Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 1010 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
C

CakeMonster

Golden Member
Nov 22, 2012
1,662
843
136
Fjodor2001 said:
9850X3D Reviewed:


”3% more performance, 30% more power”

Let’s hope 9950X3D2 performs better.
Click to expand...
Just watched the GN review as well. The new part was not interesting, but I notice that Z4 vs Z5 (both vanilla) is still highly confusing, Z5 outperforms on average but Z4 will have better minimum frame rates in some games, as well as faster in some synthetic tests. I thought we'd have more conclusive answers by now and that Z5 would age better. Maybe it will in another 2 or 4 years with more software updates...
 
G

gdansk

Diamond Member
Feb 8, 2011
4,774
8,066
136
CakeMonster said:
Maybe it will in another 2 or 4 years with more software updates...
Click to expand...
What software change, other than compiling specifically for znver5 which is never happening on Windows, did you suppose would benefit Zen 5 more than Zen 4?

Whenever Windows schedules on the wrong CCD, it'll hurt both. Zen 5 is going to stay where it has been except for a few initial bugs that were solved by the time the 9800X3D launched.
 
J

Joe NYC

Diamond Member
Jun 26, 2021
4,216
5,818
136
Fjodor2001 said:
”3% more performance, 30% more power”

Let’s hope 9950X3D2 performs better.
Click to expand...

It's fine, 9850x3d is #1 in performance #2 in power efficiency.
The only competition is 9800x3d which is #1 in power efficiency and #2 in performance:

9800x3d leads in power efficiency by so much that even 30% reduction in efficiency is good enough for #2 for 9850x3d.

1769656535263.png

1769656571598.png
 
  • Like
Reactions: Tlh97, inquiss, Elfear and 1 other person
J

Joe NYC

Diamond Member
Jun 26, 2021
4,216
5,818
136
Fjodor2001 said:
So you think 3% perf improvement for 30% more power and higher price is fine/good? Not sure if serious.
Click to expand...

Unless you were born yesterday, you would know that the last 1% of performance is most costly in terms of $ and Watts. Except in this case, the difference in $s is trivial. 4%. So not costly at all.

I don't think there has ever been such a great deal that you would only pay 4% for the last 3% of performance. Typically, it is far more costly.

But it is ok if you want to pass on the last percent of performance. In that case, 9800x3d is there to fill all your needs.
 
  • Like
Reactions: Tlh97, Elfear and lightmanek
M

MoistOintment

Member
Jul 31, 2024
162
260
96
fastandfurious6 said:
posted on pantherlake thread with more detail,

you can get Halo 392 full laptop 'ASUS TUF' for around 1500 bucks now

pantherlake machines START from 1500, up to 2500. DOA
Click to expand...
If I wanted to spend $1500 on an ugly plastic brick with poor battery life, why wouldn't I get something with a dGPU instead?
 
I

itsmydamnation

Diamond Member
Feb 6, 2011
3,135
4,003
136
Fjodor2001 said:
So you think 3% perf improvement for 30% more power and higher price is fine/good? Not sure if serious.
Click to expand...
yes , lets remember context , this is a new BIN/ SKU , it is not a new part. remember back in like 2002 when AMD and intel would release new bins like almost every 3 months...... this is just that.......

people only pretend to care about power on desktop when they are getting beaten in T1.
 
  • Like
Reactions: Tlh97, yuri69 and lightmanek
F

Fjodor2001

Diamond Member
Feb 6, 2010
4,608
744
126
Joe NYC said:
Unless you were born yesterday, you would know that the last 1% of performance is most costly in terms of $ and Watts.
Click to expand...
Exactly why I don't think it's worth it. Who will even notice 3% perf improvement, unless you're gunning for benchmark bragging rights.
 
L

LightningZ71

Platinum Member
Mar 10, 2017
2,692
3,391
136
A straight port of Strix Point to N3P would have netted next to nothing of importance. They would have gotten maybe another 200Mhz of clock speed over Gorgon Point with the same exact logic. The RDNA 3.5 iGPU might have gained 5% peak clock over Gorgon Point. The die would have shrunk by, what, 10% for a net increase in per chip cost. Fully loaded power draw would have maybe gone down a few percent. IT wouldn't be worth it without substantive changes that would require a new floorplan. Once you start to mess with the floorplan, costs go up dramatically. If they had REALLY wanted, they could have done an optimized shrink that made the minor logic changes to get the best performance from the new node as well as changing the floorplan to accommodate a 16MB MALL cache. Doing a MALL cache like that, gaining 400 Mhz in iGPU clock speed and supporting a couple bins higher in LPDDR5X RAM would have gone a LONG way to bringing the iGPU up to Panther Lake performance levels. Maybe not beat it, but the difference wouldn't be compelling in any way. The extra few hundred MHZ of ST boost speed would have kept it ahead of PTL in most every benchmark, and the MT performance would see a notable uplift. But, BUT, none of that would have justified the increase in costs and wouldn't have done much for market volume.

In other words, a waste of money. Save the money and sell what you already have while you work on something MUCH better.
 
  • Like
Reactions: Tlh97
J

Joe NYC

Diamond Member
Jun 26, 2021
4,216
5,818
136
Fjodor2001 said:
Exactly why I don't think it's worth it. Who will even notice 3% perf improvement, unless you're gunning for benchmark bragging rights.
Click to expand...

In Zen 3, AMD introduced lower binned cheaper V-Cache models, below the 5800x3d model.

It seems that AMD is aiming to add more bins down the road, with 9850x3d as a flagship and lower binned, lower core SKUs with V-Cache.

BTW, the king of efficiency is still 7800x3d (which I have in my home system). It will not surprise me if we see 65 Watts V-Cache SKUs.

In non-V-Cache models, AMD just took off the "x" suffix to get the 65 Watt parts. Which would not work so well with "x3d". So maybe just different SKU numbers.
 
poke01

poke01

Diamond Member
Mar 8, 2022
4,864
6,194
106
fastandfurious6 said:
you can get Halo 392 full laptop 'ASUS TUF' for around 1500 bucks now
Click to expand...
I don’t want an ASUS tuf. What now?
LightningZ71 said:
A straight port of Strix Point to N3P would have netted next to nothing of importance. They would have gotten maybe another 200Mhz of clock speed over Gorgon Point with the same exact logic. The RDNA 3.5 iGPU might have gained 5% peak clock over Gorgon Point. The die would have shrunk by, what, 10% for a net increase in per chip cost. Fully loaded power draw would have maybe gone down a few percent. IT wouldn't be worth it without substantive changes that would require a new floorplan. Once you start to mess with the floorplan, costs go up dramatically. If they had REALLY wanted, they could have done an optimized shrink that made the minor logic changes to get the best performance from the new node as well as changing the floorplan to accommodate a 16MB MALL cache. Doing a MALL cache like that, gaining 400 Mhz in iGPU clock speed and supporting a couple bins higher in LPDDR5X RAM would have gone a LONG way to bringing the iGPU up to Panther Lake performance levels. Maybe not beat it, but the difference wouldn't be compelling in any way. The extra few hundred MHZ of ST boost speed would have kept it ahead of PTL in most every benchmark, and the MT performance would see a notable uplift. But, BUT, none of that would have justified the increase in costs and wouldn't have done much for market volume.

In other words, a waste of money. Save the money and sell what you already have while you work on something MUCH better.
Click to expand...
why are they porting RDNA3.5 to N3 then?
 
  • Like
Reactions: Tlh97 and MoistOintment
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom