AMD Ryzen 7 1800X (14nm Zen)

[Tunnah]

Because the problem is even 10% here or there makes a big difference - look at card launches,etc where the same thing plagues AMD. Like I said that SMT bug/optimisation issue cost them 10% in the Hardware.fr review - which is big. Its the difference in the Hardware.fr review of the R7 1800X matching a Core i7 4790K or Core i7 5820K and being within 10% of a Core i7 6900K to the latter being 20% faster and the R7 1800X be the same as a Core i7 3770K in performance.

So if people see AMD matching a high clocked Haswell part,its far more positive than a SB/IB CPU from years ago - you are starting to see people laughing that 6 year old Intel CPUs are doing well against it.

AMD is just giving more chances to Intel in a growing segment.

The whole issue,is that its enough for say an Intel CPU to get 80FPS on a £1000 card and an AMD one to get 70FPS for it to look fail,especially since it will be an older Intel CPU getting that,since realistically we have not seen a big change in single core IPC over 5 years.

Remember once you go over £200+ you are entering the enthusiast area where people will be comparing percentages.

Also I doubt not having proper windows support is a small problem - it means proper scheduling support too,or even proper power control of the CPUs under Windows,and its why some people mentioned performance was improved using a high performance profile(might be not true).

You need to realise Intel CPUs have had years of support as they are gradual improvements - Ryzen is a new core,and AMD achieves nothing by launching it earlier and making it look worse than it should.

AMD always does this - the R9 290X due to its crap cooler,made it loose performance,and look hot and noisy against Nvidia cards and often the drivers are not quite there. The same as the RX480 which meant Nvidia probably got some extra sales. Intel and Nvidia might have some bugs during their launches but as a whole seem to just be better at doing smoother ones.

Toms Hardware said the same thing:



The problem is this enforces the whole Intel/Nvidia duality against them.

Yeah I get what you're saying and it's a valid grievance, but I genuinely don't believe it'll hurt em. They needed to get this product out, they needed to show they can pee all over Intel chips at half the price. These issues will be ironed out but honestly this was the best way. Instead of releasing a patched CPU in several months to make everyone happy, the vast majority of folks who don't care can get 1 now, and have a performance increase when it is sorted.

They need to have the products out for them to be patched anyway, so the problems can be found by a much wider audience than the QC department.

EDIT: Also the enthusiast market isn't as clear cut as that. Personally if I hadn't just bought a 2700K, I'd be getting one of these, simply because they're offering such a hefty performance for half the price of the Intel model. Hell, depending on the speed of the 4c models, if they can boost higher, I might still get one. It is the clock speed which is keeping me on the 2700K, if the other parts come in at same speed, it'll be performing better at IPC due to it being a newer chip, and I want a motherboard with more up to date features, so it'll be a no brainer for me.

The reason I didn't get a 6700K was because I was looking at £600 just to boost the CPU and get M2. But if the 4c Ryzen is any good, it'll no doubt be around £200, the motherboards tend to be cheaper, and...OK I'll still have to buy DDR4 no getting away from that. But I'll be looking at £400 vs £600 for better performance. And even if the 6700K is better slightly, hopefully it'll be a low enough difference as to not make a real world dent

[Tabbykatze]

They need to have the products out for them to be patched anyway, so the problems can be found by a much wider audience than the QC department.

This, is why this launch happened. It wasn't rushed (to a degree), they needed to get it out in the wild so that these problems can be found. QA can only take you so far. I'm a Sophos Architect and the amount of products they release with a plethora of interesting bugs and hindrances makes me turn my head. I'm forever creating tickets from my installs and feeding information to the appropriate persons on what is wrong and where.

But that doesn't diminish the power and value of the product, technology is so unbelievably complex these days that even with exhaustive regression testing and QA that there are vast areas that just can't be caught in time for release when you have shareholders and people with their own personal moneys riding on what you do. Sometimes you've just got to do it and fix it when people find these issues to find out where to focus on.

Using Sophos again as an example, I was the first XG Architect in the UK and v15 of the new Firewall appliance is cutely labelled "as the child we don't talk about". It was bloody awful to work with but it was still powerful and only by it being out in the wild did the majority of the issues get ironed out within a financial year and we got v16 which was very good. Now it wasn't the best but v17 is almost ready to drop and only by 1,000s of installs and people using it will it become something that will make me go "the older SG which has an established 10 years behind it is no longer the appliance of choice".

[CAT-THE-FIFTH]

Yeah I get what you're saying and it's a valid grievance, but I genuinely don't believe it'll hurt em. They needed to get this product out, they needed to show they can pee all over Intel chips at half the price. These issues will be ironed out but honestly this was the best way. Instead of releasing a patched CPU in several months to make everyone happy, the vast majority of folks who don't care can get 1 now, and have a performance increase when it is sorted.

They need to have the products out for them to be patched anyway, so the problems can be found by a much wider audience than the QC department.

Thats the thing - people who want a change,or find a niche that these CPUs will excel in will buy them. But what about everyone else - the RX480 started strong and then Nvidia launched the GTX1060 with better cooling,etc before AMD could and stole some of their thunder.

Its just annoying to know so many gaming benchmarks out there are probably a good 5% to 15% lower than they should off just because someone at AMD forgot to test their CPUs in games with no SMT on.

Now you have this response from AMD saying they need better optimisation by devs for their CPU in games:
https://www.pcper.com/news/Processor...ng-tests-Ryzen


What they don't say is why they released their CPUs in an unsupported state under windows or ignore the fact they knew very well SMT was leading to reductions in gaming performance.

How could they just not be honest and say "please test games with SMT off,as some games might have performance oddities with SMT enabled due to unfinished support under windows CURRENTLY".

I just cannot understand what the heck they were thinking??

AMD apparently does not want to launch any of its CPUs or GPUs in the last 5 years with the best possible performance.

Compare that to Nvidia and to a slightly lesser degree Intel.

[gordon861]

So it is looking (at the moment) that as expected the AMD chips will be close to the Intel chips but at a lower price. But is that lower price going to be enough?

At the moment I can buy an i5-7600k for £220 from Scan, the AMD 'equal' will be cheaper but I doubt it'll be half the price, may 2/3 the price so about £150. Is that £70 enough for me to jump from Intel to AMD for gaming?

I don't think so, based on my existing i5-2500k that I have had since they were launched and is still going strong. From past experience I doubt that the AMD chip will still be competitive for as many years as the Intel, so spend the bit extra now and it will probably end up costing me £20 a year for four years.

I will watch how things develop over the next few months before pulling the trigger on a new PC but I think it's still gonna be an Intel.

[CAT-THE-FIFTH]

This, is why this launch happened. It wasn't rushed (to a degree), they needed to get it out in the wild so that these problems can be found. QA can only take you so far. I'm a Sophos Architect and the amount of products they release with a plethora of interesting bugs and hindrances makes me turn my head. I'm forever creating tickets from my installs and feeding information to the appropriate persons on what is wrong and where.

But that doesn't diminish the power and value of the product, technology is so unbelievably complex these days that even with exhaustive regression testing and QA that there are vast areas that just can't be caught in time for release when you have shareholders and people with their own personal moneys riding on what you do. Sometimes you've just got to do it and fix it when people find these issues to find out where to focus on.

Using Sophos again as an example, I was the first XG Architect in the UK and v15 of the new Firewall appliance is cutely labelled "as the child we don't talk about". It was bloody awful to work with but it was still powerful and only by it being out in the wild did the majority of the issues get ironed out within a financial year and we got v16 which was very good. Now it wasn't the best but v17 is almost ready to drop and only by 1,000s of installs and people using it will it become something that will make me go "the older SG which has an established 10 years behind it is no longer the appliance of choice".

That is not the point - the SMT performance regressions in games were found within a few days of testing by reviewers. Surely AMD knew about this??

How long does it take AMD to run a few games with SMT on and off to just verify its working fine??

Seriously??



Its shoddy testing - its not one or two games which are obscure. That is 8 well known games there including BF1 and W3 which are popular titles.This is not like the TLB which was not so evident.

What type of testing were they doing in their labs??

Do AMD want to sell their products or find new ways for Intel to sell less Ryzen CPUs for them??

Edit!!

That only leads to the perception AMD is for tweakers and for everybody else just get Intel/Nvidia as its less of a "problem".

We as enthusiasts might wait,but the problem is AMD by not launching their products in a polished way,is making them look worse,and hence they need to cut prices to sell their products.

Why do you think Nvidia has traditionally charged more for their cards and got away with it even if they are no better??

AMD screws up launches 9/10 in some way that their competition can breathe easier.

Intel is probably slightly less worried now,since AMD just threw them a bone.

[Tabbykatze]

That is not the point - the SMT performance regressions in games were found within a few days of testing by reviewers. Surely AMD knew about this??

How long does it take AMD to run a few games with SMT on and off to just verify its working fine??

Seriously??



Its shoddy testing - its not one or two games which are obscure. That is 8 well known games there including BF1 and W3 which are popular titles.

This is not like the TLB which was not so evident.

Don't get me wrong, I absolutely agree with you. But according to that legitreviews review of the 1800x, in the week of testing they were being fed new updates and points and things to do. Now, if AMD continues that trend then they can quite quickly ramp it up.

Now I don't know the relationship that AMD has with Microsoft regarding OS optimisation but it obviously looks like Intel/Nvidia have a better relationship in getting this stuff out. The other issue is that this is a completely new architecture and a behemoth as large as Windows needs massively exhaustive testing before they release anything in this area. On top of this, Intel's architecture hasn't changed much relatively in the past 5 years.

So you have a mature supported architecture vs. an immature architecture that was only finalised at the start of Q1 2017. So I'm reserving my spear judgment of crucifying AMD because some of this is nothing to do with AMD's responsibility.

[Tunnah]

So it is looking (at the moment) that as expected the AMD chips will be close to the Intel chips but at a lower price. But is that lower price going to be enough?

At the moment I can buy an i5-7600k for £220 from Scan, the AMD 'equal' will be cheaper but I doubt it'll be half the price, may 2/3 the price so about £150. Is that £70 enough for me to jump from Intel to AMD for gaming?

I don't think so, based on my existing i5-2500k that I have had since they were launched and is still going strong. From past experience I doubt that the AMD chip will still be competitive for as many years as the Intel, so spend the bit extra now and it will probably end up costing me £20 a year for four years.

I will watch how things develop over the next few months before pulling the trigger on a new PC but I think it's still gonna be an Intel.

Ya but there are a LOT of folks out there who are going to see a 4c/8t part costing considerably less than a 4c/4t part from Intel and it'll be a no brainer. Sure you're not going to fork out to upgrade to a part that doesn't give you much performance, but there's a HELL of a lot of people out there with Q6600s, or even i3s that aren't cutting it for games.

[Tunnah]

That is not the point - the SMT performance regressions in games were found within a few days of testing by reviewers. Surely AMD knew about this??

How long does it take AMD to run a few games with SMT on and off to just verify its working fine??

Seriously??


Its shoddy testing - its not one or two games which are obscure. That is 8 well known games there including BF1 and W3 which are popular titles.This is not like the TLB which was not so evident.

What type of testing were they doing in their labs??

Do AMD want to sell their products or find new ways for Intel to sell less Ryzen CPUs for them??

They probably didn't admit it on purpose, you don't want reviewers going into it with an issue at the forefront of their mind

Also they probably want to downplay it, make it out like sure they know about it, but it's no biggie and will be sorted out anyway.

[flearider]

as has been made clear by so many reviews this is not a gamers cpu .. yes it's just ok .. but there pushing it in a different way to those professionals out there that are hard struck by the down turn in the economy ..wanna get more done for less ££$$$ choose a ryzen .. imo it's a good play there hitting 2 segments at once ..
so it will game just fine @4-4.2ghz under water with 3200-3600 ddr4 .. now what I want to see is what the 1700x will do .. cause i'm not spending £100 more for very little oc ..

[CAT-THE-FIFTH]

Don't get me wrong, I absolutely agree with you. But according to that legitreviews review of the 1800x, in the week of testing they were being fed new updates and points and things to do. Now, if AMD continues that trend then they can quite quickly ramp it up.

Now I don't know the relationship that AMD has with Microsoft regarding OS optimisation but it obviously looks like Intel/Nvidia have a better relationship in getting this stuff out. The other issue is that this is a completely new architecture and a behemoth as large as Windows needs massively exhaustive testing before they release anything in this area. On top of this, Intel's architecture hasn't changed much relatively in the past 5 years.

So you have a mature supported architecture vs. an immature architecture that was only finalised at the start of Q1 2017. So I'm reserving my spear judgment of crucifying AMD because some of this is nothing to do with AMD's responsibility.

My main gripe is that with SMT off,Ryzen is faster in gaming,and AMD has essentially made its CPU looking terribad in many reviews,with a Core i7 3770K being close to it in some of them. This is doing damage to AMD in the gaming community,and first impressions count. This is a continuation of almost all their launches in the last 5 years - they really need to get on top of this,and have to be realistic and honest about things.

They can't afford these sort of issues,and like I said its a simple fix - they should have tested it and told reviewers that SMT might have some issues in gaming.

They probably didn't admit it on purpose, you don't want reviewers going into it with an issue at the forefront of their mind

Also they probably want to downplay it, make it out like sure they know about it, but it's no biggie and will be sorted out anyway.

But its not helping since instead of people thinking AMD is aware of it and trying to get support out,it looks like some "bug" AMD is not aware off discovered by reviewers.

[Tabbykatze]

My main gripe is that with SMT off,Ryzen is faster in gaming,and AMD has essentially made its CPU looking terribad in many reviews,with a Core i7 3770K being close to it. This is doing damage to AMD in the gaming community,and first impressions count. This is a continuation of almost all their launches in the last 5 years - they really need to get on top of this,and have to be realistic and honest about things.

They can't afford these sort of issues,and like I said its a simple fix - they should have tested it and told reviewers that SMT might have some issues in gaming.

True, but then you have to remember the i7 3770k is also close to a 7700k

If we use the cpu benchmark mega only showing desktop cpus then ordered by CPU Mark, there are 3rd generations rubbing shoulders with 7th generations. Now what does that read like to you?

AMD have an extremely good platform that's just been released, it does have kinks and SMT optimisation is not a "deal breaking" kink, far from it. Some of these things are quite odd, but issues like this aren't unexpected from something so new.

I want to see what the first 30 days give, I want to see a review in 30 days time and I want to see what changes. These reviews have almost made me push the button but I'm going to sleep on it right now. i've been AMD for 10 years and I want to replace my 8350 and Ryzen is what I want.

Edit: I also don't believe you understand the gravity of how a processor architecture is designed. "Resolving SMT is simple fix"? AMD have created SMT, it is up to Microsoft to optimise how it uses it in it's abstraction layer

[CAT-THE-FIFTH]

True, but then you have to remember the i7 3770k is also close to a 7700k

If we use the cpu benchmark mega only showing desktop cpus then ordered by CPU Mark, there are 3rd generations rubbing shoulders with 7th generations. Now what does that read like to you?

AMD have an extremely good platform that's just been released, it does have kinks and SMT optimisation is not a "deal breaking" kink, far from it. Some of these things are quite odd, but issues like this aren't unexpected from something so new.

I want to see what the first 30 days give, I want to see a review in 30 days time and I want to see what changes. These reviews have almost made me push the button but I'm going to sleep on it right now. i've been AMD for 10 years and I want to replace my 8350 and Ryzen is what I want.

Edit: I also don't believe you understand the gravity of how a processor architecture is designed. "Resolving SMT is simple fix"? AMD have created SMT, it is up to Microsoft to optimise how it uses it in it's abstraction layer

I think you misread what I said by a simple fix - the simple fix is for AMD to have told reviewers to test games with both SMT enabled and disabled. There is NOTHING complex about putting that in their reviewers guides.

The simple fix would be for AMD to be honest and say MS was having some delays in getting drivers and scheduler updates out. The simple fix would be them saying games needed updates to handle the AMD implementation of SMT.

They need to be honest about these problems - not hide them.

They know very well hiding problems has blown up in their face and they don't have the sway of Intel or Nvidia to bury them.

It shows you either AMD has not done proper internal testing or has and was trying to hide it to save face.

Saving face is irrelevant when your CPU is scoring upto 15% lower scores in games due to SMT issues.

If you look at it objectively why would you want to get a Ryzen 7 1700 or 1700X as a gamer if you just looked at many of the reviews??

We are enthusiasts,so we can kind of think its plausible for AMD to get another 10% maybe 15% out of the CPU once devs start getting to grips with optimisation,or by simply switching off SMT.

Sadly,most reviews won't mention any of that.

[watercooled]

One thing I was left asking myself after reading the review was did it live up to the hype and honestly I think it missed in a few areas.
- At least 10 Watts higher at idle than any of the Intel chips tested

Apart from the equivalent socket 2011 processors you mean?

- The PiFast result for single thread was down on what I was hoping for

Does anyone care?

- Warhammer showed better fps with SMT off (as we saw with AMD's CMT, getting anyone to support a particular uarch is not always easy so we can't just assume there will be a game patch to fix it)

This is one to watch I agree, but I don't think it's comparable to the issues AMD had with CMT support - SMT is fairly well-implemented in Windows now. Intel had problems pretty much the same as this a few years ago, so it could just be a case of Windows Update providing an update to the scheduler to properly recognise Ryzen's capabilities.

- Memory latency was horrible (hopefully something that new BIOS versions can fix and I fully understand its early doors on that)

As a synthetic benchmark it's... pretty meaningless TBH. According to the Hexus benchmark they're beating Intel across the board in terms of bandwidth with 2 channels of memory, and they quite clearly stated the unoptimised nature of the benchmark. This also seems to be what people were referring to with, like you say, last-minute BIOS patches yesterday, so I wonder if that had anything to do with it?

Surely it's not just following the herd to buy Intel it's the logical thing to do.

As others have said that very much depends what you're doing.

• Established product
• Less driver issues
• Established hardware platforms
• Know what you are going to get

Because Intel never stumbles on releases with product recalls, deteriorating chipsets, awful GPU drivers, etc? You don't 'know what you are going to get' with Intel any more than anyone else TBH - choosing them is no guarantee you won't have problems.

• And for the overwhelming majority of users (inc. gamers) they are faster.

Where's the logic on taking a punt on a brand new product that's probably going to have many teething problems unless it's much faster or massively cheaper ?

Well... they are massively cheaper? And much faster in many workloads too. As above, Intel have as many teething problems as anyone else, that's not sensible logic.

[Tabbykatze]

I think you misread what I said by a simple fix - the simple fix is for AMD to have told reviewers to test games with both SMT enabled and disabled. There is NOTHING complex about putting that in their reviewers guides.

The simple fix would be for AMD to be honest and say MS was having some delays in getting drivers and scheduler updates out.

They need to be honest about these problems - not hide them.

They know very well hiding problems has blown up in their face and they don't have the sway of Intel or Nvidia to bury them.

It shows you either AMD has not done proper internal testing or has and was trying to hide it to save face.

Saving face is irrelevant when your CPU is scoring upto 15% lower scores in games due to SMT issues.

Ah, I see what you're getting at.

But I do agree with what someone else said in this thread, why tell upfront when all you're going to do is cause undue panic. Or the flip side of the coin, tell people and people are "nice" in their reviews. Or just let it carry on as normal.

[CAT-THE-FIFTH]

Ah, I see what you're getting at.

But I do agree with what someone else said in this thread, why tell upfront when all you're going to do is cause undue panic. Or the flip side of the coin, tell people and people are "nice" in their reviews. Or just let it carry on as normal.

But the problem is like hiding a dirty family secret its different when you get on top of it,instead of somebody "finding it out".

It looks like review sites have found a "bug" in Ryzen regarding SMT and gaming and many sites are not aware of the SMT issues in games,so are showing lower than normal performance in games.

OTH,if they actually see the tree for the woods,and get ahead of the curve on it they could manage it. Plus so many people pre-ordered Ryzen for gaming,expecting it would reasonably competitive even if it did not beat Intel,but the issue is in certain reviews(who seem to be oblivious to the issue),they tested with SMT on meaning performance could regress to IB levels in certain reviews.

IB levels at the worst end are still perfectly fine but,the issue it sounds like AMD Ryzen is more like a 5 to 6 year old than a newer one and people will subconsciously link the two.

Edit!!

The Hexus review is fairly solid as they are seasoned enough to test whether SMT was having an issue but lots of reviewers nowadays don't really bother checking these sorts of things too.

[Tabbykatze]

But the problem is like hiding a dirty family secret its different when you get on top of it,instead of somebody "finding it out".

Well, I think that's taking it a bit far but I somewhat agree.

The Hexus review is the best I've seen thus far. What I do find interesting is the fact that AMD falls behind in 1080p but locksteps at higher resolutions.