, other desktop card components will still add more power, so the whole card might easily consume more than 200W, 3DCenter predicts. We do not know if Intel’s GPU Power accounts for memory subsystem yet. Nevertheless, such a GPU power was leaked a while ago for the SOC1 GPU (ACM-G10) for desktop parts. Do note that there is a difference between TGP, TDP or GPU Power metrics, but the details on Intel’s implementation are yet to be confirmed once desktop Arc GPUs are released.

Intel DG2 GPU Power Consumption (Source: @tegra):

Furthermore, the VRAM clock of 1093 MHz suggest we are looking at an effective peed speed of 17.5 Gbps, while the fastest mobile GPU has 16 Gbps. Memory overclocking for mobile GPUs is an extremely unlikely feature to be enabled, thus this might confirm that desktop Arc will not be slower than, for instance, a GeForce RTX 3070 (14 Gbps).
 

GPU	AMD Navi 22	NVIDIA GA104-300	ACM-G10
Fabrication Node	TSMC N7	SAMSUNG 8N	TSMC N6
Die Size	336 mm²	392 mm²	406 mm²
Transistor Count	17.2 B	17.4 B	21.7 B
FP32 Units	2560	5888	4096
Memory	12GB GDDR6 129-bit	8GB GDDR6 256-bit	16GB GDDR6 256-bit
Memory Speed	16 Gbps	14 Gbps	17.5 Gbps
Card Power	230W (TGP)	220 (TDP)	175W (GPU Power)

(Source: Intel via 3DCenter.org)

Edited April 1, 2022 by LucShep

[Maverick1202]

On 1/9/2021 at 2:58 AM, Lange_666 said:

Totally agree. RTX 3090 from EVGA goes for 2299€ here (official shop price). Just plain crazy and even then it's only available on pre-order with no time window.

AMD XT 6900 X which has an official recommended price of $999 goes on sale for 1830$, also on pre-order. 

For me the crypto currency crap can't be trashed quickly enough because it's one of the major causes.

 

It's all about making the $$$.  there is a backlog of people that are waiting for cheaper prices.  Screw the miners, sell to more gamers!  

[LucShep]

NVIDIA GeForce RTX 4090/4080/4070 get new rumored specs

NVIDIA hardware leaker @kopite7kimi suggests specifications change for GeForce RTX 40 series.
 

RTX 4090 with more cores

According to Kopite, the RTX 4090 would get more cores than previously reported. Instead of 16128 CUDAs, the new specs list 16384 cores, so an upgrade from 126 SM (Streaming Multiprocessors) to 128. The full configuration of AD102 GPU supposedly has 144 cores, but such configuration is more than likely to be reserved for future TITAN/RTX4090Ti SKU.

Other parts of the specs are unchanged. This card is the flagship RTX 40 model with TDP expected to reach 450W. It should launch with 24GB GDDR6X memory clocked at 21 Gbps.
 

RTX 4080 with 10240 CUDA cores

Furthermore, we now have first rumored GPU configurations of RTX 4080 and RTX 4070 GPUs. The former is the only known AD103 based graphics card, which would supposedly debut a month after RTX 4090. According to Kopite, the RTX 4080 would feature 10240 CUDA cores and 16GB memory.

Interestingly, this card might not rely on GDDR6X like RTX 4090 supposedly does, but GDDR6 instead. This part of the specs is most likely to change though, because 18 Gbps GDDR6 memory would actually be slower than RTX 30 predecessors (RTX 3080 has 19 Gbps memory). However, one should underline the fact that AD102 GPU has larger L2 cache, which may work just as efficiently as Radeon RX 6000 series. In other words, the memory speed does not need to be that much faster.
 

RTX 4070 with 10GB and 160-bit memory

Finally, RTX 4070 sees a rumored spec update. The card is no longer expected with 12GB memory but 10 GB instead. This would also mean a narrower memory bus from 192-bit to 160-bit. It is said that RTX 4070 would feature 7168 CUDA cores, which would mean 512 disabled CUDA cores on AD104 GPU.

The RTX 40 series are now rumored to launch in late Q3/early Q4. The initial lineup would feature all three SKUs mentioned above, however their rollout would span across the fourth quarter.
 

RUMORED NVIDIA GeForce RTX 40 Series Specs
VideoCardz.com	GeForce RTX 4090	GeForce RTX 4080	GeForce RTX 4070	GeForce RTX 4060
Architecture	Ada (TSMC N4)	Ada (TSMC N4)	Ada (TSMC N4)	Ada (TSMC N4)
GPU	AD102-300	AD103-300	AD104-400 -> AD104-275	AD106
Board Number	PG139-SKU330	PG139-SKU360	PG141-SKU341	TBC
SMs	126 -> 128	80	56	TBC
CUDA Cores	16128 -> 16384	10240	7168	TBC
CUDA vs Predecessor	+56% (vs 3090)	+18% (vs 3080)	+21% (vs 3070)	TBC
Memory	24 GB G6X	16 GB G6	12 GB G6 -> 10GB G6	TBC
Memory Bus	384-bit	256-bit	192-bit -> 160-bit	TBC
Memory Speed	21 Gbps	18 Gbps	18 Gbps	TBC
TDP	~450W	~420W	~300W	220W+
Launch Date	September-October 2022	October-November 2022	November-December 2022	January 2023

Source: VideoCardz and @kopite7kimi

 

Edited July 3, 2022 by LucShep

[reece146]

I don't get why NV doesn't do 16GB cards. Not enough addressing space?

