9900k or 3800x

Dan_D said:
You mean a heat sink that no enthusiast would probably use? A heat sink that would probably gimp your boost clocks and certainly limit the shit out of an all core overclock?
Click to expand...

Ahh and this is why we love you Dan, telling it how it is.
 
Dan_D said:
You mean a heat sink that no enthusiast would probably use? A heat sink that would probably gimp your boost clocks and certainly limit the shit out of an all core overclock?
Click to expand...
It was used for reviews of the 3700X in direct comparison to the 9900K. I am sure 3800X reviews will do the same.

Keljian when are you going to correct your "last gasp" faux pas or is misquoting how you roll?
 
ChadD said:
Haven't we all learned that clock speed is far from the determining factor in regards to performance. No this isn't P4 vs Athlon... still look at the benchmarks folks AMD is besting Intel in almost everything at lower clocks. I have no idea why people are so hung up on hitting 5ghz. Does your epeen grow 10 inches when you break 4.97 gigawatts I mean gigahertz or something ?.
Click to expand...

Humans have 5 fingers on each hand, and two hands, so our number system is base10 as a result.

5 (ghz) is a big deal to humans! :blackalien::alien::greyalien::panda::penguin:
 
Last edited:
I'd probably spend the extra money to get the 3900x, that said, if it's the 3800x vs 9900k, I'd without a doubt get the 9900k. I don't count future processor upgrades on a platform.

If I'm upgrading my processor I'm generally replacing the board. I generally buy mid to lower end boards.
 
Dan_D said:
It's not as expensive as the Maximus XI Extreme.



Yes. It's the one thing I don't like about it but the thing does at least clock RAM extremely well. Which is the point.
Click to expand...

Shame that, I have the IX with a 7700k, and only regrets is that I'm tied to 16gb of fast memory (tridentz 3600 cl15 binned ones)...
Nowadays 32gb is in some cases needed... :(
 
FedericoUY said:
I know, buy I enjoy OCing, that's mostly why :)
Click to expand...

Fair enough. Case in point though, my Threadripper 2920X is actually faster using DDR4 3000MHz modules with better timings than it is using the DDR4 3600MHz modules I used in the benchmark testing against the Ryzen 9 3900X. Those modules have much worse timings. I used them to achieve memory clock speed parity. It wouldn't have done much better against the new Ryzen, as that had little impact on most of the tests but it does test better with the slower clocked yet better latency modules.
 
Dan_D said:
Fair enough. Case in point though, my Threadripper 2920X is actually faster using DDR4 3000MHz modules with better timings than it is using the DDR4 3600MHz modules I used in the benchmark testing against the Ryzen 9 3900X. Those modules have much worse timings. I used them to achieve memory clock speed parity. It wouldn't have done much better against the new Ryzen, as that had little impact on most of the tests but it does test better with the slower clocked yet better latency modules.
Click to expand...
How are new ryzens doing with timings? 3600 @ ?
I've been rock solid on my 7700k with 3600 15-15-15 35 1t at 1,36vram. Totally pleased with it, but would need some more gigs.
 
FedericoUY said:
How are new ryzens doing with timings? 3600 @ ?
I've been rock solid on my 7700k with 3600 15-15-15 35 1t at 1,36vram. Totally pleased with it, but would need some more gigs.
Click to expand...

They'll run with tighter timings if that's what you are wondering. However, I don't yet know if they really benefit from tighter timings the way other Ryzen's seem to. That's something I've wanted to test and haven't yet.
 
Dan_D said:
They'll run with tighter timings if that's what you are wondering. However, I don't yet know if they really benefit from tighter timings the way other Ryzen's seem to. That's something I've wanted to test and haven't yet.
Click to expand...
Ok, good. Keep updated. Cheers!
 
go with the AMD because it's probably cheaper and you can put the rest of your money into a better GPU. This will probably give you the best peformance per dollar.
 
funkydmunky said:
It was used for reviews of the 3700X in direct comparison to the 9900K. I am sure 3800X reviews will do the same.

Keljian when are you going to correct your "last gasp" faux pas or is misquoting how you roll?
Click to expand...

I wouldn't have done that and didn't. I've got some upcoming CPU reviews and have zero intention of using the stock cooler for anything. I would rather show you what the CPU can do without the restrictions of some sad ass bundled CPU cooler. Even AMD says that PB2 and PBO reward users for going beyond minimum spec cooling via better boost clocks. AMD's example only included air coolers for reference.

upload_2019-7-16_9-38-24.png


Now, if we were doing a cooler round up, then yeah, the included CPU coolers would have a place showing you what you get for your upgrade.
 
Dan_D said:
They'll run with tighter timings if that's what you are wondering. However, I don't yet know if they really benefit from tighter timings the way other Ryzen's seem to. That's something I've wanted to test and haven't yet.
Click to expand...

This is something I'd really like to see rounded up. I feel heavily pressured to pay nearly double for a bdie kit than current prices for "other" 3600 kits up to c19.

Is the performance gap between 3600 c16 (or less) to c19 really going to mean anything in actual use?
 
Mode13 said:
This is something I'd really like to see rounded up. I feel heavily pressured to pay nearly double for a bdie kit than current prices for "other" 3600 kits up to c19.

Is the performance gap between 3600 c16 (or less) to c19 really going to mean anything in actual use?
Click to expand...

I am actually working on an article revolving around the topic of memory latency vs. clock speeds. I can't say definitively, but for the most part I believe the answer is no. While Ryzen or any other AMD architecture since the Athlon 64 has always seemed to benefit some from lower latency modules, I've never seen anything to indicate that the gains received would have an impact in any real world situation to a degree that you could perceive. Meaning, its something that certainly can show up in benchmarks, but it isn't going to translate to the user experience in any meaningful way. I still have lots of testing to do, but that's what I believe is the case based on my current data.
 
Dan_D said:
I am actually working on an article revolving around the topic of memory latency vs. clock speeds. I can't say definitively, but for the most part I believe the answer is no. While Ryzen or any other AMD architecture since the Athlon 64 has always seemed to benefit some from lower latency modules, I've never seen anything to indicate that the gains received would have an impact in any real world situation to a degree that you could perceive. Meaning, its something that certainly can show up in benchmarks, but it isn't going to translate to the user experience in any meaningful way. I still have lots of testing to do, but that's what I believe is the case based on my current data.
Click to expand...

I tend to agree from my own experience, looking forward to the article! You'll probably finish before 3700x ever comes back in stock for me to buy something anyway :p
 
Thanks for the tidbits about 3200/timings! (now I don't feel bad for the 3200 c16 I got in yesterday)
 
So Kajun614 what did you end up getting?


Dan_D Since you run the 2920xTR what are your thoughts on a refresh of those CPUs on 7nm? Right now at Microcenter I can buy a 2920x and board (Taichi, Auros pro, or Phantom Gaming) for less than a 3900x and board. 610 with the Taichi, 589 with Auros Pro, or 535 with the Phantom Gaming. Could be worth getting on that platform if we see some sweet new TR CPUs in a year or so.

Apologies for vearing of topic but TR4 has some deals man. Maybe not a gaming first build but damn.
 
I think the 2920 is extremely relevant since it's priced in the ballpark with the 9900k if you find a deal. Lots of people are back and forth between TR4 and AM4. If this is a workstation and gaming isn't your primary compute need, I would do the 2920 myself.. 10-20% less frames in games is a small price to pay for the benefits of HEDT. Figure out what the most time consuming / demanding use is for you and buy around that, chances are either system will get the job done unless you need either better single core OR quad channel ram / more pcie lanes.

For myself I'm probably going for a 3700x + aorus elite for my gaming rig and maybe x399 taichi and 1950x or 2950x when price comes down for my workstation.. I try to keep my work and play in separate rooms because I'm too prone to slacking. Only issue is TR will be good enough to run any game I want so I may slack anyway, have to go light on the GPU!
 
Call me crazy but I went with the 9900kf when it was marked down for a couple days. I was thinking about a 3900x or 3800x but they were not in stock anywhere a few days ago when I needed to pull the trigger. Plus watching reviews, Steve from Hardware Unboxed/ Techspot had his 3900x die on him and there was some other little quirks that had me apprehensive about going with Ryzen right at launch.
 
misterbobby said:
Call me crazy but I went with the 9900kf when it was marked down for a couple days. I was thinking about a 3900x or 3800x but they were not in stock anywhere a few days ago when I needed to pull the trigger. Plus watching reviews, Steve from Hardware Unboxed/ Techspot had his 3900x die on him and there was some other little quirks that had me apprehensive about going with Ryzen right at launch.
Click to expand...

As impressive as zen is, all this agesa and bios shit is inexcusable. You're right to be apprehensive.
 
Dan_D said:
You mean a heat sink that no enthusiast would probably use? A heat sink that would probably gimp your boost clocks and certainly limit the shit out of an all core overclock?
Click to expand...
Indeed, enthusiasts with lots of money.

But keep in mind the costs of some of those aftermarket coolers. And also the skewed perspective some enthusiasts have. Especially reviewers, who don't pay for a lot of the stuff they use. Hardware unboxed for example, figured the price of a Dark Rock Pro 4 into their cost analysis of a 9900k Vs. a 3700x with the stock cooler. He called the Dark Rock Pro4 a "decent" cooler (its actually one of the most premium air coolers) and also low balled the price a lot, at $75. Dark Rock Pro 4 are actually about $90 before tax and shipping. Which not only skews the actually cost/value of that cooler with a 9900k, but was also an overkill comparison Vs. the cooler included with a 3700x. He should have priced something in the $30-$40 range (if he still wanted a relative improvement over a "stock" cooler). There are a few great options there, which would be fine for a 9900k.

For $90 extra plus $10 more, you could go from a 3800x to a 3900x. That's a huge jump. Likewise with Intel, you need to be super dedicated to the highest framerates, to justify a 9900K with a $90 cooler, rather than 4 more cores in a 3900x with the stock cooler.

And that's the whole point of this talk about stock coolers. AMD gives you one, an actual decent one. Hell, the Wraith Spire fits into a Node 202 or RVZ02, if you take the top portion off the fan shroud. And considering that some of the better low profile coolers have been discontinued and are now extremely priced, if you can find them at all: The vapor chambered Wraith Spire ain't a bad value and is actually a really good value, for smaller cases. Similarly, the Wraith Stealth could be used in the absolute smallest cases. And smaller cases are absolutely enthusiast use scenarios.

Again, this may be irrelevant to people who use cases the size of suitcase and have money to spare. But these coolers absolutely have relevance.

Also, since AMD's has a pretty sharp upper limit on clocks, a 360mm water cooler only gives 100mhz more average on the 3900x in Blender (Hardware unboxed's test), than the stock cooler. and averaged 1fps more in games. The value really shows itself with the stock coolers.
 
Last edited:
chameleoneel said:
Indeed, enthusiasts with lots of money.


For $90 extra plus $10 more, you could go from a 3800x to a 3900x. That's a huge jump. Likewise with Intel, you need to be super dedicated to the highest framerates, to justify a 9900K with a $90 cooler, rather than 4 more cores in a 3900x with the stock cooler.

Again, this may be irrelevant to people who use cases the size of suitcase and have money to spare. But these coolers absolutely have relevance.
Click to expand...

While I never profess to be anything but an enthusiast, I have had the D15 since the 4790~ era about 4-5 years ago. It's not new by any means and has been on a 3770k, 5775c, Ryzen 1700, as well as the 9900k. I have a mid-tower case. Point is, I've not needed to change it in years.
 
Keljian said:
While I never profess to be anything but an enthusiast, I have had the D15 since the 4790~ era about 4-5 years ago. It's not new by any means and has been on a 3770k, 5775c, Ryzen 1700, as well as the 9900k. I have a mid-tower case. Point is, I've not needed to change it in years.
Click to expand...
Yeah a good cooler can last you many many years and through multiple builds. I used a Corsair A70 with my 2500k build and carried it over to my 4770k build. And I actually could still use the cooler again as it works perfectly but I decided with the 9900k build I just did to step up to something a little beefier (D15S) to cover me for the next couple of builds. And I guess if I decide to sell my previous PC with a 4770k locally then I will just leave that old cooler on there too. But no matter how you look at it getting 8 years out of a cooler sure was worth it.
 
chameleoneel said:
Indeed, enthusiasts with lots of money.

But keep in mind the costs of some of those aftermarket coolers. And also the skewed perspective some enthusiasts have. Especially reviewers, who don't pay for a lot of the stuff they use. Hardware unboxed for example, figured the price of a Dark Rock Pro 4 into their cost analysis of a 9900k Vs. a 3700x with the stock cooler. He called the Dark Rock Pro4 a "decent" cooler (its actually one of the most premium air coolers) and also low balled the price a lot, at $75. Dark Rock Pro 4 are actually about $90 before tax and shipping. Which not only skews the actually cost/value of that cooler with a 9900k, but was also an overkill comparison Vs. the cooler included with a 3700x. He should have priced something in the $30-$40 range (if he still wanted a relative improvement over a "stock" cooler). There are a few great options there, which would be fine for a 9900k.

For $90 extra plus $10 more, you could go from a 3800x to a 3900x. That's a huge jump. Likewise with Intel, you need to be super dedicated to the highest framerates, to justify a 9900K with a $90 cooler, rather than 4 more cores in a 3900x with the stock cooler.

And that's the whole point of this talk about stock coolers. AMD gives you one, an actual decent one. Hell, the Wraith Spire fits into a Node 202 or RVZ02, if you take the top portion off the fan shroud. And considering that some of the better low profile coolers have been discontinued and are now extremely priced, if you can find them at all: The vapor chambered Wraith Spire ain't a bad value and is actually a really good value, for smaller cases. Similarly, the Wraith Stealth could be used in the absolute smallest cases. And smaller cases are absolutely enthusiast use scenarios.

Again, this may be irrelevant to people who use cases the size of suitcase and have money to spare. But these coolers absolutely have relevance.

Also, since AMD's has a pretty sharp upper limit on clocks, a 360mm water cooler only gives 100mhz more average on the 3900x in Blender (Hardware unboxed's test), than the stock cooler. and averaged 1fps more in games. The value really shows itself with the stock coolers.
Click to expand...

I paid for most of the hardware I used for the entire time I worked for HardOCP. CPU's, RAM, graphics cards, PSU's etc. were always purchased by me. The only thing I got for free was motherboards. Occasionally I'd get PSU's or something else when Kyle would clean out his office but those were always things I received after my main machine was built. So that hardware always ended up in other machines. The point is, I always paid for my cooling. That's why I am not running a custom loop right now. If I could score one for free, I would. My second point, I've never seen anyone who can afford a 9900K and buy CPU's at that price point balk at a $90 cooler or even a $100-$150 AIO setup.
 
Dan_D said:
I paid for most of the hardware I used for the entire time I worked for HardOCP. CPU's, RAM, graphics cards, PSU's etc. were always purchased by me. The only thing I got for free was motherboards. Occasionally I'd get PSU's or something else when Kyle would clean out his office but those were always things I received after my main machine was built. So that hardware always ended up in other machines. The point is, I always paid for my cooling. That's why I am not running a custom loop right now. If I could score one for free, I would. My second point, I've never seen anyone who can afford a 9900K and buy CPU's at that price point balk at a $90 cooler or even a $100-$150 AIO setup.
Click to expand...
I see that point presented, pretty often. If you can spend X amount, you can spend Y more. While it is valid for some people, I don't think its a thoughtful point to make.

There are people all over this forum looking to maximize their dollar. People with budgets. For some people, $90 might be nothing. That's great for them. For many people, $90 is not trivial. And for anyone, $90 could easily be the difference to a higher product level. In the situation of a budget, If you can shift that $90 to a better piece of hardware, rather than a nice accessory which may not really be necessary; that's a smart move. or maybe just outright save the $90, because you don't actually need to spend it. Money saving moves like that are often looked for. I don't think you have to buy all the expensive stuff, to be an enthusiast. Anyone building or upgrading piece by piece is an enthusiast. A literal example would be that enthusiasts often buy the cheaper parts and tweak them to be better than stock. But sometimes that's only for the sake of tweaking. Because you might need other expensive gear, to be able to tweak. I mentioned earlier that you can modify the fan shroud on a Wraith Spire and it fits into one of the most popular, affordable SFF cases. The Node 202. That's an enthusiast move.

