seven months ago at CS, we had our first meetings with AMG, officially regarding there's into architecture running on seven centimeter Finn fete Well, and he promised us that it was gonna be worth the wait that in Q three of this year we would see amazing improvements over the first generations and architecture.
Well, did they deliver?
Corsair is proud to present their new hydro X line of custom water cooling products.
The new X revive.
Siri's radiators offer the perfect balance of Finn density and airflow to keep your loop cool and quiet.
The extra seven water blocks feature full coverage cooling for GPU while also maintaining an aesthetically pleasing design.
While the XY seven XY nine CP blocks keep All your modern CP is nice and cool To see the complete lineup of high directs cooling products from Corsair, click the link in the description below.
You know it's funny when it comes to seep you reviews.
There's lots of different ways you can do this.
The problem with CP reviews is there is a 1,000,000 different ways that you can use the CPU.
I mean, we're talking from basic Internet browsing email machines.
Remember the machines back in the day they called the machines.
It was super cheap and entry level because you could, like, go online and write email, right?
It didn't have to be powerful or anything like that.
But as we've moved forward, we've started to see the multi threaded Revolution, where programs are starting to use threads over just I, P.
C and core clock and all that sort of stuff.
And now we see a huge benefit to both multi threading and high core clocks.
Seeing five gigahertz now on si pues especially like the 9900 K is nothing special.
In fact, the fact that Intel even came out with the 9900 ks um, running a five gigahertz out of the boxes is just kind of dumb, in my opinion, because I've yet to see a 99 100 Kate, I couldn't run five gigahertz out of the box.
It takes more to really a sort of wow, I think, the mainstream audience and especially the enthusiast when it comes to see pews.
Like I said, there's there's live streaming, there's gaming, there's productivity, there's workflow.
There's photo shop.
There's editing that there's so many different things you could be doing with your machine.
It's very, very difficult to find any sort of testing methodology.
That's gonna sort of touch on all of that.
So what we did today is we focused specifically on this CPU.
It's gonna talk about a test bench with this because we have multiple motherboards that were sent for review, we're going to testing.
Not only the obviously the X 5 70 which is the latest PC I express for porno compliant, utilizes all the features built in to your CPU.
We tested with the G skill Royal tried at Ze Memory, and and all these great things were going to steal it back as we go forward in.
These reviews were gonna go and test not such high end hardware with BC pews to see how performance scales.
But for today we used the crosshair hero.
It's the next 5 70 We ran 16 gigabytes of trade and Z memory, the Royal Gold.
It's just really pretty, but it's still tried any memory.
And then we ran a 39 100 X and a 37 100 x versus kind of a weird lineup, I guess, are 9900 k and our 8700 case.
So Ninth Jen and Eighth Jen the 9900 k is an eight course 16 thread CPU compared to our 3900 X, which is a 12 Court 24 thread CPU.
And we did that for a couple of reasons.
They're more directly comparative in pricing.
If we wanted to do Corporal Core, we'd have to compare it to an H P D T processor, which is a high performance desktop, and do something like a 99 20 x, which is significantly more expensive than the $499 price point of the 3900 X.
Now, 3700 X is being compared to, well, again, the 1998 100 Kate and then 8700.
Really, the 8700 for us has just thrown out there one, because it's a six court 12 3rd processor, although this is an eight course 16 Fred Processor court for core, this more directly compares to the 9900 k.
But then, when you start comparing pricing, it doesn't really line up.
It's significantly cheaper because remember, the 3900 ex compares to the Intel's pricing point.
So what we're seeing here is Mork, or count at a lesser price.
Nothing new, though They did that with Zen one.
The 1818 100 X and 1800 Siris are the 1000 Siri's and then the 2000 Siri's.
We saw the same thing Maur course for less money.
But what we saw then was that I p C, although extremely close to Intel, was still slightly behind.
Now the 1900 k wasn't out yet then that was a seventh Gen processor and then the eighth gen processor in the ninth Gen processor.
So we saw slight I P C.
Improvements as time went from Intel, although not a lot, but still there It meant that the 1800 x in that 1000 Siri's n fell behind a little bit more night PC, but still gave us a massive improvement over FX, which is something that they had promised us and they delivered on that well, this time around, they're saying not only did we catch up, but we passed intel.
Now, when you look at the benchmarks is one thing to keep in mind Here we ran the memory at 2133.
That is based clock for DDR four.
Now I know a lot of you already are gonna be like What the heck, man?
You know, Zen runs better with fast memory, and that's because of the infinity fabric and all that sort of stuff built into the architecture with a M D.
But we tried to do here was give ourselves as fair of a fight as possible.
We're going to be addressing things like over clocking memory and over CLA Kings and two in a separate over clocking video to see how that performance skills, because one thing we saw with first Gen.
Or 1000 Siri's in Horizon was the faster the memory, the better.
Now the run, a whole new architecture.
I don't know exactly how that's gonna hold up, so we want to do some more testing with that.
And then we'll include that in her over clocking Guide for Zen, too, because over clocking your memory is always a part of over clocking your seat, Pierce.
We also ran everything out of the box with the exception of one test, you're going to see a sin.
A bench ar 15 results, not our 20 ar 15 because we've got previous CPU that have run it.
We can kind of see where it stacks up, but because we want to get single core performance to see how I peces have really held up and what we did was we locked both of the SI pues Well, I'm putting these to the 9900 k at four gigahertz versus the 3900 X at four gigahertz.
And yes, we're well aware that that's gimp ing the 9900 k.
But it's also giving this guy because it would run up to 4.6 gigahertz on single course.
So we just wanted an arbitrary number that locked them where we could get a comparison there on my PC.
And remember, whatever that performance differences as the clock speeding cruise increases, the eye PC difference and the score difference would be approximately the same percentage difference.
So if the 1939 100 exes beating it and then you speed it up in the clock, stay the same, the difference just it kind of does it whitens as it goes.
It's a percentage delta so that some percentages work the higher percentage.
Here the turns.
When it comes to CPU usage, you're going to see Obviously there's synthetic benchmarks in there like a geek bench and blender and sin adventures.
Stuff like that.
Those what they give us is a bit of control because it's extremely consistent between runs.
So what we're looking for here is not so much what the actual score was, but with the difference between the sea pews are because that gives us a at least some sort of a metric to compare.
So that's how you should be looking at that.
The real World World test gonna be things like the premier export or their warp stabilize.
And gaming is something we'll talk about in a separate video, because this video could be an hour and 45 minutes long.
Easy if we talked about everything a CPU could do and how they compare.
So we just kind of kept it.
A simple is possible here because what we really want to do here right now is we sort of wanna do a checks and balances when it comes to aim these claims on how the CPS perform versus um, you know, the real world experience and what we're getting because as we always preach, take the actual marketing material from any brand with a grain of salt because they're always going to present the best possible numbers because that's how marketing works.
In fact, gamers Nexus did a great piece showing why you would never use the slow setting on an O.
B s Lifestream or capture or whatever, because it's just unrealistic.
And the performance difference was really in proportionate to the quality difference.
In fact, in a blind taste test, as he showed, you could even tell the difference between fastest and slow versus the actual render quality.
It was nothing more than a test to show that the horsepower of the multi threading of the CPO name decide was was very, very good.
It was impractical to use that and say that the 1900 K was bad, so that's the kind of stuff we're looking for in these types of tests.
We also did, as you guys saw a single core I P C.
Using sentiments are 15th and spoiler alert.
The night 3900 X, which is the same architecture.
It's in the 3700 eggs and all that beat the 1900 K's I P.
So that is huge, because that is not something that is necessarily promised to us.
But it's something that we saw and in multi threaded performance, it makes no no surprise that the 3900 X is just running rings around the 1900 k because it has 24 threads versus 16.
So that is where the CPU reviews start.
To become difficult to kind of give us direct comparisons is comparing it to what?
So we chose price point and at the 400 $100 price point, give or take 20 or 20 or so bucks money on sales.
That's what the 1900 k costs.
And this was giving us better results in single thread as well as multi threat.
So when it comes to a premiere export, we used a 4.5 minute four k h 40.264 100 MBPs video that we did ah has, ah, full color grade on it, plus a small montage.
We use the built in YouTube 10 80 p preset.
Now you might be asking yourself, J, why did you go from four K to 10 80?
Well, as we're transporting footage like this and we're changing its skill and using maximum render, we're putting Max load on the sea pews.
We want to see how the si pues handle that sort of re scaling and trance coating so that we could see which did a better job of it.
If we went from four K to four K, there's a lot less load on the CPU.
We wanted the CPU to really be taxed in this test.
It's not a realistic thing for us to go from four K to 10 80 a lot of time to go from 10 84 k.
But that's besides the point.
But that is maximum render quality software encoding using the Mercury encoder, and we're the Mercury playback engine nog pew on either the MD or the Intel and no, I G.
Pew on the Intel whatsoever.
So this is just CPU course now.
The night of her, Kay took 22 minutes in 19 seconds, whereas the 3900 ex took 19 minutes and 56 seconds.
So it's kind of interesting to see that they were fairly close, although this had many more course.
A lot of that still comes to the fact that the premier software definitely favors core clock as well as multi threaded.
So it's one of those things that we're gonna look at as we over clock this in the future to see if we can get that number down and in the gap to be, you know, kind of reminiscent and linear with the core count of the SI pues.
But we know for a fact.
Premier loves core clock and four point a dish on all cores on the 900 k, which is kind of work goes to buy itself versus a just over four gigahertz on all course on the 3900 X.
So the 900 k was 800 ish megahertz faster but still lost to this running significantly slower but with more course.
So that's something we're talking about now.
If you've ever had to do any sort of warp stabilized fixing or processing inside of premiere, you know, depending on the length of the clip, it could be significantly taxing, not only to your system but time consuming.
So the faster that process could be done, the better it is for your real world scenario.
So the 9900 k performed that 12th 4 cage 40.264 100 MBPs video with no effects or any of that.
We saw two minutes and 22 seconds to process a 12th clip, and that's because warp stabilizer is more so a single threaded function, so it doesn't completely tax your system.
You could still do other background tests and other things while that's happening.
So this is more indicative of that single corer kind of an I.
Thing where the 3900 X performed it in two minutes in 15 seconds.
So although faster, once again with a slower clock speed, if we get the clock speeds up on the 3900 X, then we'll see an improvement to that score as well.
So something That's when the reasons why warp stabilizes is in this test, because it's kind of indicative again of that single threat performance.
Now let's go and talk about geek bench again.
They use a single core and then multi corps.
So the multi corps the 3900 X just obliterated the 9900 k 66 31,007 for the Intel I nine part vs the 3900 exits.
43,000 960 When it comes to the blender bm BMW tests the lower the number, the better 900 k again.
209 seconds versus 153 seconds for the MD part.
What about the I p C comparison directly?
Well, you guys saw on the charts there.
We had a better eye PC performance at four gigahertz on the 3900 x versus the 9900 case.
So that tells us the amount of work that see this CP is capable of doing the amount of instructions per clock is higher on this versus the intel part.
So that means that this whole time that Intel has just kind of been sitting back like yeah, yeah, OK, whatever.
An AMG has just kind of been spending the last few years just refining, refining, refining, which is what they've promised the community.
They've actually delivered on that promise.
Okay, but J if ABC is better on the 3900 X.
Why is the 3700 X, which is the same core count and thread count, as in any 100 K losing in many of the tests?
Well, that comes all the way down to clock speed.
So that is where AM D is.
Still kind of lagging behind Intel is the fact that Intel on again, we have not to over clock.
These have not done any over clock testing, so we still have to spend time in that to see how over clock abilities are.
Intel is still superior in the all core clock speed they're allowing.
I keep doing this to think it's Intel, but I'm in my mind.
I've got until the name be here.
Sorry until doesn't give us retail boxes.
We don't have one.
I could put a cardboard box here.
That's kind of what they send them to us in.
But actually, that's what they said.
The CPS to us, and they don't even care enough for the reviewers to actually give us pretty boxes.
So that's there.
Well, one could argue that the 3700 ex price point of $329 being on Lee, a handful of percentage points behind the 9900 case 400.
What, $84 is the best price we could find a foreign Amazon $484.99.
That is a $150 price difference between the two.
So I am the kind of shopper that would go well if I can get basically all of the performance of a 1900 cape as soon as I turn up the dial on this a little bit with over clocking, you know me and have 150 bucks to now put towards a better Jeep.
You or more ram for heck, more storage or pay for apart entirely.
Like my case.
That's the better by and I am confident that as soon as we start over clocking the CPS that we're going to see them catch an exceed the 999,900 case, fairly high strung as it is out of the box because it'll go for 849 all chorus.
But going above 51 is difficult.
You can get 5152 on water once you start chilling it, but day to day you're gonna be pushing enough voltage in that toe where you're gonna be degrading your C p over time.
And I would not trust running and then in 100 K at five point to every single day.
In fact, getting above five, too, was very difficult when we were doing our over clock.
Testing on that five gigahertz is easy.
51 is depending on your lottery and 52 and above is just so you don't have a lot of room to play with on a 900 k.
So if a 99 100 is already showing basically all of its potential and this is only a couple of percent behind it and more even multi threaded performance and we still have a lot of over collecting headroom into this are on this, then that makes this obviously the better by but what AMG has done here, it's delivered on the I.
P C promise it has delivered on the multi threaded promise.
And now that we know that things like games and your productivity programs are certainly leveraging multi threading our multi core availability, then that only makes this the better by now where the real story is on this and I really can't wait to get my hands on.
This are the Are five and our three models because they are three model coming later.
We know that the 3400 g in the 3200 years still going to using Vega part.
But you know as well as I do as soon as those parts get refreshed, probably later on this year.
And this is all hypothetical.
I have no information right here is that it's going to be utilizing the new radio and 5700 ext architectures or whatever they're calling it.
Navin is gonna be using the new Navid architectures and that making that a significant budget system when it comes to performance, you're gonna get amazing performance out of it without even had to buy a GP because that's exactly what the Vega part did for the 242,200 g.
But J, what if I've already got like a B 3 50 mother border next 3 70?
I mean, back when first risen came out well aimed, he made a promise that that was gonna be a five year socket.
The best sock.
It would be completely usable and valid for five years.
Intel likes to change the socket up just by a couple of pins.
Sometimes that you have to buy a new motherboard for a new C P.
Or the architecture demanded a different pin count, which means you need a different motherboard.
No backwards compatibility.
We saw a couple generations where you could use the new C P with an old one, but you lost various functionality.
The A M D Zen to architecture is backwards compatible all the way back to first Gen Risen.
So you could be using your X 3 70 or your B 3 50 motherboard, and you can still use the modern CP use.
And they will work just fine with nothing more than a BIOS update.
The only thing you're going to give up quite honestly, and I asked a nd this point blank in our meeting, what do you lose?
All you lose?
Is that PC for where did you know that PC I three is not being saturated by graphics cards yet, so we know it's not necessarily for peace before, Although the new radio and stuff That's a pizza before ready, unless you're using crazy, fast storage drives using PC I express.
That's probably a feature that you're not going to care about buying a new motherboard for.
So unless you care about any of that sort of stuff, you could take a new Zen to CPU and plop it in your old motherboard with the BIOS Update.
Update it first.
Before you get the new C.
You're not watching swap your CPU twice to do it.
It's completely backwards.
Compatible with motherboards.
AMG has promised US forward support with CPU socket I p C improvements and bringing the fight back to the consumer, the mainstream and enthusiast CPU market.
Well, I feel like a MD is delivered on all three of those.
This is simply a case of not being enough competition in the market and became complacent and didn't really feel the need to innovate.
Well, you don't innovate, you stagnate.
And that's exactly what's happened to Intel.
Although Intel is still a great performing CPU, it costs more and gives you nothing extra for it.
So this is where you guys sound off in the comments below did Intel fall behind or did a M d promise and overdeliver.
I kind of feel inclined to say that they kept all their promises, but sort of kept it quiet about how far they've really come so that we would all go Holy crap rather potentially over promise and under deliver.
This is kind of the opposite.
So you guys sound off in the comments below coming forward on this channel.
There's a lot of discussion to happen here.
We're gonna talk about backwards compatibility of those motherboards, which I just mentioned.
We're gonna do testing and compare it to these results to see if you truly lose anything.
We're also gonna be doing a complete over clocking guide for sent to, and we're also going to be doing I don't think I'll do any crazy custom cooling, but we're gonna be just testing obviously gaming performance and that sort of stuff because I know someone already typed it.
And if it's you Okay, fine.
You already typed it.
Yeah, but intel so much better at gaming.
Well, that was because Intel was superior in I p c.
So as long as we get the clock speeds up where we want them or even with current base clock speeds.