• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Ryzen benchmarking and overclocking results

"One thing I did notice is that all the games I have looked at so far -- which is considerably more than the four shown here -- were smooth on the Ryzen processors. GTA 5 for example plays really well on the Core i7-7700K, but every now and then a small stutter can be noticed, while the 1800X runs as smooth as silk, sans stuttering from what I observed

Thread scheduling works more efficient is my guess (16 threads vs 8). Hence less switching latency.
 
1080p's where it's at, imho, i can't afford anything higher resolution, still on a 2011 era TN panel

says the man with a X99/GTX1080 system
 
It is starting to sound like it is a software issue causing lousy gaming performance specifically with SMT. With it off games go quite a bit higher (At least some of the ones I have seen do). It at least closes the gap a bit which is a good thing to note, however it means that SMT is not yet optimized well. Not going to be to the end all fix, but sounds like there are definitely some patches needed as even some sites are reporting that updating motherboard bios's is also improving performance by 5%.

Hope they iron things out in a timely manner!
 
Last edited:
for a new architecture launch these chips are performing better than expected.

i think the noticed issues at gaming(and not only) will disappear once the soft's will be patched to use correctly the cpu architecture&instructions

well done AMD and hope future cpu prices will never reach 1000$ ! (not server ones ofc)
 
for a new architecture launch these chips are performing better than expected.

i think the noticed issues at gaming(and not only) will disappear once the soft's will be patched to use correctly the cpu architecture&instructions

well done AMD and hope future cpu prices will never reach 1000$ ! (not server ones ofc)
Yes, there's a good chance the gaming charts will show 5~15% more fps, at a minimum, when SMT quirks are patched over.
The multitude of MT tests show that AMD's SMT yields better returns than SMT even on KL. I'm hoping this is ironed out soon enough, looking at you MS o_O
 
My little Ryzen review.

I have provided a handful of 3dmark Firestrike runs as well each run was done once with the GPU at stock and once with the GPU overclocked. All GPU overclocked runs had the GPU boosting to 2138mhz on the core and ram at 2250. Maximum load temp on the GPU was 45C, maximum power consumed was 370w at the wall for the full system.

firestrike-stock-gpu-ryzen.png


firestrike-oc-gpu-ryzen.png


firestrike-extreme-stock-gpu-ryzen.png


firestrike-extreme-oc-gpu-ryzen.png


firestrike-ultra-stock-gpu-ryzen.png


firestrike-ultra-oc-gpu-ryzen.png


My overall impression of the platform is wait a few weeks for the kinks to be pushed out. Right now it is not ready for retail sales in my opinion. Performance however was excellent. All of the chips regardless of model seem to clock the same, so save a couple bucks and grab the R7 1700 and use that saved money for a good cooler.

Also big thanks to chew* on XS as per usual he was able to walk me through some difficulties I had with the new product.

Thanks for this man. As I expected, Ryzen will change the landscape of 3dmark benching allowing a lot more ppl to get those kinds of physics scores for a hell of a lot less money.
 
So, despite the outcome, it still seems more of a bang for buck than the marginal performance difference... but since most of the reviews are using the same Asus board, the results were somewhat "incoherent". I smell the need of "AMD dual-core optimizer" all over again like the old Athlon X2 era for some sort of optimization in certain games and apps. But on the other hand, It's still a good value buy than going the Intel route, unless Intel significantly lower the price points of its CPU line-up..
 
i kind of regret getting the 1800X over the 1700 - looks like its out of stock now tho... might see if i can pick one up retail and send the 1800x back.
 
It is starting to sound like it is a software issue causing lousy gaming performance specifically with SMT. With it off games go quite a bit higher (At least some of the ones I have seen do). It at least closes the gap a bit which is a good thing to note, however it means that SMT is not yet optimized well. Not going to be to the end all fix, but sounds like there are definitely some patches needed as even some sites are reporting that updating motherboard bios's is also improving performance by 5%.

