Info 64MB V-Cache on 5XXX Zen3 Average +15% in Games

[Kedas]

Well we know now how they will bridge the long wait to Zen4 on AM5 Q4 2022.
Production start for V-cache is end this year so too early for Zen4 so this is certainly coming to AM4.
+15% Lisa said is "like an entire architectural generation"

 

[Insert_Nickname]

EDIT: So there may be a nugget of truth to this. I still find it hard to believe as there would be very different cache latencies and therefore performance in certain situations.

A dual chiplet 8 core with the full 64MB cache could have worked. But with the X3D there is very little case for such a model. It would also suffer from the latency issues inherent in the dual chiplet design.

 

[leoneazzurro]

Proof? The performance would be quite different. Just look at the 3300X compared to the 3100. I've never heard of a dual chiplet 5600X or 5800X.

EDIT: So there may be a nugget of truth to this. I still find it hard to believe as there would be very different cache latencies and therefore performance in certain situations.

What Techpowerup is describing is a different thing. The parts described there haven't both CCDs active. Only one of the CCD works, the other is completely disabled. These may be parts that were built to be 5900X or 5950X but for some reasons (packaging scrap) don't work as intended and thus they are recycled as 5600X or 5800X, and behave as them.

 

[FangBLade]

You don't know crap do you? A 5800X only has one CCD, just like the 5700X. It is not a 4 + 4 design. Yet you continue to bash AMD and promote Intel while claiming that you will continue to buy AMD. This post sums it up perfectly.

Let's use that and replace CPU's with TV's.

X brand TV's have had a good run, but Y brand TV's are doing way better. Y brand TV's will advance at a rate the X brand TV's simply cannot. Yet I will but brand X's TV within the next year, validating my opinion.

You can't make this stuff up! I'm going to buy into what I think is an inferior platform despite Intel being so great?

He lives in bulldozer era, just like his fellow Henry, unfortunately for them AMD is in excellent shape now, they bought Xilinx and Pensando which is biggest purchase in the history, They are growing at a rapid rate every year, and Intel fans will have to accept that, no more small AMD, they will fight evenly with Intel every year, and seeing future data center products, I predict a bright future for them and more $billions in wallet.

 

[Thibsie]

What Techpowerup is describing is a different thing. The parts described there haven't both CCDs active. Only one of the CCD works, the other is completely disabled. These may be parts that were built to be 5900X or 5950X but for some reasons (packaging scrap) don't work as intended and thus they are recycled as 5600X or 5800X, and behave as them.

Indeed, in the case of my 5600x at least, this a dual CCD with (this is how I understood it), both active CCDs but no active core on the second CCD.
As both CCDs are active it eats more power though.

 

[MadRat]

/wrong thread

 

[moinmoin]

And they need to shrink the IOD.

I/O interfaces essentially can't really be shrunken. That's makes IODs big.

 

[Hans Gruber]

You don't know crap do you? A 5800X only has one CCD, just like the 5700X. It is not a 4 + 4 design. Yet you continue to bash AMD and promote Intel while claiming that you will continue to buy AMD. This post sums it up perfectly.

Let's use that and replace CPU's with TV's.

X brand TV's have had a good run, but Y brand TV's are doing way better. Y brand TV's will advance at a rate the X brand TV's simply cannot. Yet I will but brand X's TV within the next year, validating my opinion.

You can't make this stuff up! I'm going to buy into what I think is an inferior platform despite Intel being so great?

Please explain the significant increase in power draw for the 5800x. 105w tdp vs. 65w tdp for the 5700x? The max boost power draw for the 5700x is 76w (PPT) vs. 142w (PPT) for the 5800x. The 5800x offers 100mhz more on the boost clock and a 400mhz higher base clock than 5700x. 5800x owners have to manually under volt their CPU's to achieve power usages and temps similar to the 5700x. You can manually OC a 5700x (B2 stepping) to 5800x stock levels and beyond without the same stock power usage of the 5800x.