A 9900k (doesn't come with a cooler) + a $90 cooler, is the cost of a 3900x (comes with adequate cooler) which has 4 more cores. If you don't really need the 4 extra cores, you could save that $90 and get a 3800x. Or you could save even more and get a 3700x for virtually the same performance. And then really have some extra cash. Right this moment, AMD has the value lead. The reasons to get Intel, are to squeeze every last frame from games. Or certain productivity situations. Example, Adobe makes really good use of Quicksync in their suite. Certain Adobe tasks via Quicksync, are actually faster than the 12 core 3900x on its own. However, many people have dedicated GPUs which can do the same thing (assuming Adobe supports it). So that might be a moot point, in the end.

Ok fine, maybe some hardware reviewers are actually buying much of that hardware. You certainly were. I'll skip unpacking more on that, because I wasn't talking about you (I wasn't even specifically aware you were a reviewer for anyone). There are definitely many youtubers thanking companies for hardware, in their videos.

My point, with Hardware Unboxed as a specific example; is that some of these people are so used to having the best stuff on hand. or even all the stuff. best or not best. And its clear that some of them are either out of touch with what this stuff actually costs for people to own, or are hyper focused on selling the branding and lifestyle attributes of building with aforementioned stuff. Such that, they automatically go for the best thing and show that or use that. Something I liked about Hardocp's articles is that even though the name is "hard", they often pointed out value and the actual experience to the user.

Hardware Unboxed was way off on the price of the Dark Rock cooler and also verbally discarded it as "decent". Really pushing the water cooling. To his credit, he did highlight that with AMD, there isn't much benefit beyond potential noise reduction, to using a better cooler. Which is a very interesting and important point to highlight.

However, it was pretty tough to decipher what his thoughts were about the in between. If the stock cooler works, A Dark Rock 4 is just "decent", and the 360mm watercooler is admittedly very expensive: Where should we land? Some of these people are out of touch, I say. It overlaps into their communities. And I think it makes it really difficult for the average buyer to make sensible choices.
 
Last edited:
M76 said:
Really hard to even consider anything but 3900X in face of things like this:
Click to expand...

Those framerates on that 2.0GHz Xeon!

I get what you are saying with respect to pure number crunching per socket, but you're also not comparing a commercial CPU with commercial motherboards / platforms when using the 3900X. Which would actually be fine if you weren't also comparing price... :)
 
1080p is still the most popular primary res on the June 2019 Steam survey by a huge margin. I realize that’s just one data point but it’s a gaming platform so probably fairly relevant.

Now enthusiast gamers have moved to higher res yes. I’m one of them. BUT 1080p/60 is still a price/performance target. And of course CSGO bros at 240hz etc.
 
Keljian said:
In the face of a synthetic benchmark?
Click to expand...
I know it has become trendy to dump on synthetic benchmarks but they are still a far superior indicator of computational power than any game.
Besides I haven't played anything in 1080p since 2014. So excuse me if I give zero craps about 1080 benchmarks on a 2080ti. That's like using a sports car as a school bus.

With games it is always dependent on how they are optimized and are they gpu or cpu limited, and where are the differences come from. Because it is certainly irrelevant to me that the 9900K gets better average fps if the 1% low numbers are the same. So these game benchmark averages are completely misleading. You won't get a better gaming experience because your top 1% frames are better. The only scenario where I'd base my CPU purchase on game benchmarks is where I know that I'd only use the PC exclusively for gaming.

But for rendering, and video editing, every bit actually counts.
 
Don't disagree with any of that M76, in fact, I'll second the 1% low numbers point. I still want to see average framerates too, but if 1% lows are acceptable then I'm happy.

And I agree, if 1% lows are borked, the part is written off in my book. We don't have to settle for that anymore and we absolutely should call out any part / configuration / software / etc. that produces poor 1% low results.
 
You must log in or register to reply here.
Back
Top