[SOLVED] Questioning the Ryzen 3000 Series, 12 Cores/16 Cores (7nm)? FX Stunt?

[valeman2012]

The Ryzen 3000 series with 12 cores/ 24 threads or 16 Cores/32 Threads? That all i been seeing...all computer news articles seem to be following it...is it true or going be 8 Cores /16 Thread again....

 

[TerryLaze]

If that were true, an overclocked 7700K would routinely roughly equal the 9900K's frame rates...(when both are running at/near same clock speeds, ~all core of 4.6-4.7 GHz)

It does not, and particularly in gaming/streaming scenarios (naturally, they are pretty close in 4k gaming, i.e., in massively GPU-limited scenarios).

Several games are doing well on more cores/threads, such as BF1 and BF5 among others...; comparing CPUs that lack SMT/hyperthreading such as the 9700K, however, do seem to indicate that real cores are more important than just thread count, as the 'only' 8 c/8t CPU still smashes in performance...

"an overclocked 7700K would routinely roughly equal the 9900K's frame rates"
it would and it does,even at stock most people would call this even,it's exactly 10% behind,the 2700x is farther behind the 9900k than the 7700k is.
This is an extensive multiplayer bench.
Yes of course,doing CPU streaming (like it's 1999) won't be as good as on a CPU with more cores but nobody,certainly not me,argued that.
https://www.techspot.com/review/1754-battlefield-5-cpu-multiplayer-bench/

 

[mdd1963]

BF1 and BF5, even at 'only' 1080P Ultra are still partially GPU limited, although we likely can't really prove this until the next even faster batch of GPUs arrive...

The above BF5 charts, if accurate, really show a larger disparity with Intel/AMD than I was expecting with BF5 multipayer, so it would be interesting to see those results confirmed/mirrored elsewhere. (as 64 player games are very common in BF1/BF5, those results are indeed important)

Certainly, I am not 'hating on the 7700K' , as I've been using one the last 26 months....; indeed it is still quite competent. (Hope not to still be on it until Intel finally releases 10 nm desktop CPUs in ..2021?) I do notice that I'm often first into most BF1 servers waiting on all other players to join over the next 10-15 seconds, but, I have both a fast connection (700-750 Mbps)/good ping (30 ms), and still relatively fast storage for OS and game (960 EVO)

 

[TerryLaze]

BF1 and BF5, even at 'only' 1080P Ultra are still partially GPU limited, although we likely can't really prove this until the next even faster batch of GPUs arrive...

As are a lot of games,a fact that most people are more then happy to fully ignore just so they can spam that amd is within 10% of intel and doom and gloom and cats and dogs...
Right now we have exactly zero idea on how fast the 9900k actually is in gaming because there is no GPU in the world that can keep up with it.

 

[NightHawkRMX]

I love the 9900k, but for most times a 2700x will rival it in high res gaming.

 

[drivinfast247]

Not really Intel Fanboys issue, but "Most" AMD Fanboys" issue are just plain stupid or something near that. AMD Reviewers bias, AMD Fanboys Believes "AMD being ahead" means better than Intel (CPU) or NVIDIA (GPU) in gaming performance overall. https://forums.tomshardware.com/threads/how-many-years-will-it-always-take-amd-cpu-to-beat-intel-in-gaming-who-is-the-real-sub-par-product-here.3413935/post-20699249

Calm down, kid! No need for name calling.

 

[digitalgriffin]

I do not think AMD will even beat Intel in gaming ....they can be at 7nm++ for sake and Intel can be at 14nm+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ and still be able to beat.

AMD is good at beating Intel at Cinebench (a benchmark software) mostly. That about it.

And this is where you lose objectivity.

IF there is a 15% ipc gain and it does clock to 5 GHz then all bets are off.

The source of the latest amd leaks comes from a major supplier (distributor) of computer hardware. So this leak is taking credibility.

If Intel still holds the crown then good on them. Ill be happily compiling and ai training on 12/24 threads when they get here.

 

[valeman2012]

And this is where you lose objectivity.

IF there is a 15% ipc gain and it does clock to 5 GHz then all bets are off.

The source of the latest amd leaks comes from a major supplier (distributor) of computer hardware. So this leak is taking credibility.

If Intel still holds the crown then good on them. Ill be happily compiling and ai training on 12/24 threads when they get here.

AMD Fanboys : Cinebench scores is better means Intel is beaten rofl

Those rumors are not creditable, but AMD fanboys would say so...rofl...

 

[digitalgriffin]

AMD Fanboys : Cinebench scores is better means Intel is beaten rofl

Those rumors are not creditable, but AMD fanboys would say so...rofl...

We just have wait and see if AMD is really going pull off another bulldozer stunt.

