News New Third-Gen Ryzen CPUs Listed by ECC: Ryzen 9 3900, Ryzen 7 3700, Ryzen 5 3500

The Ryzen 5 3500, on the other hand, could be the successor to the Ryzen 5 2500X. If so, the processor would only be available to OEMs. The Ryzen 5 3500's specifications remain a mystery as the EEC listing only exposed the chip's 65W TDP. Since even the lowest Ryzen 3000-series part has six cores and 12 threads, the Ryzen 5 3500 could arrive with the same core and thread count.
The Ryzen 1500X and 2500X were both 4-core, 8-thread parts, so logically I would expect the 3500 to be as well. I suspect it would be Zen 2 based, and lack integrated graphics, using a cut-down version of the same chiplet used for the 6 and 8-core parts. While the 2500X might have been a bit redundant for consumers due to the 2400G offering relatively similar performance along with integrated graphics, and the 2600 not costing much more, a 2500X could stand out over the current Zen+ based APUs due to the larger performance improvements brought by Zen 2. So, I could see them potentially making it available to consumers as well, perhaps somewhere around a $150 price point.

The 3700 would be a welcome addition as well, since there's currently a relatively large price rift between the Ryzen 3600 and 3700X.
 

InvalidError

Titan
Moderator
The Ryzen 1500X and 2500X were both 4-core, 8-thread parts, so logically I would expect the 3500 to be as well.
Nah, there is one other way that AMD could differentiate a 3500 from the 3600: L3 cache. AMD has no SKU with reduced cache to funnel all chiplets with one or more defective L3 slices to. A 6C12T chiplet with 16MB of L3 instead of 32MB would make plenty of sense.
 
Reactions: Rdslw

hannibal

Distinguished
Apr 1, 2004
2,409
70
19,890
14
Interesting... then there will be 3 different 8 core prosessors... how close 3700 will be 3700x... the 3600 is almost the same as 3600x so it can be the real budget 8 core King! But it can make 3800x to look even worse, so have to see how low in binning can those really will be.
The 3500 is the odd bird in here. Four core or otherwice cut down version?
All in all if it is four core it we be super usefull in budget builds! Not for modern Gaming, but older games would run very well indeed when overclocked and below 3600 price point with same ips advancements. It can be really good indeed...
if it has that suggested cache cut, harder to say. I still Expect less cores but who knows.
 
Reactions: keith12

InvalidError

Titan
Moderator
But it can make 3800x to look even worse, so have to see how low in binning can those really will be.
Much like how the original 1800X also made very little practical sense. Seems like AMD wants to make the *800X the model it wishes to sell but the market has no reason to want unless there is a shortage of *700(X) parts.
 
I find that the 3700 at 65W wouldn't make much sense. With the 1700 & 2700 those were 65W versions compared to the 1700X & 2700X which are 95/105W versions. In this case the 3700X is already 65W so making a 3700 a 65W chip doesn't make any sense. If AMD were to make the 3700 a 35W or 45W CPU then they could have a chip that would make sense.
 
how difficult is it to OC these new cpu's? I have heard that you basically can't. Is that true? I keep seeing people say OC the 3600 for it to match the 3600X or is that just off the top of your head?

i am curious, not that I would OC anything, the last time I tried a few decades ago i fried the mobo... lol, never tried since, I just buy faster hardware and don't worry about it.
 

InvalidError

Titan
Moderator
how difficult is it to OC these new cpu's? I have heard that you basically can't. Is that true? I keep seeing people say OC the 3600 for it to match the 3600X or is that just off the top of your head?
AMD's Power Boost is already clocking 3rd-gen chips pretty close to the maximum possible without sub-ambient cooling, leaving next to nothing to gain from manually overclocking to the point that siliconlottery (a site that specializes in aftermarket binning of chips for overclocking potential) has admitted that their relevance is nearly over.

Most reviewers who have compared the 3600 with the 3600X have found that both already perform about the same (within 3% in most cases) so paying $50 (25%) extra for the 3600X makes very little sense. AMD should drop the 3600X's price to $20 above the 3600 mainly for the upgraded HSF and call it a day.
 
Nah, there is one other way that AMD could differentiate a 3500 from the 3600: L3 cache. AMD has no SKU with reduced cache to funnel all chiplets with one or more defective L3 slices to. A 6C12T chiplet with 16MB of L3 instead of 32MB would make plenty of sense.
a bin for failed chips that are almost as good previous gen R7 and cost half ? A lof of people will buy this with a smile. AMD will bump the yeld per wafer utilizing what would go to trash otherwise, people will get cheap cpu that is still better than intel i5....
 
Reactions: salgado18
AMD's Power Boost is already clocking 3rd-gen chips pretty close to the maximum possible without sub-ambient cooling, leaving next to nothing to gain from manually overclocking to the point that siliconlottery (a site that specializes in aftermarket binning of chips for overclocking potential) has admitted that their relevance is nearly over.

