EVGA GeForce GTX 1650 Super SC Ultra

[W1zzard]

EVGA's GeForce GTX 1650 Super SC Ultra is priced at NVIDIA MSRP, yet offers all the important extras. It comes with a dual-fan cooler that achieves excellent temperatures and has idle-fan-stop capability. A factory overclock is included, too, and EVGA even managed to squeeze in a metal backplate.

Show full review

 

[Deleted member 50521]

Would it be possible to have a hard line at 60FPS for the average FPS graph? Just way to emphasize it is the minimum playable framerate on monitors? Same can be done for 144Hz high end high refresh gaming?

Something like this

Then maybe take all the GPUs' average FPS metrics and put them through the 2 filter mentioned above, rate the GPUs either pass the minimum 60FPS or not. And for 144FPS as well.


I find percentage deceiving some time. 10% difference at 30FPS range is just 3FPS. While 10% difference at 60FPS range is 6FPS (14FPS difference for 144FPS gaming). Lot of people would take those percentage difference out of context when recommending GPUs which is really misleading.

 

[jabbadap]

Quite interesting that this uses out of spec 14Gbps memories. Is there some shortage of 12Gbps gddr6, or does manufacturers just slap card which ever memory they have on stock.

Either way a bit too loud for what it is. There's a typo on Fan noise page too:"With 38 dBA, the card is the second-noisiest GTX 1660 Super we tested so far, and it wouldn't have to be."

 

[bug]

SC used to be my go-to line, but it seems like these days SC Ultra only earns you an extra 30MHz (a whopping +1.7%) and a nice cooler.

Also, how is this card almost 8% better than the reference in perf/$ when it's priced the same and in the performance charts is barely 2% faster?

 

[Rowsol]

Hopefully you can get your hands on the gigabyte windforce. That is also priced at 160 and turns off the fans.

 

[Assimilator]

@W1zzard Did you feed back to EVGA about the overly-aggressive fan profile? If they can fix that with a new BIOS, no reason why they shouldn't.

 

[nguyen]

wow, 15.8% overclocking headroom is quite awesome, probably a golden sample.

 

[W1zzard]

hard line

I looked into those lines and it makes the charts way too busy, and scary for novice readers. The Y-scale is created such that there's always a line at 60 fps.

typo

fixed

@W1zzard Did you feed back to EVGA about the overly-aggressive fan profile? If they can fix that with a new BIOS, no reason why they shouldn't.

Of course.. for the GTX 1660 Super already. They said they'd look into it, I offered to wait with testing until a new BIOS, they said don't wait. That was for 1660S, and no change with this 1650S either..

 

[jabbadap]

wow, 15.8% overclocking headroom is quite awesome, probably a golden sample.

Well that memory makes a lot of that from stock 12Gbps*128/8 = 192 GB/s to 4*2*1.960GHz*128/8 = 250.88 GB/s. Just makes me wonder if every card have 14Gbps gddr6, or is it just some early patch.

 

[Assimilator]

Quite interesting that this uses out of spec 14Gbps memories. Is there some shortage of 12Gbps gddr6, or does manufacturers just slap card which ever memory they have on stock.

My guess is that the memory manufacturers have stopped producing 12Gbps chips in favour of the 14Gbps ones, as it simplifies logistics and inventory. (This is the same reason NVIDIA axed TU117 in favour of using die-harvested TU116.) The fact that all the 1650 Super cards reviewed so far use 14Gbps memory supports this. But NVIDIA can't allow 1650S cards to ship as 14Gbps because then the only difference between those cards and the 1660S would be the number of CUDA cores enabled on the GPU (1280 vs 1408) and the amount of memory and bandwidth (4GB vs 6GB/192GB/s vs 336GB/s), which would cannibalise 1660S sales. So, 1650S cards get a BIOS limiting their memory to 12Gbps despite the fact that they can all do 14Gbps no problem, in order to keep the marketing segmentation.

The end result is that the 1650S is probably the bang-for-buck card in this market segment, assuming you overclock the memory to get the ~10% free extra performance that nets you. Unless the mass-market RX 5500 is significantly faster than the OEM version W1zz reviewed, or comes in at $140 or less, it's going to have a hard fight against the 1650S.

 

[jabbadap]

My guess is that the memory manufacturers have stopped producing 12Gbps chips in favour of the 14Gbps ones, as it simplifies logistics and inventory. (This is the same reason NVIDIA axed TU117 in favour of using die-harvested TU116.) The fact that all the 1650 Super cards reviewed so far use 14Gbps memory supports this. But NVIDIA can't allow 1650S cards to ship as 14Gbps because then the only difference between those cards and the 1660S would be the number of CUDA cores enabled on the GPU (1280 vs 1408) and the amount of memory and bandwidth (4GB vs 6GB/192GB/s vs 336GB/s), which would cannibalise 1660S sales. So, 1650S cards get a BIOS limiting their memory to 12Gbps despite the fact that they can all do 14Gbps no problem, in order to keep the marketing segmentation.

The end result is that the 1650S is probably the bang-for-buck card in this market segment, assuming you overclock the memory to get the ~10% free extra performance that nets you. Unless the mass-market RX 5500 is significantly faster than the OEM version W1zz reviewed, or comes in at $140 or less, it's going to have a hard fight against the 1650S.

Uhm yeah, though I'm not sure the problem is really 1660S, it would still have massive memory bandwidth difference. But vanilla 1660 which memory bandwidth 1650S now equals with 12Gbps gddr6s and there have no word about EOL of vanilla 1660. As for 5500xt, I really hope it's a full Navi14 chip and not some rumored nonsense OC 5500.

About gddr6 at least Samsung and Micron web sites have those still in production(Micron even have 13Gbps memory). But yeah those sites can be not so up-to-date. I would still believe that if either IHV contacts gddr6 manufacturer with >100 000 chip order, they will gladly make them. But every time I see out of spec memory used in review sample, I'm a bit skeptic if that is true on retail products as well: Card is spec'ed for 12Gbps memory, AIB does not have to use 14Gbps memory on retail card to achieve that. If I may, could you @W1zzard ask for that from EVGA?