What is Radeon RX 7000 series for you? Please elaborate

[DemonicRyzen666]

Or it has transistors which don't add to the overall performance - transistors for computing loads (not gaming, or at least not games which have already been released), and ray-tracing, ML, DL, etc...

Add Magic software like, D.L.S.S, & Frame generation, NVEC, NV-Voice, just to use those shaders when gaming, because most shaders some around 40-30% are idle when gaming. Thats basically what Nvidia is doing to increase the shader use.

 

[Space Lynx]

Amd driver improvements have given the 7900 xt 14% gain in last of us. And 19% gain in f1, and I would bet money if more games were tested we would see these gains elsewhere too.

AMD Radeon RX 7900 XT Re-Review

Let's pretend for a minute it's 2022 again, and the Radeon RX 7900 XT is just getting released. But in this scenario, the GPU is introduced with...

www.techspot.com

At one point techspot says 10% gain is not as good as the gain observed at 1080p.... Are they fucking nuts? Drivers that give these kind of performance gains is like getting a whole new jump in card performance for free, not sure why heir connotation is negative about that.

@W1zzard this is really my only criticism of this website, I think once every 15 months you should do a retest of a few games (not all games) and just like 3 gpu's, just to show us how much driver improvement has changed the benchmark pages. Some of the numbers observed here at techspot as linked above are absolutely mind blowing. Amd should be commended on such fine tuning and gains across the board.

 

[W1zzard]

this is really my only criticism of this website, I think once every 15 months you should do a retest of a few games (not all games) and just like 3 gpu's, just to show us how much driver improvement has changed the benchmark pages.

You're not serious are you? I retest ALL CARDS every few months, with new drivers, new games, new patches, new hardware, everything, last retest was in May, after RX 7600 launch. The one before that was in January

 

[Space Lynx]

You're not serious are you? I retest ALL CARDS every few months, with new drivers, new games, new patches, new hardware, everything, last retest was in May, after RX 7600 launch. The one before that was in January

I did not know you did this, when a new card comes out, I thought you just tested the new card. That is an insane amount of work...

 

[W1zzard]

I did not know you did this, when a new card comes out, I thought you just tested the new card. That is an insane amount of work...

I've been doing that for over a decade .. every few months .. two weeks of full-time work .. and I keep getting this feedback from people .. even regulars like you .. I assume you're reading the "Test Setup" pages of my reviews, how can I make it more clear that everything is retested?

 

[Space Lynx]

@W1zzard I'm a bit confused. I don't see any data from higher end cards in your May review. You only post midrange cards. So it's been 8 months since 7900 xt since the reader has had access to fps numbers.

AMD Radeon RX 7600 Review - For 1080p Gamers

Today AMD is launching their Radeon RX 7600, which is targeted at gamers with 1080p Full HD screens. Our review confirms that thanks to the new RDNA3 graphics architecture, the new card is highly energy efficient.

www.techpowerup.com

 

[W1zzard]

I only include cards that are around +-25% to the tested card, the summaries have everything though.

There's this recent review: https://www.techpowerup.com/review/zotac-geforce-rtx-4070-amp-airo/9.html

 

[Space Lynx]

I only include cards that are around +-25% to the tested card, the summaries have everything though.

There's this recent review: https://www.techpowerup.com/review/zotac-geforce-rtx-4070-amp-airo/9.html

I understand now, thanks, this does help.

 

[W1zzard]

I understand now, thanks, this does help.

http://imgur.com/xAnIXvT

http://imgur.com/p8DZVZi

Adding this to the Test Setup page of all reviews will help?

 

[Space Lynx]

http://imgur.com/xAnIXvT

http://imgur.com/p8DZVZi

Adding this to the Test Setup page of all reviews will help?

it will probably help some people certainly.

I am still confused though, if you re-tested all gpu's in May, why not make that data available to everyone in May? your work is just going to waste if no one can see it, doesn't have to be attached to the midrange gpu review since it is over 25% threshold, but you could not let that extra work go to waste, and post a second article just called "gpu re-cap with updated drivers"

i don't know, just my thoughts

edit: to be more clear, I understand the updated ones will get posted eventually with the other incoming gpu's, I just don't see why not post them right then when all done nice and fresh. TechSpot does this, as I linked a few posts above, and I particularly enjoyed reading it because it was so fresh/recent.

