Best Eight-Core CPU Battle: AMD Ryzen 7 3800X vs Intel Core i7-9700K

[Nick_C]

However, upon testing we found very little meaningful variation between the Corsair and the stock cooler. We went with the Corsair results to provide a level playing ground. You can see similar variation in the charts with the 3600X with the Spire vs the Corsair charted out - The difference between Spire and the Corsair, with PBO active (peak consumption), amounts to a 0.2 fps variance in 99th percentiles, while looking at average fps yields a 0.8 fps variation. AMD's coolers are superb.

Little meaningful variation for the AMD CPUs, maybe - however the Intel CPU would seem to use significantly more power, based on your own review that I linked to, so it's not exactly a level playing field just because the AMD CPUs don't really need water-cooling - and omits to test whether the Intel CPU would behave in the same way on an air cooler equivalent to that supplied with the 3800X.

 

[PaulAlcorn]

Little meaningful variation for the AMD CPUs, maybe - however the Intel CPU would seem to use significantly more power, based on your own review that I linked to, so it's not exactly a level playing field just because the AMD CPUs don't really need water-cooling - and omits to test whether the Intel CPU would behave in the same way on an air cooler equivalent to that supplied with the 3800X.

Therein lies the crux of the issue. Assuming we test with a cooler of purportedly equal rating, then we go down the rabbit hole of using multiple different coolers for test pools of 18 processors. Different coolers for each price band? Then there will be endless complaints of which coolers are chosen, the fact that we're using different coolers for different price bands thus making all results incomparable, and claims of bias and handicapping. Better to remove cooling from the equation if you can and provide one centralized cooling solution that provides full performance for all chips. This is one of the foundational principles of comparing things: Create a level playing field, and let the silicon compete on its own merits. That's complicated by the fact that Intel doesn't provide coolers. Again, this is not perfect, but I'm not aware of a perfect answer. I'm all ears if you have an idea, but you have to think from the viewpoint of someone that opposes your opinion, as no matter what we do, someone will oppose based on various factors. The counter-arguments can be made from either side in convincing fashion, as there are both positive and negative aspects to any number of possible approaches.

At times we will do drill-downs on the impact of cooling, and we do present test results with most high-end processors with multiple coolers throughout a subset of workloads for all the most recent articles to indicate the merits of stepping up to a better cooler. Also, if a stock cooler causes significant performance reductions, we'll drill down hard on that, like we did here: https://www.tomshardware.com/reviews/intel-core-i7-8700-cpu-review,5638-2.html

 

[Redneck5439]

We used a corsair h115i for both stock and OC on all processors, except where listed in the charts as otherwise.

Was a little confusing with how the test bench were listed. Good to know that the same cooling was used for both.

 

[PaulAlcorn]

Was a little confusing with how the test bench were listed. Good to know that the same cooling was used for both.

Yeah, sorry about that. I could've done a better job there. I'll work on improving there and include some more information in the test system area (hopefully in a less confusing way) moving forward.

 

[Redneck5439]

The stock cooler for the 3800X is designed for 124W, whereas the PPT (the peak power consumption possible) is 142W, as mentioned in the original review:

"Like the other Ryzen 7 and 9 chips, the 3800X comes with the capable Wraith Prism RGB cooler which is rated to dissipate up to 124W if you crank the fans up to high. Given the 3800X's maximum 142W PPT measurement, that means, at least on paper, that the Wraith Prism might come up a tad shy of dissipating the full heat output of the 3800X at stock settings. "

However, upon testing we found very little meaningful variation between the Corsair and the stock cooler. We went with the Corsair results to provide a level playing ground. You can see similar variation in the charts with the 3600X with the Spire vs the Corsair charted out - The difference between Spire and the Corsair, with PBO active (peak consumption), amounts to a 0.2 fps variance in 99th percentiles, while looking at average fps yields a 0.8 fps variation. AMD's coolers are superb.

You are correct, the bundled cooler provides an exceptional value, as we covered in the article under the "cooling" section. Hence, AMD wins that category.

These results are very consistent with XFR / PBO behavior on Ryzen 3000. Better cooling will net better boosts to a point and then even if there is overhead for further boosting behavior its just not going to happen. I have very good aftermarket cooling on my 3800X, however with PBO / AOC / manually adjusting EDC limits the best all core boosting behavior I could get was 4.35Ghz all core but would fall to 4.3Ghz all core under sustained load. Single core boosting would hit 4.5Ghz at times but other times would only be ~4.45Ghz. From countless forums and posts of other people on Ryzen 3000 this is very consistent with what other people have seen as well. This is also why to get the best overall performance out of Ryzen you have to manually overclock the processor. By overall I mean productivity and gaming.

Saying that in this review, even though a top of the line motherboard was used the manual overclock wass only 4.3Ghz PBO would have yielded better results, but only if configured properly. An optimized PBO will provide all core boosts of 4.3 - 4.35Ghz however if not set up properly will typically see all core boosting behavior of 4.2 - 4.25Ghz.

XFR / PBO with Ryzen 3000 in my opinion can be problematic, but the biggest problem I keep having is the manual overclock on the Ryzen 3800X. This could be attributed to early silicon and a poorly binned chip (ie loosing the silicon lottery) or it could be early drivers and bios. The linked article "https://www.tomshardware.com/reviews/amd-ryzen-7-3800x-review,6226-2.html" was referenced as "our overclocking efforts" leads me to believe that no new overclocking efforts were made in this current article. Please correct me if I'm wrong, but if this is the case the overclocking results would have been done using early bios and early chipset drivers on early silicon. Those 3 factors together I can understand how 4.3Ghz was the overclock limit. With the latest bios and latest chipset drivers on one of the best X570 boards available, unless that particular processor totally lost the silicon lottery it should have no issues hitting 4.4Ghz at no more than 1.35V. Using a relatively unsafe 1.41- 1.42V should be able to get to at least 4.45Ghz, but with that motherboard 4.5Ghz would almost be expected.

There is still some confusion here... What chipset drivers were used for the overclocking on the 3800X? Which bios was used? There has to be a reason why that processor needs 1.42V to only get to 4.3Ghz, literally every 3800X I have worked on have all hit 4.4Ghz with much less voltage.

 

[rickmeister]

I buy my components based on bang for the buck because I have a fixed budget for my builds. At this point AMD is offering better bang for the buck. I bought my 3800X from bestbuy two weeks ago for $280 (price match from Micro Center.) And an MSI B450 motherboard for under $100. RAM and HDD are a wash so in my price range I was able to get a lot more performance for my money especially when reusing my GTX 1060.

 

[RodroX]

The conclution should said: for best game performance at 1080p (since thats the only resolution covered, I wonder why...), the i7 9700K still shown a small advantage when overclocked over the Ryzen 7 3800X on those 9 games tested (2 of which are really old).

But I still don't get why someone will spend a huge amount of many to pair the i7 9700k or R7 3800X and the RTX 2080TI to only play at 1080p, makes no sense at all.
I also don't think anyone will waste money on the R7 3800X when the R7 3700X can provide amost the same performance for less money (of coruse as long as the diference is worth it).

Why recommend a core i7 9700K (8c/8t) when intel is to launching in april 2020 a Core i7 with 8 cores and 16 threads that will require a new motherboard?, thats not a smart choice, not when the Ryzen 3800X has a new generation CPU coming later this year that will work on the same socket.

Personal opinion been a former intel user and willing to go back when is worth it: I think the whole intel desktop line (the current one) is completely broken and not worth the attention of anyone smart enough. Not unless intel does a big cut on the price, and not when there will be new CPUs (that will require a new motherboard) coming in 2 months). Of course there will be no big cut in price till the new CPU launch, and so anyone buying this old core i5, i7 and i9 today probably wont be very happy later on (in two months time)

This is not fanaticism, just the facts, Are some intel cpu still a bit better than AMD in some games?, yes they are. Is it worth/wise/smart to buy a dead platform (intel) with limited thread counts when a new CPU is coming in two months? No, is not.

Cheers

 

[jimmysmitty]

The conclution should said: for best game performance at 1080p (since thats the only resolution covered, I wonder why...), the i7 9700K still shown a small advantage when overclocked over the Ryzen 7 3800X on those 9 games tested (2 of which are really old).

But I still don't get why someone will spend a huge amount of many to pair the i7 9700k or R7 3800X and the RTX 2080TI to only play at 1080p, makes no sense at all.
I also don't think anyone will waste money on the R7 3800X when the R7 3700X can provide amost the same performance for less money (of coruse as long as the diference is worth it).

Why recommend a core i7 9700K (8c/8t) when intel is to launching in april 2020 a Core i7 with 8 cores and 16 threads that will require a new motherboard?, thats not a smart choice, not when the Ryzen 3800X has a new generation CPU coming later this year that will work on the same socket.

Personal opinion been a former intel user and willing to go back when is worth it: I think the whole intel desktop line (the current one) is completely broken and not worth the attention of anyone smart enough. Not unless intel does a big cut on the price, and not when there will be new CPUs (that will require a new motherboard) coming in 2 months). Of course there will be no big cut in price till the new CPU launch, and so anyone buying this old core i5, i7 and i9 today probably wont be very happy later on (in two months time)

This is not fanaticism, just the facts, Are some intel cpu still a bit better than AMD in some games?, yes they are. Is it worth/wise/smart to buy a dead platform (intel) with limited thread counts when a new CPU is coming in two months? No, is not.

Cheers

This comes up every single time there is any review of CPUs. The biggest reason to run the games at 1080p is to create a bottleneck at the CPU and not the GPU. If you are testing a CPU you want to see what the performance from the CPU being hit at 100% would be and not if the GPU is where the bottleneck exists. Pumping a game up to 4K max settings would be placing the bottleneck on the GPU and at that point we wont see any difference between most CPUs. This is also why most CPU reviews use the best motherboard and best GPU, to help keep the focal point on the CPU itself.

Now what does this show? From my experience it shows a CPUs performance and potential longevity. Normally CPUs that perform better now will still perform well a few years from now. The best proof I can think of is second generation Core i series. Still perform pretty well today. In fact most of the Core i series performs well today, the only downside is core count.

And the advantage is actually quite big. The 9700K is probably the best gaming CPU especially when overclocking, it has a pretty big lead when overclocking in most games.

I do agree though if there is a new platform coming out to not buy right now if the plan is to buy Intel. I would go a step further and say if the plan is to go with Intel to wait until they have something to be more competitive on mainstream and HEDT platforms as they are lagging and until 7nm is rolling I do not see anything Intel does as worth it. Keep in mind that I am holding off till we possibly have NVDIMMs like Optane as an option on mainstream so I am a tad weird.

 

[RodroX]

This comes up every single time there is any review of CPUs. The biggest reason to run the games at 1080p is to create a bottleneck at the CPU and not the GPU. If you are testing a CPU you want to see what the performance from the CPU being hit at 100% would be and not if the GPU is where the bottleneck exists. Pumping a game up to 4K max settings would be placing the bottleneck on the GPU and at that point we wont see any difference between most CPUs. This is also why most CPU reviews use the best motherboard and best GPU, to help keep the focal point on the CPU itself.

Now what does this show? From my experience it shows a CPUs performance and potential longevity. Normally CPUs that perform better now will still perform well a few years from now. The best proof I can think of is second generation Core i series. Still perform pretty well today. In fact most of the Core i series performs well today, the only downside is core count.

And the advantage is actually quite big. The 9700K is probably the best gaming CPU especially when overclocking, it has a pretty big lead when overclocking in most games.

I do agree though if there is a new platform coming out to not buy right now if the plan is to buy Intel. I would go a step further and say if the plan is to go with Intel to wait until they have something to be more competitive on mainstream and HEDT platforms as they are lagging and until 7nm is rolling I do not see anything Intel does as worth it. Keep in mind that I am holding off till we possibly have NVDIMMs like Optane as an option on mainstream so I am a tad weird.