Most reviewers who have compared the 3600 with the 3600X have found that both already perform about the same (within 3% in most cases) so paying $50 (25%) extra for the 3600X makes very little sense. AMD should drop the 3600X's price to $20 above the 3600 mainly for the upgraded HSF and call it a day.
It behaves better than OC-ed as it clocks down when idling, saving the thermal headroom for time you need it. Unless you have overkill cooling, this is more desirable than OC.
About 3600 vs 3600X its just cooler (I agree a pricey one for $50), If you plan on using aftermarket one, those 2 chips are almost identical.
 

salgado18

Distinguished
Feb 12, 2007
625
57
19,070
4
Nah, there is one other way that AMD could differentiate a 3500 from the 3600: L3 cache. AMD has no SKU with reduced cache to funnel all chiplets with one or more defective L3 slices to. A 6C12T chiplet with 16MB of L3 instead of 32MB would make plenty of sense.
You mean like the Athlon II and Phenom II? I believe using dies with 4 Zen 2 cores makes more sense. Lower L3 cache, especially when it is already big, won't give much performance impact, and wouldn't lower prices too much.

About 3600 vs 3600X its just cooler (I agree a pricey one for $50), If you plan on using aftermarket one, those 2 chips are almost identical.
From some tests I've seen, even when exchanging coolers they behave almost identically, and $50 can get you a way better cooler. The 3600X doesn't have much reason to exist, except to fill a supply issue (in Brazil only the 3600X is available, all others are basically sold out).
 

InvalidError

Titan
Moderator
I believe using dies with 4 Zen 2 cores makes more sense. Lower L3 cache, especially when it is already big, won't give much performance impact, and wouldn't lower prices too much.
The 32MB of L3 cache accounts for about 2/3 of Zen 2 chiplets' die area and there are no SKUs using chipets with less than 32MB of L3 per chiplet. Since you are roughly twice as likely to get defective cache than defective cores, it would make sense to have at least one SKU to spend all the chiplets with some defective L3 on. The alternative is throwing all of those otherwise perfectlyfine dies in the garbage can.

Making marketable products out of what would otherwise be very expensive garbage can enable considerably cheaper parts and I bet AMD has many more chiplets with some bad L3 than chiplets with more than one bad core per CCX.
 
Much like how the original 1800X also made very little practical sense. Seems like AMD wants to make the *800X the model it wishes to sell but the market has no reason to want unless there is a shortage of *700(X) parts.

I have been hoping AMD does a thin bin like Intel did to make the 9900K and makes a higher end 8-core above the 3800x. I would step up into that if It was doing 4 to 4.1Ghz base and say 4.8Ghz boost.
 

InvalidError

Titan
Moderator
I have been hoping AMD does a thin bin like Intel did to make the 9900K and makes a higher end 8-core above the 3800x. I would step up into that if It was doing 4 to 4.1Ghz base and say 4.8Ghz boost.
If AMD had the yields to do that, that's what the 3800X which currently makes next to no sense over the 3700X and even the plain 3700 would have been.
 

hannibal

Distinguished
Apr 1, 2004
2,409
70
19,890
14
Best cores goes to Roma prosessors, but it seems that the quality of the chips is reasonable stable so the is not too big variance to play with. When amd does get new steppings out maybe early next year before 4000 series those may run a Little bit higher, but I don`t personally Expect to see much higher clockspeeds untill the next production node upgrade. Next gen versions on 7nm production node could improve the performance but that most likely goes to 4000 and 5000 series untill we see those.
 
Best cores goes to Roma prosessors, but it seems that the quality of the chips is reasonable stable so the is not too big variance to play with. When amd does get new steppings out maybe early next year before 4000 series those may run a Little bit higher, but I don`t personally Expect to see much higher clockspeeds untill the next production node upgrade. Next gen versions on 7nm production node could improve the performance but that most likely goes to 4000 and 5000 series untill we see those.
Didn't know the CPU core was a type of tomato.
 

InvalidError

Titan
Moderator
Next gen versions on 7nm production node could improve the performance but that most likely goes to 4000 and 5000 series untill we see those.
I'm not expecting significantly higher clocks with 4th-gen. Why? Because I'm expecting AMD to use most of the transistor speed improvements to cram more logic in the critical path instead of increasing clock frequencies the same way it did from 2nd to 3rd. People were expecting 500-800Mhz clock gains from 12nm to 7nm, we only got 200-300Mhz extra because AMD used over half of the process speed gains on making paths and various architecture structures bigger/wider to increase IPC. While AMD may not increase cache sizes again this time around, adding more execution ports to the back-end, increasing the register file size, increasing cache bandwidth and fattening up the scheduler some more to better support SMT3/4 would be high on my suspect list.
 
I find that the 3700 at 65W wouldn't make much sense. With the 1700 & 2700 those were 65W versions compared to the 1700X & 2700X which are 95/105W versions. In this case the 3700X is already 65W so making a 3700 a 65W chip doesn't make any sense. If AMD were to make the 3700 a 35W or 45W CPU then they could have a chip that would make sense.
TDPs listed for processors typically don't correspond to the exact power that they draw under load. Both the Ryzen 3600 and 3700X currently have a "65 watt" TDP, for example, despite the 3600 having two less active cores and in turn drawing less power. The same goes for the 8-core 3800X, 12-core 3900X, and 16-core 3950X, which all share the same "105 watt" TDP. It's possible they could give the 3700 something like a "45 watt" TDP, but I suspect they are not going to reduce the clocks enough to where it would be drawing less power than a 3600, so a "65 watt" TDP seems more likely, not that it matters much.