In past generations AMD explained their binning process. The 3600's were cut down chips that could not pass the binning process for higher level chips but worked perfectly fine for a 6 core 12 thread CPU. You know better than me. I don't know crap as you said. Explain the increased power draw of the 5800x over the 5700x? Then explain why 5800x owners can manually under volt their CPU's without any issues. Basic math tells us that a 16 core 32 thread Zen 3 CPU is double an 8 core 16 thread Zen 3 CPU. So cutting in half or disabling 8 cores is not the same thing but the results are the same. The 5950x is the best binned CPU's. If you have CPU's that do not pass the binning process for the 5950x, they cut that chip in half giving you a 5800x.

FYI, I bought into the AMD platform from the original (R31200) Zen CPU. When people struggle to upgrade an AMD bios on older boards. I pop in my R3 1200.

 

[coercitiv]

Please explain the significant increase in power draw for the 5800x.

I don't know crap as you said. Explain the increased power draw of the 5800x over the 5700x?

Here's the crap you need to know:


See the ~500Mhz delta? There's your power right there. You're welcome.

Then explain why 5800x owners can manually under volt their CPU's without any issues.

Not only 5800X owners can undervolt their CPUs, ALL CPUs can be undervolted provided the users have good quality components in their builds and they're willing to eat into the safety margins put in place to ensure reliability using a myriad of PC components and cooling conditions.

You buy a better than average quality PSU and better than average motherboard, you put better than average cooling in place... and voila, your compounded margins are significantly better than someone else's even with identical quality CPU. On top of that the person who undervolts also performs extensive stability testing to ensure the thinner voltage margins are enough for THEIR use case.

 

[Thunder 57]

What Techpowerup is describing is a different thing. The parts described there haven't both CCDs active. Only one of the CCD works, the other is completely disabled. These may be parts that were built to be 5900X or 5950X but for some reasons (packaging scrap) don't work as intended and thus they are recycled as 5600X or 5800X, and behave as them.

Thanks. That's what I get for second guessing myself and not reading the whole thing first.

Please explain the significant increase in power draw for the 5800x. 105w tdp vs. 65w tdp for the 5700x? The max boost power draw for the 5700x is 76w (PPT) vs. 142w (PPT) for the 5800x. The 5800x offers 100mhz more on the boost clock and a 400mhz higher base clock than 5700x. 5800x owners have to manually under volt their CPU's to achieve power usages and temps similar to the 5700x. You can manually OC a 5700x (B2 stepping) to 5800x stock levels and beyond without the same stock power usage of the 5800x.

In past generations AMD explained their binning process. The 3600's were cut down chips that could not pass the binning process for higher level chips but worked perfectly fine for a 6 core 12 thread CPU. You know better than me. I don't know crap as you said. Explain the increased power draw of the 5800x over the 5700x? Then explain why 5800x owners can manually under volt their CPU's without any issues. Basic math tells us that a 16 core 32 thread Zen 3 CPU is double an 8 core 16 thread Zen 3 CPU. So cutting in half or disabling 8 cores is not the same thing but the results are the same. The 5950x is the best binned CPU's. If you have CPU's that do not pass the binning process for the 5950x, they cut that chip in half giving you a 5800x.

FYI, I bought into the AMD platform from the original (R31200) Zen CPU. When people struggle to upgrade an AMD bios on older boards. I pop in my R3 1200.

OK, explain the reason why the 3700X is 65W vs the 3800X at 105W? @coercitiv already has this covered though.

 

[Hans Gruber]

Thanks. That's what I get for second guessing myself and not reading the whole thing first.



OK, explain the reason why the 3700X is 65W vs the 3800X at 105W? @coercitiv already has this covered though.

You quote a person not directly in this conversation stating exactly what I said about how AMD bins their chips. You wrongly assume that AMD makes CPU's on silicon platters for specific models. They make Zen 3 chips on the same silicon and bin them according to their performance after manufacturing. You should second guess yourself.

When I write what I write, I ballpark my statements until someone insults my knowledge of how things work. Below you can see a massive 1-2% performance increase that contradicts the Gamers Nexus single slide offered by another forum member. The max speed for the 5800x is 4.75ghz and the max speed for the 5700x is 4.7ghz in the CPU benchmark submissions.