Hope they iron things out in a timely manner!
Yes, there's a good chance the gaming charts will show 5~15% more fps, at a minimum, when SMT quirks are patched over.
The multitude of MT tests show that AMD's SMT yields better returns than SMT even on KL. I'm hoping this is ironed out soon enough, looking at you MS o_O

And fixing/disabling HPET.
 
one of the reviews has a graph with smt on/off for all cpus in the test. intel's ht is affected the same way, only usually to a slightly lesser degree.
 
It seems to do IPC alright in certain things. It scores higher from a singlethreaded IPC perspective in the CPU-Z benchmark at 4Ghz than a Skylake CPU does.
Looking at it, very good IPC in this case, i can only think it's faster than a Oced 7700K.
 
These are brand new chips that were just released. Intel has been tweaking the same design for years. Some bugs are to be expected, as with any brand new tech.
 
I am disappointed. Lousy gaming performance over Kabey Lake. I thought about upgrading to this but held off on preordering to wait for reviews. I think i'll wait for Canon Lake or maybe even x299.
Lousy? It's 5% behind a 7700k at .5ghz lower clock speed and nearly double in multi thread, I have no issues fps wise gaming with a 3570k as do others with 2500k etc but I could sure use an extra 8 threads and double performance in other tasks for the same price....
 
@hat even still my 4790k is very relaxed today, not like it has been pre ryzen release. It was worried about losing its job. Now it knows at the very least I'll have to keep it around for gaming, even if I do get a ryzen productivity rig.
 
Better than I expected. I think once the teething issues are corrected and the next revision comes out it will be a serious contender to the best Intel can make. Clearly needs more clockspeed, better mobo bios, turbo, and ram compatibility.
 
My little Ryzen review.

0301171640a-hdr.jpg


To start off I am overall very impressed with the product, however the thing that left me angry was the feeling that it was released before it was done. To start things off I had a retail sample board and CPU no ES markings just retail style, but the CPU and board arrived pre-used and updated to a newer BIOS than what ships to the retail market.

0227171412-hdr.jpg


0227171413-hdr.jpg


Couple of distinct things to notice here the new bracket for AM4 works just fine with the older AM3 coolers that use the normal mounting style. I went ahead and tested this with the older generation Wraith cooler.

0227171540-hdr.jpg


However it turns out the Asus is way ahead of that and has double drilled the board for AM3's older style. Just a heads up using the other hole set is a bit particular with coolers. The height is not the same so I ran into an issue when using EK's waterblock where if you tighten the fasteners more than one to two turns the system would not post. Be very careful when selecting coolers for these systems and if it doesn't post back the screws out.

0227172029-hdr-crop.jpg


Finished product yielded excellent temperatures. Under load the CPU did not exceed 61C with an overclock on both the CPU and GPU. This is using the standard EK 280 kit with a couple modifications, using the shorter reservoir to clear the card and downgraded the fans for LED Thermaltake rings. There is next to no noise produced and the BIOS is set to extreme silence.

0227172014-hdr-crop100.jpg



0301172149-hdr-crop.jpg


0301172149a-hdr-crop.jpg


Now onto the actual usage of the parts. I went ahead and stuck to the normal batch of benchmarks found on hwbot. Performance was all done at my maximum stable clock of 4075mhz@1.475v, ram@3200 18-17-17-29 1T. Maximum temperatures again were 61C during all of the testing and maximum power pulled at the wall for the whole system under CPU only testing was 260w.

cpuz-ryzen.png


x265-1080p-ryzen.png


x265-4k-ryzen.png


cb15-singlecore-ryzen.png


cb15-multicore-ryzen.png


wprime155-ryzen.png


superpi-ryzen.png


Getting this unit stable is what brings me to the "unfinished product" belief the BIOS was nothing short of a joke. I have played with some pre-release ECS products that had a more complete BIOS than this does. Half of the settings wouldn't save and the first board I had just up and died on me. This led to the swap to an actual retail board that wouldn't post until it was flashed using the Asus crash free utility. Currently the only useful BIOS I have found is 0601 and 0702 I wouldn't bother with any of the others.