Hey my jury is out. But using an analytical eye is crucial. My main rig is a 3770K @ 4.4GHz. So guess what that doesn't make me an AMD fan boy. I'm about delivering the best value for the money. Right now I know Intel can't deliver that kind of value that AMD can at a similar price point. And I may have deep pockets, but even I'm not "Money is no object." builder. I build with quality components that will last me at least 5, 6, or 7 years. If I were "money if no object" kind of person I would admit tingly get an intel 9700K and a 360mm radiator (Not an AIO) with a water cooled NVIDIA 2080Ti (With an extended warranty on the later). But hey I'm not money foolish.

Cinebench is just one metric and doesn't account for the average IPC of 15% that is rumored. But you are right these are just rumors. But every day we get closer, the more these rumors are collaborated from big players. And we are very close.

The IPC improvements comes from improvements in CPI average but also a larger Cache with an improved latency coherency system for the infinity fabric.

 

[valeman2012]

We just have wait and see if AMD is really going pull off another bulldozer stunt. 🤣
Upgrade Now : Buy Ryzen 7 3700 (a 12C/24T) if is not stunt. If it is...then you wasted money.
Save Money Now : If is a Bulldozer stunt Stick with Intel 8th Gen or buy 9th Gen less spending on upgrade as their performance last a long time,

"an overclocked 7700K would routinely roughly equal the 9900K's frame rates"
it would and it does,even at stock most people would call this even,it's exactly 10% behind,the 2700x is farther behind the 9900k than the 7700k is.
This is an extensive multiplayer bench.
Yes of course,doing CPU streaming (like it's 1999) won't be as good as on a CPU with more cores but nobody,certainly not me,argued that.
https://www.techspot.com/review/1754-battlefield-5-cpu-multiplayer-bench/

BF5 is AMD game too...and intel once again beats them despite being behind in nm process rofl.

 

[Finstar]

Zen+ is 3% behind coffee lake in ipc. Zen 2 has been rumoured to have 13-20% higher ipc over Zen+ which is a very realistic guess as Zen+ had 4% better ipc than Zen and that was just an improved die shrink. That alongside 4.5-4.7GHz rumoured clock speeds and Zen 2 will be walking all over Coffee lake.

 

[Karadjgne]

Wayne: So Garth, which is better? The red car or the blue car?
Garth: Well Wayne, the blue car is a 2016 ZR-1 Corvette pushing 600Hp and will do 250mph. The red car is a 2012 Dodge Carger R/T pushing 400Hp and only does 180mph, so the blue car is better.
Officer: Yes, but the speed limit is 55mph....

 

[NightHawkRMX]

That analogy is very true.

 

[NightHawkRMX]

AMD=Intel at least until new GPUs become available that can cause a CPU bottleneck on one but not the other.

 

[rigg42]

Zen+ is 3% behind coffee lake in ipc.

I was about to tell you that the IPC difference in gaming workloads was probably closer to the 10% that was discussed earlier in the thread. Then I decided to poke around for a good real world comparison. I came across an interesting video from Hardware Unboxed:

View: https://www.youtube.com/watch?v=hmhBgLSIneQ&t=628s


The 2600x at stock all core is only 100mhz faster than the 9400f. I was surprised to see how close the results were over 18 games. I was also surprised to see how much stepping up to a z390 and 3400 ram improved performance on the 9400f. Gaming load IPC might be considerably closer than I originally thought.

 

[valeman2012]

Zen+ is 3% behind coffee lake in ipc. Zen 2 has been rumoured to have 13-20% higher ipc over Zen+ which is a very realistic guess as Zen+ had 4% better ipc than Zen and that was just an improved die shrink. That alongside 4.5-4.7GHz rumoured clock speeds and Zen 2 will be walking all over Coffee lake.

and will remain behind Intel 9th Gen (Coffee Lake R ~ 14nm) 🤣

 

[rigg42]

and will remain behind Intel 9th Gen (Coffee Lake R ~ 14nm) 🤣

Are you just trolling?

Nobody knows this for sure just yet. All signs point to AMD either beating Intel strait up, or closing the gap to the extent that the value argument makes any performance differences irrelevant to anyone but the most stubborn Intel fanbois.

 

[valeman2012]

Are you just trolling?

Nobody knows this for sure just yet. All signs point to AMD either beating Intel strait up, or closing the gap to the extent that the value argument makes any performance differences irrelevant to anyone but the most stubborn Intel fanbois.

Mostly AMD fanboys are being blindsided by AMD trolling.
Let wait and see if they are not going pull a FX stunt,.

 

[TerryLaze]

I was about to tell you that the IPC difference in gaming workloads was probably closer to the 10% that was discussed earlier in the thread. Then I decided to poke around for a good real world comparison. I came across an interesting video from Hardware Unboxed:

View: https://www.youtube.com/watch?v=hmhBgLSIneQ&t=628s


The 2600x at stock all core is only 100mhz faster than the 9400f. I was surprised to see how close the results were over 18 games. I was also surprised to see how much stepping up to a z390 and 3400 ram improved performance on the 9400f. Gaming load IPC might be considerably closer than I originally thought.

for a good real world comparison. I came across an interesting video from Hardware Unboxed:

Yeah you shouldn't look at canned in game benchmarks if you want a real world comparison,they are only good for comparing different GPUs on the same CPU.

