Especially when Intel's next Gen will be supporting 8800+
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Zen 5 rumours
- Thread starter Grim5
- Start date
Especially when Intel's next Gen will be supporting 8800+
Probably not across 4 slots. That would push Intel’s power consumption to new levels of insane.
Caporegime
Brut force.
Typically Intel, AMD gave us 3D Stacking and Intel are trying to compete by turning the IMC up to 11 along with the CPU cores.
Ok, you have to work with what you've got but i also find it quite annoying when tech tubers justify this lack of real progress by trying to tell their audience its all good because if you run 8000Mhz RAM on Intel.....
So that's £250 ram, to go with your £200 cooler and your £400 motherboard to get with in 5% of something that is happy with cheap and cheerful RAM, board and cooler....
Give me a break you people are desperate.....
Last edited:
Soldato
- Joined
- 28 Sep 2018
- Posts
- 2,515
how are the two related?
Soldato
- Joined
- 9 Jun 2011
- Posts
- 3,606
Brut force.
Typically Intel, AMD gave us 3D Stacking and Intel are trying to compete by turning the IMC up to 11 along with the CPU cores.
Ok, you have to work with what you've got but i also find it quite annoying when tech tubers justify this lack of real progress by trying to tell their audience its all good because if you run 8000Mhz RAM on Intel.....
So that's £250 ram, to go with your £200 cooler and your £400 motherboard to get with in 5% of something that is happy with cheap and cheerful RAM, board and cooler....
Give me a break you people are desperate.....
The speculated 25% IPC increase on 16-core CCXs with extended cache, along with reduced power consumption, could potentially lead to a game-changing outcome. Intel is trying to do what it does best BS marketing - Just like the 48 core 5ghz [ with the phazzed chiller .... ]
The MC is on chip, adding more and/or faster memory, the more work needs to done and this need power.
Soldato
- Joined
- 28 Sep 2018
- Posts
- 2,515
On Intel?
Do you know the system agent power draw on an intel cpu going from stock 0.95v to 1.35v? I want a number. Not a wall of text.
I see. Let me think about it…
8 core CCx not 16
First zen 5 benchmark
128 zen5 cores gives 123k cinebench r23 score. Operating at 3.85ghz and with 512mb L3 cache. This is slightly faster than the 110k score that Zen4 EPYC Genoa gets, which also has 512mb L3 cache and a clock speed of 3.75ghz
videocardz.com
128 zen5 cores gives 123k cinebench r23 score. Operating at 3.85ghz and with 512mb L3 cache. This is slightly faster than the 110k score that Zen4 EPYC Genoa gets, which also has 512mb L3 cache and a clock speed of 3.75ghz
Early Dual AMD Zen5 CPU system scores 123K points in Cinebench - VideoCardz.com
AMD Zen5 allegedly tested with Cinebench R23 AMD Zen5 architecture debut is not expected anytime soon, but early engineering samples are apparently already running in the labs. Moore’s Law is Dead claims to have a screenshot featuring a Zen5 pre-release hardware running a popular benchmark. The...
videocardz.com
Last edited:
Caporegime
Its difficult to know....
Intel's Sapphire Rapids, with 56 cores scored 132K with LN2 at 5.4Ghz
Zen 3 with 64 cores scored 121K at the same 5.4Ghz on LN2.
There are no Zen 4 Threadrippers yet.
This as an early engineering sample on standard air cooling scored 123K, its impressive but its also 128 cores, clock speed is frankly unknown, windows is just reporting the highest boost clock. how MLID can draw any conclusions from that is frankly ridiculous.
PS: Zen 5c will have 16 core CCX.
Intel's Sapphire Rapids, with 56 cores scored 132K with LN2 at 5.4Ghz
Zen 3 with 64 cores scored 121K at the same 5.4Ghz on LN2.
There are no Zen 4 Threadrippers yet.
This as an early engineering sample on standard air cooling scored 123K, its impressive but its also 128 cores, clock speed is frankly unknown, windows is just reporting the highest boost clock. how MLID can draw any conclusions from that is frankly ridiculous.