It is a common issue to set something in the BIOS have it reboot and appear to be at that setting in the BIOS and end up in windows at stock settings. This is especially true with the ram. The ram on this is just a joke it is lacking all tertiary timing control and of the timings that were provided Cas Latency doesn't even work. If you set the ram speed expect to change it to one lower, reboot and then set what you actually wanted otherwise it will stick at 2133mhz.

After all of that nonsense I was lucky to know chew* who lovingly informed me of the memory hole that exists for 3200-3600mhz. That fixed a handful of stability problems I was having and led to my final choice of clocks (40.75x100). I had been testing the unit at 40.25x101 prior. I went as far as 362x for memory speed at 18-19-19-39 1T which is the rating of this kit, it would not stay stable in super pi 32m so it isn't worth really noting outside of one strange little thing. Remember when I said I was using a "retail sample" CPU? Well the actual retail CPU I grabbed which was a 3 week newer chip actually couldn't even POST with the ram at 36xx. That CPU was set aside and all testing was completed on the original sample provided.

I have provided a handful of 3dmark Firestrike runs as well each run was done once with the GPU at stock and once with the GPU overclocked. All GPU overclocked runs had the GPU boosting to 2138mhz on the core and ram at 2250. Maximum load temp on the GPU was 45C, maximum power consumed was 370w at the wall for the full system.

firestrike-stock-gpu-ryzen.png


firestrike-oc-gpu-ryzen.png


firestrike-extreme-stock-gpu-ryzen.png


firestrike-extreme-oc-gpu-ryzen.png


firestrike-ultra-stock-gpu-ryzen.png


firestrike-ultra-oc-gpu-ryzen.png


My overall impression of the platform is wait a few weeks for the kinks to be pushed out. Right now it is not ready for retail sales in my opinion. Performance however was excellent. All of the chips regardless of model seem to clock the same, so save a couple bucks and grab the R7 1700 and use that saved money for a good cooler.



Also big thanks to chew* on XS as per usual he was able to walk me through some difficulties I had with the new product.
Finally a TRUSTED Source review and info... @W1zzard you need to Hire this MAN!!!

Thanks @cdawall for all your help and info :)
 
Last edited:
It is starting to sound like it is a software issue causing lousy gaming performance specifically with SMT. With it off games go quite a bit higher (At least some of the ones I have seen do). It at least closes the gap a bit which is a good thing to note, however it means that SMT is not yet optimized well. Not going to be to the end all fix, but sounds like there are definitely some patches needed as even some sites are reporting that updating motherboard bios's is also improving performance by 5%.

Hope they iron things out in a timely manner!
wait... Doesn't intel chips display the exact same behavior? In some titles HT hinders performance??
 
wait... Doesn't intel chips display the exact same behavior? In some titles HT hinders performance??

Not anymore - in many/most modern games Intel HT either doesn't affect performance or even leads to a small plus.

https://www.computerbase.de/2017-02/cpu-skalierung-kerne-spiele-test/

For Intel on one side and devolpers on the other one there was enough time to iron out the HT issues.
AMD being new to HT on the other hand ... . Ofc a new platform with iffy BIOS, strange memory behavior and what not doesn't help.


edit:

And to those who think that the 6C/12T part will "really overclock" ... keep dreaming. Maybe in a year or two
 
Last edited:
AMD is new to HT hum?
 
Not anymore - in many/most modern games Intel HT either doesn't affect performance or even leads to a small plus.

https://www.computerbase.de/2017-02/cpu-skalierung-kerne-spiele-test/

For Intel on one side and devolpers on the other one there was enough time to iron out the HT issues.
AMD being new to HT on the other hand ... . Ofc a new platform with iffy BIOS, strange memory behavior and what not doesn't help.


edit:

And to those who think that the 6C/12T part will "really overclock" ... keep dreaming. Maybe in a year or two
It did as of mid 2016...

http://www.overclock.net/t/1588555/gaming-benchmarks-skylake-core-i7-hyperthreading-test

Also, not sure your link tested HT directly, unless I missed it in translation? Was just a cpu and total threads test, no? Did they disable HT in any of that testing?
 
Back
Top