Sorry if I was not clear enough, I understand why reviewers test CPUs at 1080p highest settings with the fastest GPU available at the time, but the point still remains, Who will spend that huge amount of money to play games at 1080p, specially games that aren't esport titles?

So to tell the whole story I think the inclusion of atleast 1440p results would give a better look at to what cpu is worth it or not.

And I don't agree a CPUs with fewer resourses in 2020 will do better than others with more in the future, is simple enough to look at the results of R5 1600 vs Core i5 7600 in current games to realise the Ryzen 5 1600 was the better deal back then. Not to mention you can now upgrade on the same motherboard if you had buyed the AMD chip back then.

 

[Phaaze88]

Wow. [Insert surprised Pikachu face]
I really shouldn't be surprised at how long this went on, but I am.

Here's the bottom line:
Money is no object gaming performance: 9700K & 9900K - or the K(eep) S(pending) version
Better cost per frame gaming performance: Ryzen 3700X & 3600

The review doesn't run nearly enough gaming samples; the overall difference between the 3800X and 9700K is much smaller than that - like less than 5%.
I can't blame them for not doing that - it's a lot of work after all, but if people aren't aware of that, it can be very misleading.

Get a 9700K or 9900K combo for over 30% more than the cost of a Ryzen 3700X combo for a meager single digit better performance...
Me: 🤣🤣🤣

 

[Darkbreeze]

Therein lies the crux of the issue. Assuming we test with a cooler of purportedly equal rating, then we go down the rabbit hole of using multiple different coolers for test pools of 18 processors. Different coolers for each price band? Then there will be endless complaints of which coolers are chosen, the fact that we're using different coolers for different price bands thus making all results incomparable, and claims of bias and handicapping. Better to remove cooling from the equation if you can and provide one centralized cooling solution that provides full performance for all chips. This is one of the foundational principles of comparing things: Create a level playing field, and let the silicon compete on its own merits. That's complicated by the fact that Intel doesn't provide coolers. Again, this is not perfect, but I'm not aware of a perfect answer. I'm all ears if you have an idea, but you have to think from the viewpoint of someone that opposes your opinion, as no matter what we do, someone will oppose based on various factors. The counter-arguments can be made from either side in convincing fashion, as there are both positive and negative aspects to any number of possible approaches.

At times we will do drill-downs on the impact of cooling, and we do present test results with most high-end processors with multiple coolers throughout a subset of workloads for all the most recent articles to indicate the merits of stepping up to a better cooler. Also, if a stock cooler causes significant performance reductions, we'll drill down hard on that, like we did here: https://www.tomshardware.com/reviews/intel-core-i7-8700-cpu-review,5638-2.html

I agree with this. So long as everybody is using the same cooler, and it is a cooler that is capable of not being the limiting factor, it shouldn't matter. Same cooler is fine for any comparison so long as it is a cooler model capable of doing the job. Otherwise, there is way too much door opening for crying and whining about selection.

 

[Darkbreeze]

Get a 9700K or 9900K combo for over 30% more than the cost of a Ryzen 3700X combo for a meager single digit better performance...
Me: 🤣🤣🤣

That probably won't even make a difference anyhow because you're likely already past your minimum/1% FPS threshold anyhow with the Ryzen SKU.

 

[Nick_C]

Therein lies the crux of the issue. Assuming we test with a cooler of purportedly equal rating, then we go down the rabbit hole of using multiple different coolers for test pools of 18 processors. Different coolers for each price band?

Probably - in the linked test the cheapest processor costs about the same as the cooler used - which, I would suggest, is not "normal" in terms of how a buyer allocates a limited budget between components.

Arguably the cheapest cooler that is able to meet the cooling requirements of the CPU ought to be used for each - even if this leads to needing to determine the best cooler for each and every CPU.

Better to remove cooling from the equation if you can and provide one centralized cooling solution that provides full performance for all chips. This is one of the foundational principles of comparing things: Create a level playing field, and let the silicon compete on its own merits.

Then be clear to add the cost of the cooler to the price of the CPU when comparing cost and performance differentials.

While using the same cooler for all CPUs levels the playing field in the test environment, that doesn't translate to examination of real world relative performance - unless the same cooler is used there too. Using a cooler that is "too good" can also hide disparities in power consumption / heat output between CPUs.

That's complicated by the fact that Intel doesn't provide coolers.

Which Intel does for its own reasons - one might suspect that it allows poor test results to be blamed on "using the wrong cooler" and also that it reduces the immediately visible price of ownership as the end-user needs to buy a cooler separately from the CPU.

Those CPUs supplied with a stock cooler should be tested using it as well - to allow the reader to compare performance between the CPU using the "free" cooler and the common cooler used in the test and make the value proposition of both clearer.

 

[Gurg]

Those CPUs supplied with a stock cooler should be tested using it as well - to allow the reader to compare performance between the CPU using the "free" cooler and the common cooler used in the test and make the value proposition of both clearer.

Reviewers use open test bench systems for CPUs and GPU to make it easier to swap components and eliminate the impact of heat trapped in case enclosures. I haven't had an air cooler hanging on my motherboard blowing hot air around inside my PC case for ten years.

That 10 year time period includes 5 different CPUs that I overclocked. I even did the "mod" on GPUs to allow those to be overclocked with attachment of my old 120mm water cooler to remove GPU heat from inside my case.

Personally I overclock to get the max out of the systems I build and the PC investments I make, generally replacing CPUs roughly every two year and GPU during the intervening years. You don't buy a new AIO closed water cooler for every CPU change so how do you allocate its cost over multiple different CPU purchases and and years of usage?

AMD only includes the cheap air cooler (that it probably gets in bulk for $5-10 each ) with its CPU for gullible people. From the comments above AMDs strategy works well.

 

[King_V]

Yeah, it would be a lot to start having to deal with multiple coolers. As long as the articles note "this will be fine with a stock cooler" and "that NEEDS a bigger cooler, or AIO/liquid/etc" it's sufficient I think, along with noting thermals in general. It would be mayhem to deal with a lot of varying coolers - just another variable that would make it NOT be "like vs like."

AMD only includes the cheap air cooler (that it probably gets in bulk for $5-10 each ) with its CPU for gullible people. From the comments above AMDs strategy works well.

Asserts facts not in evidence.

 

[Redneck5439]

Reviewers use open test bench systems for CPUs and GPU to make it easier to swap components and eliminate the impact of heat trapped in case enclosures. I haven't had an air cooler hanging on my motherboard blowing hot air around inside my PC case for ten years.

That 10 year time period includes 5 different CPUs that I overclocked. I even did the "mod" on GPUs to allow those to be overclocked with attachment of my old 120mm water cooler to remove GPU heat from inside my case.

Personally I overclock to get the max out of the systems I build and the PC investments I make, generally replacing CPUs roughly every two year and GPU during the intervening years. You don't buy a new AIO closed water cooler for every CPU change so how do you allocate its cost over multiple different CPU purchases and and years of usage?

AMD only includes the cheap air cooler (that it probably gets in bulk for $5-10 each ) with its CPU for gullible people. From the comments above AMDs strategy works well.

The Prism is the best factory cooler ever provided with a CPU. The cooler isn't for "gullible people" and it isn't just a gimmick. It is a capable cooler for stock operation. While it may lack the cooling potential of the big heat sink Noctua coolers (of which I upgraded to) or an AIO it does do a fine job at stock speeds and is what I would consider a good mid-range cooler. I have even seen some light overclocking done with the Prism (PBO / AOC). AMD even went as far as to include some surprisingly good RGB elements to the Prism. It is not a "cheap" cooler like Intel bundles with some of its processors. Short and long is don't confuse the Prism with the cheap $5 crap coolers Intel bundles.

View: https://www.youtube.com/watch?v=0JRY0Ri2VoM


Do you get AIO or high end Noctua like performance, no- but you do get very good performance considering the cooler is free and really will be fine at stock speeds. AMD stock coolers are in a league of their own for stock cooling. For anyone who wants to overclock I personally highly recommend the Noctua NH-D15 or NHU-14S (or of course the highly impressive Assassin III). Short of going to a custom loop these air coolers can provide all performance of an AIO yet last much longer and are more reliable as they don't have pumps that can fail.