PS: Zen 5c will have 16 core CCX.
Last edited:
Caporegime
I suppose....
EPYC 9564 X2 (Zen 4) scored 108K at 3.7Ghz reported, that's actually 192 cores but R23 is limited to 128 cores.
We also don't know if that was overclocked.
The difference between 121K and 108K is +12%.
EPYC 9564 X2 (Zen 4) scored 108K at 3.7Ghz reported, that's actually 192 cores but R23 is limited to 128 cores.
We also don't know if that was overclocked.
The difference between 121K and 108K is +12%.
Last edited:
Caporegime
If we are really adventurous....
Add the Zen 3 vs Zen 4 9% IPC in R23 to Zen 3 you get 132K < that's the only thing we can be sure of*
Add 12% to that you get 148K.
Add 12% to 38,650 for 16 cores you get 43,300.
*we have no idea of the real clock speeds for the Zen 4 example scoring 108K or the Zen 5 sample seen here. Wich just shows how little we can learn from this, almost nothing.
Add the Zen 3 vs Zen 4 9% IPC in R23 to Zen 3 you get 132K < that's the only thing we can be sure of*
Add 12% to that you get 148K.
Add 12% to 38,650 for 16 cores you get 43,300.
*we have no idea of the real clock speeds for the Zen 4 example scoring 108K or the Zen 5 sample seen here. Wich just shows how little we can learn from this, almost nothing.
Last edited:
Its hard to make conclusions about either IPC or clock speed (or power efficiency, affecting boost clocks), but it is obvious that there is an increase. MLID cautiously said 15%
I think it is more.
If you take a core running a single thread for 100%, same core running SMT 2 threads has 140% performance
Zen4 scores 108k on 2x96 cores running a 256 thread workload. 128 cores are running a single thread, 64 run 2 threads on average. So normalised to single core its 496 points per core (108000/(128+64*1.4))
Zen5 scores 123k on 2x64 cores, all run 2 threads, 686 per core (123000/(128*1.4))
this translates Zen5 is 38% faster ((686-496)/496) when normalised to single core. That includes clock speed and IPC (P.S. and memory, rumored to support faster DDR5)
Its back of a napkin math, but clearly Zen5 will be substantial. This is engineering sample and it is probably a year until desktop release.
I think it is more.
If you take a core running a single thread for 100%, same core running SMT 2 threads has 140% performance
Zen4 scores 108k on 2x96 cores running a 256 thread workload. 128 cores are running a single thread, 64 run 2 threads on average. So normalised to single core its 496 points per core (108000/(128+64*1.4))
Zen5 scores 123k on 2x64 cores, all run 2 threads, 686 per core (123000/(128*1.4))
this translates Zen5 is 38% faster ((686-496)/496) when normalised to single core. That includes clock speed and IPC (P.S. and memory, rumored to support faster DDR5)
Its back of a napkin math, but clearly Zen5 will be substantial. This is engineering sample and it is probably a year until desktop release.
Last edited:
The biggest issue for me is that clocknspeed; there's just no way to know if it's accurate in windows and whether or not that's the clocks under load or idle. I've seen a lot of engineering sample cpu leaks over the last few years and what is common amongst them is that the first couple ES revisions have extremely low clocks - for example I've seen an Intel CPU go from 1.6ghz on an ES to over 5ghz at launch.
If the windows clock speed is accurate, then zen5 is quite far along already and this is likely not an early engineering sample and a Zen5 launch is closer than people expect.
If the windows clock speed is accurate, then zen5 is quite far along already and this is likely not an early engineering sample and a Zen5 launch is closer than people expect.
Last edited:
Soldato
I know it. Down to the dot
I think AMD have mentioned new AM4 parts aren’t out of the question. Some time ago IIRC.
A strong APU would be great, but I think we might be in the minority.
Caporegime
AM4 is still an extremely popular platform, AMD can't ignore it...
They can, the question is will they. AMD would probably like to do more AM4 chips, but they would need to support all the AM4 chipsets, or they would be a lot of hate. The board makers would not like having to add support for new chips, they want to sell new stuff and forget about the older stuff.