Please click the green human to continue - UserBenchmark

In the past AMD admitted they juice their chips because some CPU's performed better with lower power and other required more power in their binning process. They did this to maximize silicon yields. Juicing chips means they apply more stock voltage than required. They do this to offset the limited number of CPU's that require more voltage. I have no issue with this practice.

During Zen 2, review sites said to skip the 3800x because the 5700x was a better value. In Zen 3, AMD removed the 5700x SKU until well over a year after the release of Zen 3.

On a side note I said that ram is ram and the memory issues that plagued AMD in earlier versions of the Zen architecture (B350/X370) were recently fixed in the latest AGESA bios. This means that all Zen 2 and later CPU's can run memory @ 3800mhz without infinity fabric issues. I would also like to point out that TSMC silicon has been excellent. AMD used to own Glo Flo and sold it off when they were almost bankrupt.

I am also the forum member who told people to buy 3600mhz binned ram and call it a day. All 3600mhz can run 3800mhz speeds without any issues. To get the most out of Ryzen CPU's you want your memory running @ 3800mhz. But I know nothing.

I am an Intel fanboy. My last intel build was in 2013.

 

[DAPUNISHER]

LULZ at Userbenchmark! Worst of the worst garbage tier site.

 

[Makaveli]

You quote a person not directly in this conversation stating exactly what I said about how AMD bins their chips. You wrongly assume that AMD makes CPU's on silicon platters for specific models. They make Zen 3 chips on the same silicon and bin them according to their performance after manufacturing. You should second guess yourself.

When I write what I write, I ballpark my statements until someone insults my knowledge of how things work. Below you can see a massive 1-2% performance increase that contradicts the Gamers Nexus single slide offered by another forum member. The max speed for the 5800x is 4.75ghz and the max speed for the 5700x is 4.7ghz in the CPU benchmark submissions.

Please click the green human to continue - UserBenchmark

In the past AMD admitted they juice their chips because some CPU's performed better with lower power and other required more power in their binning process. They did this to maximize silicon yields. Juicing chips means they apply more stock voltage than required. They do this to offset the limited number of CPU's that require more voltage. I have no issue with this practice.

During Zen 2, review sites said to skip the 3800x because the 5700x was a better value. In Zen 3, AMD removed the 5700x SKU until well over a year after the release of Zen 3.

On a side note I said that ram is ram and the memory issues that plagued AMD in earlier versions of the Zen architecture (B350/X370) were recently fixed in the latest AGESA bios. This means that all Zen 2 and later CPU's can run memory @ 3800mhz without infinity fabric issues. I would also like to point out that TSMC silicon has been excellent. AMD used to own Glo Flo and sold it off when they were almost bankrupt.

I am also the forum member who told people to buy 3600mhz binned ram and call it a day. All 3600mhz can run 3800mhz speeds without any issues. To get the most out of Ryzen CPU's you want your memory running @ 3800mhz. But I know nothing.

I am an Intel fanboy. My last intel build was in 2013.

Source userbenchmark vs GN I know who I trust more....

 

[coercitiv]

I am an Intel fanboy. My last intel build was in 2013.

Maybe you should buy Intel more often, if only to get a better grasp of what you're talking about.

Las bit of advice, never quote Userbenchmark data again. That "Effective speed" rating you see is NOT what you think it is.

 

[moinmoin]

When I write what I write, I ballpark my statements until someone insults my knowledge of how things work. Below you can see a massive 1-2% performance increase that contradicts the Gamers Nexus single slide offered by another forum member.
UserBenchmark: AMD Ryzen 7 5700X vs 5800X

userbenchmark.com is not a site you want to quote seriously if you don't want your knowledge to be insulted.

 

[Makaveli]

Maybe you should buy Intel more often, if only to get a better grasp of what you're talking about.

Las bit of advice, never quote Userbenchmark data again. That "Effective speed" rating you see is NOT what you think it is.

This if you ever want to be taken seriously never ever have Userbenchmark as a source. I wasn't even part of this discussion and as soon as I saw that I had to post.

 

[Ranulf]

Its been awhile since I've read a userbenchmark review. Their take on the 5800x in 2020 is rather hilarious both then and now. "It still offers far better value than a 5900x".