 

[Gurg]

The Prism is the best factory cooler ever provided with a CPU. The cooler isn't for "gullible people" and it isn't just a gimmick. It is a capable cooler for stock operation. While it may lack the cooling potential of the big heat sink Noctua coolers (of which I upgraded to) or an AIO it does do a fine job at stock speeds and is what I would consider a good mid-range cooler. I have even seen some light overclocking done with the Prism (PBO / AOC). AMD even went as far as to include some surprisingly good RGB elements to the Prism. It is not a "cheap" cooler like Intel bundles with some of its processors. Short and long is don't confuse the Prism with the cheap $5 crap coolers Intel bundles.

View: https://www.youtube.com/watch?v=0JRY0Ri2VoM


Do you get AIO or high end Noctua like performance, no- but you do get very good performance considering the cooler is free and really will be fine at stock speeds. AMD stock coolers are in a league of their own for stock cooling. For anyone who wants to overclock I personally highly recommend the Noctua NH-D15 or NHU-14S (or of course the highly impressive Assassin III). Short of going to a custom loop these air coolers can provide all performance of an AIO yet last much longer and are more reliable as they don't have pumps that can fail.

Personally I would not use a Prism Wraith or any other cooler that holds back and throttles my CPU from achieving common peak performance. If the CPU is running even close to throttling temps at expected OC rates, I would replace the cooler with a better one. Evidently I'm not alone as there are pages of owners trying to dump these new AMD Prism Wraith coolers on E-bay before throwing them in the trash.

 

[RodroX]

Personally I would not use a Prism Wraith or any other cooler that holds back and throttles my CPU from achieving common peak performance. If the CPU is running even close to throttling temps at expected OC rates, I would replace the cooler with a better one. Evidently I'm not alone as there are pages of owners trying to dump these new AMD Prism Wraith coolers on E-bay before throwing them in the trash.

I would also be hesitant to try to save a few dollars by buying a new x470 M/B rather than a x570 M/B for a top AMD CPU knowing that a bios flash/update may be necessary to get the promised CPU performance with the new motherboard. If I already owned the x470 I would be willing to take a chance and then just buy a x570 if the bios flash went squirrelly. If you don't perform the update you can be leaving 3+% performance on the table.

The Prism cooler doesn't hold back the Ryzen CPU at stock settings, PBO and AOC are not stock settings. Stock setting is using PB only.

Also with my very cheap B450 mobo, after a BIOS update was getting around the same results or even higher on all the benchamrks I tested like most well known reviews out there.
In fact some reviewers including GamerNexus and HardwareUnboxed (among others, can't remember if Tomshardware found something similar) found out that some B450 mobos gave the same or better results than most X570 mobos, even some very high end models (using exactly the same components, only swapping the mobo).

Heck the very basic wraith stealth cooler was very decent for my CPU. Since day one with the stealth I got 4192MHz on single core and ~3918MHz on all cores. I installed a new one mainly cause of the pitch noise, some fans (even expensive ones) at the same dB level can have a more or less pleasant sound, and the stealth was annoying for me.

Yes if you wana enable PBO and/or AOC, then an aftermarket solution is the way to go. But those are not stock settings.

 

[King_V]

Personally I would not use a Prism Wraith

This shouldn't ever be a problem. Judging by the various posts that I recall seeing from you, I have doubts that you'd purchase any Ryzen.

 

[cryoburner]

I think you guys are reading too much into this article. These "VS" comparisons are always kind of silly and inaccurate. It's mainly just a way for them to recycle old content. You shouldn't expect much more from them.

About the only issue I would have is that one is not going to see those kinds of performance differences in today's games with most setups, simply because most people are not getting an enthusiast-level graphics card to run games at 1080p. It can make sense to test that way to show the potential upper bounds of performance differences in current games, but there should arguably be a disclaimer pointing out that most setups will tend to be more graphics-limited than anything, and under those conditions the processors will tend to perform rather similar.