 

[W1zzard]

I just don't see why not post them right then when all done nice and fresh.

Great point. Quite often the rebench is done leading up to a launch, which means I'm kinda busy with testing the new cards that'll come out soon. I'll think about how this can be improved.

 

[nguyen]

So are we gonna follow cherry picking benchmarking that HUB/Techspot like to do now?

I mean the Techspot is doing a fine job of cherry picking games in their re-review of 7900XT
Initial review: 16 games
Re-review: 15 games, games that were not in the original review (but show large gain from new driver) just conveniently show up at the top of the article too .

The new driver gains would amount to less than 1% if they test with 50 games

 

[ARF]

They should test the 12GB RTX 4070 Ti at 4K in games which require more than 12 GB and then compare with the 20GB RX 7900 XT.
Simply to emphasize on the drawbacks of that nvidia decision to equip the card with so little VRAM.

Or better come back in another year or two to revisit the results with some fresh, newly released titles.

 

[Space Lynx]

So are we gonna follow cherry picking benchmarking that HUB/Techspot like to do now?

I mean the Techspot is doing a fine job of cherry picking games in their re-review of 7900XT
Initial review: 16 games
Re-review: 15 games, games that were not in the original review (but show large gain from new driver) just conveniently show up at the top of the article too .

The new driver gains would amount to less than 1% if they test with 50 games
View attachment 308112

The only point of the techspot article was to find which games improved the most with drivers... If I had OLED and a 4090 like you I'd spend more time gaming and less time talking about a few 7900 xt owners.

 

[nguyen]

The only point of the techspot article was to find which games improved the most with drivers... If I had OLED and a 4090 like you I'd spend more time gaming and less time talking about a few 7900 xt owners.

I have been playing lots of games recently, but never do I play (nor replay) very specific few games that gain some perf via driver updates .

 

[Space Lynx]

I have been playing lots of games recently, but never do I play (nor replay) very specific few games that gain some perf via driver updates .

normally I get along with you mate, but your fanboyism and clear love for Nvidia is getting exhausting. Nvidia makes great product, AMD makes a slightly lesser product at a cheaper price, its been this way for time immemorial, lets move on shall we

Im happy with my 7900 xt for the price i paid.

 

[nguyen]

normally I get along with you mate, but your fanboyism and clear love for Nvidia is getting exhausting. Nvidia makes great product, AMD makes a slightly lesser product at a cheaper price, its been this way for time immemorial, lets move on shall we

I
m with my 7900 xt for the price i paid.

Likewise, I'm just here because you are asking Wizzard to do what is essentially an act of cherry picking that HUB/Techspot is doing.

Or better, just ask Wizzard nicely to do 50 games benchmark between 4070Ti vs 7900XT and see if which one actually had better driver improvement

 

[Space Lynx]

Likewise, I'm just here because you are asking Wizzard to do what is essentially an act of cherry picking that HUB/Techspot is doing.

Or better, just ask Wizzard nicely to do 50 games benchmark between 4070Ti vs 7900XT and see if which one actually had better driver improvement

I never said cherry pick anything, I even said in my original message pick any few games (but not all), it was only from his post that I learned he retests all, which is excessive imo, I just want a general idea, but its nice that he does that.

 

[nguyen]

I never said cherry pick anything, I even said in my original message pick any few games (but not all), it was only from his post that I learned he retests all, which is excessive imo, I just want a general idea, but its nice that he does that.

Picking a few games that AMD specifically wrote in their driver blog that state to improve performance is definitely cherry picking, I'm not quite sure what you think cherry picking is.

Anyways, hope you enjoy such games, I already voted so I'm out.

 

[Space Lynx]

Picking a few games that AMD specifically wrote in their driver blog that state to improve performance is definitely cherry picking, I'm not quite sure what you think cherry picking is.

Anyways, hope you enjoy such games, I already voted so I'm out.

if it makes you feel better, techspot was negative in attitude about the gains made at 1440p. I honestly don't understand you or them, I don't care if it is 10% or 19% or 14% gains are gains, and it won't just be this game, those adjustments will benefit a few other games here and there that never get tested, but thats beside the point, its great to see this kind of service compared to years past. Ratchet and Clank was fixed within 72 business hours, which was also impressive to me.