 

[Thunder 57]

You quote a person not directly in this conversation stating exactly what I said about how AMD bins their chips. You wrongly assume that AMD makes CPU's on silicon platters for specific models. They make Zen 3 chips on the same silicon and bin them according to their performance after manufacturing. You should second guess yourself.

When I write what I write, I ballpark my statements until someone insults my knowledge of how things work. Below you can see a massive 1-2% performance increase that contradicts the Gamers Nexus single slide offered by another forum member. The max speed for the 5800x is 4.75ghz and the max speed for the 5700x is 4.7ghz in the CPU benchmark submissions.

Please click the green human to continue - UserBenchmark

In the past AMD admitted they juice their chips because some CPU's performed better with lower power and other required more power in their binning process. They did this to maximize silicon yields. Juicing chips means they apply more stock voltage than required. They do this to offset the limited number of CPU's that require more voltage. I have no issue with this practice.

During Zen 2, review sites said to skip the 3800x because the 5700x was a better value. In Zen 3, AMD removed the 5700x SKU until well over a year after the release of Zen 3.

On a side note I said that ram is ram and the memory issues that plagued AMD in earlier versions of the Zen architecture (B350/X370) were recently fixed in the latest AGESA bios. This means that all Zen 2 and later CPU's can run memory @ 3800mhz without infinity fabric issues. I would also like to point out that TSMC silicon has been excellent. AMD used to own Glo Flo and sold it off when they were almost bankrupt.

I am also the forum member who told people to buy 3600mhz binned ram and call it a day. All 3600mhz can run 3800mhz speeds without any issues. To get the most out of Ryzen CPU's you want your memory running @ 3800mhz. But I know nothing.

I am an Intel fanboy. My last intel build was in 2013.

OK so now we're completely forget that we were talking about the 5800X being a 4+4 CPU while the 5700X was a "true" 8 core. That fell apart really fast so now we're trying to argue TDP. Userbenchmark is not a source anyone should use. Since I can't find a review of the 5700X that tested power, I'll do the next best thing.

See the performance of the 5700X vs the 5800X in multithreading.

"The multi-threaded results reveal that the Ryzen 7 5800X is 12% faster than the stock Ryzen 7 5700X, but engaging PBO shrinks the delta between the chips to less than one percent. But, again, the 5700X and 5800X are remarkably similar after tuning."

So there is a performance difference at stock that PBO largely eliminates. I guess PBO just bumps the clock and power up to near 5800X levels? Nah, the 5700X uses less power because its a true 8 core.

And why are you bringing RAM into this? I never mentioned RAM and agree with what you said about it. I also never called you an Intel fanboy, but your posts just do not make sense.

 

[Hans Gruber]

Its been awhile since I've read a userbenchmark review. Their take on the 5800x in 2020 is rather hilarious both then and now. "It still offers far better value than a 5900x".

View attachment 63693

There is no user benchmark review. It's called a benchmark like Cinebench which is a waste of time. It's simple an apples to apples comparison running the same feature set. The reason people hate userbenchmark simply for a snapshot. In the past it was always slanted towards Intel CPU's. We are comparing Zen 3 AMD CPU's vs AMD CPU's. I look at the best benchmarks and clocks. It's not scientific.

 

[Hans Gruber]

userbenchmark.com is not a site you want to quote seriously if you don't want your knowledge to be insulted.

Now I can see why your post count is so high in just 5 years here on the forum.

 

[moinmoin]

There is no user benchmark review. It's called a benchmark like Cinebench which is a waste of time. It's simple an apples to apples comparison running the same feature set. The reason people hate userbenchmark simply for a snapshot. In the past it was always slanted towards Intel CPU's. We are comparing Zen 3 AMD CPU's vs AMD CPU's. I look at the best benchmarks and clocks. It's not scientific.

Just in case anybody needs an update on userbenchmark.com:

They are the laughing stock among benchmark sites.

Some threads for your entertainment (all between 2019-2020, after which we as a forum apparently finally stopped caring about the site):

Ryzen 3000: userbenchmark changed the score weights in favor of intel

As predicted benchmark tool(s) are changed after Ryzen 3000 (to favor intel) The first one, and there is always a reason why they needed to change it now. https://wccftech.com/userbenchmark-updates-cpu-ranking-by-lowering-multi-core-importance-and-raising-single-core/

forums.anandtech.com

Question - Could Userbenchmark be more dishonest?