Making marketable products out of what would otherwise be very expensive garbage can enable considerably cheaper parts and I bet AMD has many more chiplets with some bad L3 than chiplets with more than one bad core per CCX.
Yep, they could definitely cut the cache for a 3500, though I suspect they would cut the processor down to a quad-core as well. So far, the last two generations of Ryzen "500" parts have both had 4-cores with 8-threads, and I don't see why they would change that now, especially since they already have two 6-core processors, however similar they might be.

As for 4th-gen Ryzen, my best guess is that it will be a lot like 2nd-gen, offering slightly more performance than the current 3000-series processors, but probably not a whole lot more on the feature front. That could take the form of slightly higher clock rates due to a matured process, along with small architectural improvements. Maybe they'll move the IO chip to 7nm if doing so is cost-effective by that point. One possible major change I could see would be for them to move their APUs to use the same chiplets as the rest of their lineup, along with a separate graphics chiplet. That could allow them to have processors with integrated graphics and up to 8-cores. And maybe they could even use the same graphics chiplets in their next generation of dedicated cards.
 

InvalidError

Titan
Moderator
Yep, they could definitely cut the cache for a 3500, though I suspect they would cut the processor down to a quad-core as well.
And what would the price point be?

With previous generations, you had APUs using one monolithic die and mid-range using another monolithic die, placing them on roughly equal footing in terms of manufacturing cost. With Zen 2, mainstream starts with a significant cost handicap due to far more complex multi-die package. Being a quad-core without IGP, it cannot cost more than the 3400G and being multi-chip, it cannot cost much less either. A very awkward place to be.

As a hex-cores with halved cache though, it could easily slot somewhere between the 3400G and 3600 at something like $175, a far more profitable place for AMD to be at.
 
Being a quad-core without IGP, it cannot cost more than the 3400G and being multi-chip, it cannot cost much less either. A very awkward place to be.
Not really. It still has one major thing going for it. The 3400G is still based on the 12nm Zen+ architecture, while a quad-core 3500 based on Zen 2 would be around 15% faster at similar clocks. For a system with dedicated graphics, such as a lower-end gaming system, one would be better off with the Zen 2 part. I would expect such a processor to be priced similar to the 3400G, trading the integrated graphics for more performance per core in systems where the IGP wouldn't be getting used anyway. Again, the 4-core, 8-thread 2500X exists, along with the 4-core, 4-thread 2300X, both slightly faster than their 2400G/2200G counterparts, but without integrated graphics.

And just as those processors are only sold in bulk to OEMs, the 3500 might be as well. But I could also see it having a place on the consumer market, due to the new 7nm architecture bringing greater performance gains than we saw with the previous generation. And its difficult to say exactly how the manufacturing cost of these new processors compares. It might be more expensive, but I doubt it's prohibitively so. And while a $150 quad-core might not seem as profitable as a $175 six-core processor, such a processor would likely cut into sales of the 3600 more than anything.
 

IceMyth

Honorable
Dec 15, 2015
571
3
11,165
97
I hope this will be true because I want to get the 12 core CPU (Upgrade from Ryzen 1700). The current 3900X is not available anywhere, and if it does, it does not last long. I hope the 3900 will have the same base clock as the 3900X while the boost clock not below 4.4Ghz.
 

InvalidError

Titan
Moderator
Again, the 4-core, 8-thread 2500X exists, along with the 4-core, 4-thread 2300X, both slightly faster than their 2400G/2200G counterparts, but without integrated graphics.
The 2500X costs about the same make as the 2400G - similar process, similar die size, simple single-die package. A Zen 2 3500 would be significantly more expensive to make than the 3400G due to requiring a far more complex package substrate (almost double the layer count) and two chips. AMD is aiming to increase its gross profit to 45% and you can't increase gross profit by selling parts that are more expensive to manufacture for the same price. So, a quad-core 3500 would need to either cost $10-20 more than a 3400G for AMD to maintain gross profit or be a cut-down 3400G (no IGP, higher CPU clocks) instead.
 

rigg42

Commendable
Oct 17, 2018
639
233
1,390
36
Most reviewers who have compared the 3600 with the 3600X have found that both already perform about the same (within 3% in most cases) so paying $50 (25%) extra for the 3600X makes very little sense. AMD should drop the 3600X's price to $20 above the 3600 mainly for the upgraded HSF and call it a day.
The 3600x is more pointless than the 3800x. Since the auto oc feature doesn't really do anything for any CPU but the 3600, turning it on at 200 mhz makes the 2 CPU's basically identical. The 3600 will hold at 4275-4300 under your typical game load with a decent cooler and 200 mhz auto OC enabled. It gets the same cinebench single thread score as the 3600x and 3700x as well when you do this.
 

ASK THE COMMUNITY

TRENDING THREADS