I think Nvidia and AMD are doing great on drivers these days, and its nice to see that is all

 

[Assimilator]

So are we gonna follow cherry picking benchmarking that HUB/Techspot like to do now?

I mean the Techspot is doing a fine job of cherry picking games in their re-review of 7900XT
Initial review: 16 games
Re-review: 15 games, games that were not in the original review (but show large gain from new driver) just conveniently show up at the top of the article too .

The new driver gains would amount to less than 1% if they test with 50 games
View attachment 308112

Just let the AMD fans believe in their mythical fine wine, alright? They tend to cry less when they do that, so we all benefit.

 

[Space Lynx]

Just let the AMD fans believe in their mythical fine wine, alright? They tend to cry less when they do that, so we all benefit.

I paid $580 for 7900 xt... go pay 1200 for for your 4080 and buy jensen another leather jacket.

 

[Kyan]

Likewise, I'm just here because you are asking Wizzard to do what is essentially an act of cherry picking that HUB/Techspot is doing.

Or better, just ask Wizzard nicely to do 50 games benchmark between 4070Ti vs 7900XT and see if which one actually had better driver improvement

He never specified any particular game, he just thought that re-tests hadn't been done and that it would be good to do them again from time to time. You know, to see how the old cards perform with new games.

 

[wheresmycar]

it will probably help some people certainly.

I am still confused though, if you re-tested all gpu's in May, why not make that data available to everyone in May? your work is just going to waste if no one can see it, doesn't have to be attached to the midrange gpu review since it is over 25% threshold, but you could not let that extra work go to waste, and post a second article just called "gpu re-cap with updated drivers"

i don't know, just my thoughts

edit: to be more clear, I understand the updated ones will get posted eventually with the other incoming gpu's, I just don't see why not post them right then when all done nice and fresh. TechSpot does this, as I linked a few posts above, and I particularly enjoyed reading it because it was so fresh/recent.

ooooh i like the sound of that!!

I totally get what you're suggesting and its a valid request. Dedicated post driver update articles are always welcome providing the reviewer is willing or can manage the extra workload. I often view some of HUBs content for these little extras but can't always be bothered with their exhaustive review layout via techspot and not so big on watching videos either (well fast fwd and timestamps are excellent). Anyway i'd fancy seeing TPU flexing their muscles here too... or exclusively when driver update performance variance is significantly worthy of a shout-out.

 

[ARF]

What is the chance that the GCD of Navi 31 is being made on the TSMC 3nm process, and not as claimed officially the 5nm process?

I found something in the data which doesn't add up.

Official data is:

Process Type N5 / N6 FinFET
Transistors 57,700 million
GCD Transistors 45,400 million
MCD Transistors 2,050 million x6
Density 109.1M / mm²
GCD Density 150.2 M / mm²
MCD Density 54.64 M / mm²
Die Size 529 mm²
GCD Die Size 304.35 mm²
MCD Die Size 37.52 mm² x6

AMD Navi 31 GPU Specs

6144 Cores, 384 TMUs, 192 ROPs

www.techpowerup.com

The thing is that there is no such high transistor density which can be achieved using the TSMC 5nm process - the maximum claimed is 138.2 MTr/mm² https://en.wikipedia.org/wiki/5_nm_process
and 143.7 MTr/mm² on TSMC 4nm https://en.wikipedia.org/wiki/5_nm_process

However, the TSMC 3nm process allows it - maximum is 224.2 MTr/mm² on N3P and N3X. https://en.wikipedia.org/wiki/3_nm_process

I think GCD is made on 3nm process, while MCD is made on 7nm but extremely far from its potential maximum 114 MTr/mm².


To summarise.
GCD 150.2 MTr/mm²
TSMC 5nm process - maximum is 138.2 MTr/mm² https://en.wikipedia.org/wiki/5_nm_process
TSMC 3nm process - maximum is 224.2 MTr/mm² (N3P | N3X) https://en.wikipedia.org/wiki/3_nm_process
TSMC 7nm process - maximum is 114 MTr/mm² https://en.wikipedia.org/wiki/7_nm_process
MCD 54.64 MTr/mm²

For a comparison, nvidia's AD102 transistor density is 125.3 M/mm², and offically claimed 4nm process - maximum is 143.7 M/mm².