First, they changed their scoring methodology just after Zen 2 came out. Now, I was looking for a laptop for my sister, and was searching for info about the Ryzen 3500U. These clowns try to claim that a Whiskey Lake chip would be better for gaming than a Ryzen 3500U. She certainly doesn't need...

forums.anandtech.com

Discussion - UserBenchmark: 4 cores are enough

UserBenchmark.com apparently removed 64 thread comparison from their CPU benchmark, maximum is now 8. I guess nobody needs more than 4 cores, right??? :)

forums.anandtech.com

Question - More anti-AMD "Stealth Marketing" by UserBenchmark...

Basically, a blog post by "CPUPro", I guess, on UserBenchmark, talking about the new 3300X CPU, and basically dissing it, and saying that AMD is now effectively walking back their "moar coars" marketing, and now saying that "only quad-core is enough", and then claiming that the (as of yet...

forums.anandtech.com

 

[FangBLade]

Now I can see why your post count is so high in just 5 years here on the forum.

You can cry all day, but you won't see AMD bulldozer era again, and learn to live with it, Pat Gelsinger also accepted that, you should too.

 

[Thunder 57]

https://www.reddit.com/r/intel/comments/g36a2a

Says about all you need to know.

 

[Hans Gruber]

OK so now we're completely forget that we were talking about the 5800X being a 4+4 CPU while the 5700X was a "true" 8 core. That fell apart really fast so now we're trying to argue TDP. Userbenchmark is not a source anyone should use. Since I can't find a review of the 5700X that tested power, I'll do the next best thing.

See the performance of the 5700X vs the 5800X in multithreading.

"The multi-threaded results reveal that the Ryzen 7 5800X is 12% faster than the stock Ryzen 7 5700X, but engaging PBO shrinks the delta between the chips to less than one percent. But, again, the 5700X and 5800X are remarkably similar after tuning."

So there is a performance difference at stock that PBO largely eliminates. I guess PBO just bumps the clock and power up to near 5800X levels? Nah, the 5700X uses less power because its a true 8 core.

And why are you bringing RAM into this? I never mentioned RAM and agree with what you said about it. I also never called you an Intel fanboy, but your posts just do not make sense.

Great so we are now arguing over how a 5950x is cut in half to create a 5800x. I can live with that. Instead of cutting vertically, we are cutting it horizontally making a a 2x CCD 4 core vs. a 8 core single CCD. But power consumption issues still comes about. Normally people add power to get better performance. In the case of the 5800x they are reducing power to get it inline with the power consumption of a 5700x. On paper the 5800x should be a better chip than the 5700x. In reality the extra 5800x (50mhz) after manually overclocking 5700x utilizes nearly the same results.

Instead of being a 4x4 design (cut in half 5950x) is that no really a 8+0 instead of a 4x4? AMD only knows but the benefit of the 5800x on paper is for reduced core to core latency that does not show up when you compare a overclocked 5700x vs. a 5800x. Based on thousands of 5800x userbenchmark submissions. The 5800x maxes out at 4.75ghz vs 4.7ghz for an overclocked 5700x that uses 40% less power unless the 5800x is undervolted.

If you cut a 5950x in half vertically you have a single CCD 8 core CPU. If you cut it horizontally you would have a an extra functional CCD core that in theory is supposed to reduce latency between cores. I don't know if that means 2x2 or 4x4. One could argue that only one CCD is functioning in both but the 5800x has an extra functioning CCD with half the core count for a total of 8 cores.

The reason why the power consumption is higher on the 5800x is because it's still functioning with the same power envelope as a 5950x with either half the cores cut or disabled.

 

[Hans Gruber]

https://www.reddit.com/r/intel/comments/g36a2a

Says about all you need to know.

Reddit has a reputation that is less than stellar. Being banned from reddit is like being discredited by the National Enquirer in favor of the New York Post. For the record, I did not say userbenchmark was a complete and comprehensive system review. It's a quick benchmark that assess your system.

 

[igor_kavinski]

I have a question. What's 5700X and 5800X got to do with this thread?