The 2600x at stock all core is only 100mhz faster than the 9400f

Yeah and it "only" has twice the threads and SMT supposedly "only" gives a ~35-40% boost,real "close" there on the IPC front...

 

[NightHawkRMX]

We can argue all you want but we cant say anything definitive since it is all speculation.

If AMD does have a big IPC gain and increases clocks, they will will certainly be appealing to gamers and content creators alike. Will this happen? Only god knows. (Well maybe amd)

Lets wait, but i dont think we will be too disappointed. Fingers crossed.

 

[NightHawkRMX]

I also wouldnt try to compare IPC of intels current chips to AMDs current chips. For one, IPC is workload dependent and i would assume depends on what instructions are being used.
Even if you look at 2 "simmilar on paper" cpus, you cant determine ipc overall. In one app a r5 might have higher ipc than an i5. In another app an i5 might have higher ipc than a r5

I7 8700: 6/12 4.6 turbo
R5 2600x: 6/12 4.3 turbo
Since there clocks are only a few 300mhz apart and the 8700 is much faster, people draw conclusions immediately that the ipc must be much higher on the 8700 in every scenario.

I believe intels cpus do tend to execute more instructions per clock, but there is more to it than saying the 8700 is just 300mhz faster and performs a lot better in a few games so the ipc must be x% higher.
The 2 chips have way different architectures. Memory latency is different. Speculative memory control is different. The Cache config is different. Smt is different from ht. The turbo algorithms are different... Etc.
Since these also factor into the total chips performance, you cannot directly compare performance difference and chalk it up to different IPC.

I think ipc is an elusive thing that can not be used to compare cpus of completely different architectures.

 

[rigg42]

Yeah and it "only" has twice the threads and SMT supposedly "only" gives a ~35-40% boost,real "close" there on the IPC front...

I wasn't ripping the 2600x for only being 100 mhz faster. I was pointing out that the clock speeds are really close which makes this a good example of the actual IPC difference in real world gaming. Anyone investing in 4 thread, 6 thread, or 4 core/8 thread CPUs in 2019 is being extremely short sighted IMO. Unless you are buying a 1600 for $80 from micro center you shouldn't be buying any new CPU until after zen 2 specs are released anyway.

 

[rigg42]

I also wouldnt try to compare IPC of intels current chips to AMDs current chips. For one, IPC is workload dependent and i would assume depends on what instructions are being used.
Even if you look at 2 "simmilar on paper" cpus, you cant determine ipc.

I7 8700: 6/12 4.6 turbo
R5 2600x: 6/12 4.3 turbo
Since there clocks are only a few 300mhz apart and the 8700 is much faster, people draw conclusions immediately that the ipc must be much higher on the 8700k. This is true, but there is more to it than saying the 8700 is 300mhz faster and performs x% better so the ipc must be x% higher. The 2 chips have way different architectures. Memory latency is different. Speculative memory control is different. The Cache config is different. Smt is different from ht. The turbo algorithms are different... Etc.

I think ipc is an elusive thing that can not be used to compare cpus of completely different architectures.

I agree. I pointed this out earlier in the thread. I still think that the video I linked is a good example of the IPC difference between 2 very closely clocked CPU's from each team when both are presented with the same gaming load. Average this over an 18 game benchmark and it gives you a pretty good picture of how far AMD is behind in gaming performance clock for clock.

 

[valeman2012]

Yeah you shouldn't look at canned in game benchmarks if you want a real world comparison,they are only good for comparing different GPUs on the same CPU.

Yeah and it "only" has twice the threads and SMT supposedly "only" gives a ~35-40% boost,real "close" there on the IPC front...

Yup, AMD fanboys saying being close does mean it bested their completion (intel)

AMD good at getting "close" but does not make it better at anyway as AMD says.

 

[TerryLaze]

I agree. I pointed this out earlier in the thread. I still think that the video I linked is a good example of the IPC difference between 2 very closely clocked CPU's from each team when both are presented with the same gaming load. Average this over an 18 game benchmark and it gives you a pretty good picture of how far AMD is behind in gaming performance clock for clock.

You can't know that unless you see the CPU usage on both,new game engines SCALE meaning that they run more software threads the more hardware threads they find,that's what my previous comment was all about,if all these games scale and run twice the threads on the ZEN then it will get 35-40% more workload making the 10% less performance much less impressive.

 

[mdd1963]

if the recent Tom's article about Ryzen 3000 is accurate, now even the new R5 will have 8 cores (3600X at 4.8 GHz boost?), with the former lowly R3 newly having 6c/12t? Impressive!

(Intel has best move up their 10 nm desktop launch ASAP...; delaying 2021 is now just sad!)