Always a missed opportunity IMO.

[LucShep]

Here it comes.........

https://videocardz.com/newz/gigabyte-geforce-rtx-4090-cards-now-listed-in-australia

 

 Placeholder pricing for RTX 4090 is here

Australia becomes the first country to list upcoming NVIDIA RTX 40 graphics cards.

Gigabyte models are now confirmed by first retailer in Australia. The SKU models have become known thanks to the other retailer from Vietnam. So, not only are retailers already booking RTX 40 series to fill their ‘next-gen’ inventory, they are actively listing the new series on their websites already.

The RTX 4090 GAMING OC is listed at 4270 AUD (2916 USD), which is of course just a placeholder. Probably in line of what we usually see before each high-end GPU launch.



Meanwhile, another Australian store appears to have RTX 4090 AORUS Master (again from Gigabyte). It is not stated if 3,900 AUD price includes tax though, but this is equal to 2,615 USD, so in territory of peak RTX 30 series pricing during ‘complicated times’.

Gigabyte GeForce RTX 4090 AORUS Master, Source: PCAOS

The RTX 4090 does not have a rumored price yet, nor has NVIDIA confirmed one to board partners. Early retailer pricing is just a good indicator that the launch is approaching and in this case, a confirmation on specific SKUs.

Source: PCAOS, Computers Perth

 

Edited September 11, 2022 by LucShep

[Lange_666]

At the current exchange rate that would be 2861 and 2613 euro's. No thank you, nVidia doesn't have Apple status.

[LucShep]

New RTX 4090 graphics card: crazy performance and a price that is just as much

(September 9, 2022 by archyde)

 

While the new range of graphics cards made in Nvidia will be presented on September 20 at a conference, the RTX 4090 is leaking some information! And not the least since these are its potential performances, as well as its price!

The Nvidia RTX 4090, twice as powerful as the RTX 3090?!

All PC gamers are waiting for only one thing: to know if the RTX 4090 will be worthy of their expectations! And for good reason, the overpowered 3090 and 3090 Ti surprised us with their performance, but Nvidia does not want to stop there.

All this is just rumors, since nothing has been formalized, but a benchmark has been published on a Chinese site. The RTX 4090 is said to score 19,000 points on 3DMark TimeSpy Extreme software for testing the computing speed of a chip.

On another benchmark, the score would have reached no less than 20,192! So if we believe in these sources, the RTX 4090 could show us a power almost 2 times higher than the RTX 3090 (+90%).

On the frequency side, the new graphics card from Nvidia would show a frequency of 3015 MHz (3Ghz), which is extremely high for a GPU. More surprisingly, even with this frequency the RTX 4090 would not exceed 55°C, even pushed to the maximum of its capacities.

According to the person who performed the benchmark, the RTX 4090 would have a “very large” heatsink that could not fit in all the towers. Regarding consumption, the TDP would not exceed 450 Watts.

Source : Videocardz (this article)

The RTX 4090 has a knockout price

Now let’s talk about what annoys a little more: the price. If PC components have taken a huge price hit over the past two years, the course of the euro is mixed in with all this and offers us a more staggering price.

It was via an Australian reseller that a screenshot leaked on the internet. On an image, we can see a price of 4270 Australian dollars, or approximately 2909€ with the current rate of the euro.

As a reminder, the RTX 3090 had pointed the tip of its nose in our stores for the modest sum of €1099. You know, the days when it was thought that recommended prices would remain the norm. Even if the prices of an RTX 3090 revolve around 1200/1300€, the RTX 4090 would be worth more than double that of its older generation counterpart.

The Australian store displays an RTX 4090 from Gigabyte, with 24 GB of dedicated video memory, but above all overclocked. Another Australian shop, displays another model, still at Gigabyte, the Aorus Master, this time at 3900 Australian dollars, or around 2657€. However, no image has been released.
 

Nvidia plans to officially announce the arrival of the RTX 4000 line on September 20 at a conference. During this meeting, we will know more about the RTX 4090, which should, as traditionally, show up first, before welcoming its little sisters afterwards.

Thus, we will have confirmation of the real launch prices of the RTX 4090, hoping that the figures announced on these Australian stores are false…

Edited September 12, 2022 by LucShep

[Mr_sukebe]

If true, might seem obscene, but it's really just supply and demand.

Nvidia are probably confident of selling a bunch and the only way that the price is going to come down is by genuine competition existing.

[Lange_666]

No more EVGA GPU's...

 

 

[BitMaster]

More of this !

That's the language Jensen understands.

[skypickle]

since the gold mine of mining crypto has dried up, Nvidia wants more profits and can't stand their 'partners''stealing' their profits. Especially partners who have excellent customer support and are a real threat.

[dburne]

Very sad news, have really loved my EVGA GPU's dating back for several years.

Guess I will look at MSI for the 40 series. 

[AngleOff66]

So the next board meeting should be interesting over at N.

Bad couple of quarters in comparison to the glory days of crypto mining and losing an AIB.  Unless they want to get rid of AIB partners over time?

As Steve pointed, peoples jobs are involved here.  I wish them luck.

 

Edited September 16, 2022 by NoCoolName66

[AngleOff66]

Lenovo 4090 pictures.  It looks to be for prebuilts according to some of the comments? Hopefully the other AIBS can/will make them smaller?

So some cards look to take up 4 slots or you can have a length of 14"? so at up to 660w new psu and possibly new case?

https://videocardz.com/newz/lenovo-geforce-rtx-4090-gpu-pictured-some-more-it-is-36-cm-long

 

 

Edited September 18, 2022 by NoCoolName66

[SkateZilla]

AMD touting that rDNA3 May hit 4 GHz

[LucShep]

On 9/16/2022 at 9:35 PM, Lange_666 said:

No more EVGA GPU's...

 

 

Don't believe for a minute that EVGA are the good guys - they also ripped off everyone with prices nearly double (sometimes more) the MSRP of GPUs for nearly two years. 
If this argument dates way over two years back, then why only announcing it now? (the cashing machine noises were still too good, eh?) 

Think for a minute.... most AIBs were likely making 400%+ profit for nearly two years (actually 429% in 2021 according to ZOTAC, IIRC) specifically on the GPUs - a product that makes 70%+ of the whole sales of EVGA - and all of a sudden EVGA plays the "boo hoo" when the market has suddenly dried due to Mining taking a blow....

There's always two sides on a story, and in this case I'm not buying the "disrespect" thing for a minute (maybe it's "enjoying the humongous profits and leaving while on top"?).
There's more to it than what we're being shown.

Edited September 20, 2022 by LucShep

[Cab]

41 minutes ago, LucShep said:

Don't believe for a minute that EVGA are the good guys - they also ripped off everyone with prices nearly double (sometimes more) the MSRP of GPUs for nearly two years. 

Everyone hates free market pricing until it's their own product selling for top dollar.

[SkateZilla]

jsut to Note.

 

eVGA terminated the AIB Partnership, not nVidia.

 

nVidia did announce procing regulations started with 4000 Series a few months back.

 

eVGA likely did not like that.

[dburne]

All I know is EVGA made it possible for me to get a 3090 FTW3 Ultra at MSRP fairly early on through the Que system they had for consumers. And the several years experience I have with their GPU's and motherboards.

[unlikely_spider]

I was able to get two of the EVGA cards through their queuing system, when they were nearly impossible to find at comparable prices anywhere else. EVGA is generally at the top of my list for consideration... not anymore it seems. If I'm buying a 4000-series (who am I kidding? I will) then I'll likely look for a Founder's Edition. Those have served me well this current gen.

[LucShep]

Quote

31 minutes ago, dburne said:

All I know is EVGA made it possible for me to get a 3090 FTW3 Ultra at MSRP fairly early on through the Que system they had for consumers. And the several years experience I have with their GPU's and motherboards.

26 minutes ago, unlikely_spider said:

I was able to get two of the EVGA cards through their queuing system, when they were nearly impossible to find at comparable prices anywhere else. EVGA is generally at the top of my list for consideration... not anymore it seems. If I'm buying a 4000-series (who am I kidding? I will) then I'll likely look for a Founder's Edition. Those have served me well this current gen.

 

Their warranty was what impressed me most (best in the business).
The queuing system was a good idea, though it was often perceived as inconsistent (if not corrupt), as often the cards were out of stock for months on those in queues, while shipments seem to go onto miners hands.

And BTW.... remember Igor Wallosek being excluded from sampling products on Tom'sHardware? (arguably their best reviewer, when that place was actually worth the time)
Here's his take on this EVGA subject:  https://www.igorslab.de/en/evga-pulls-the-plug-with-loud-bang-yet-it-has-long-been-editorial/

Edited September 20, 2022 by LucShep

[AngleOff66]

"The RTX 4080 16GB is to feature an AD103 GPU with 9728 CUDA cores, a total of 16GB of GDDR6X memory clocked at 22.5 Gbps and TDP of 320W. This model is set to launch in November for at least 1199 USD."

"4090 $1599 usd" edit for typo(I was in a rush)

https://videocardz.com/newz/nvidia-introduces-geforce-rtx-4090-4080-series-rtx-4090-launches-october-12th-for-1599-usd

According to this post on reddit the 4080 12gb is the same performance as the 3090 ti?

 

Edited September 20, 2022 by NoCoolName66