diff --git a/GPU/GPUState.h b/GPU/GPUState.h
index e91be6d1ab..c6fb03c4d3 100644
--- a/GPU/GPUState.h
+++ b/GPU/GPUState.h
@@ -298,9 +298,7 @@ struct GPUgstate {
 	bool isTextureAlphaUsed() const { return (texfunc & 0x100) != 0; }
 	GETextureFormat getTextureFormat() const { return static_cast<GETextureFormat>(texformat & 0xF); }
 	bool isTextureFormatIndexed() const { return (texformat & 4) != 0; } // GE_TFMT_CLUT4 - GE_TFMT_CLUT32 are 0b1xx.
-	int getTextureEnvColR() const { return texenvcolor&0xFF; }
-	int getTextureEnvColG() const { return (texenvcolor>>8)&0xFF; }
-	int getTextureEnvColB() const { return (texenvcolor>>16)&0xFF; }
+	int getTextureEnvColRGB() const { return texenvcolor & 0x00FFFFFF; }
 	u32 getClutAddress() const { return (clutaddr & 0x00FFFFF0) | ((clutaddrupper << 8) & 0x0F000000); }
 	int getClutLoadBytes() const { return (loadclut & 0x3F) * 32; }
 	int getClutLoadBlocks() const { return (loadclut & 0x3F); }
diff --git a/GPU/Software/FuncId.cpp b/GPU/Software/FuncId.cpp
index 276e4f0820..3ee0d68655 100644
--- a/GPU/Software/FuncId.cpp
+++ b/GPU/Software/FuncId.cpp
@@ -427,6 +427,9 @@ void ComputeSamplerID(SamplerID *id_out) {
 	if (id.texFunc > GE_TEXFUNC_ADD)
 		id.texFunc = GE_TEXFUNC_ADD;
 
+	if (id.texFunc == GE_TEXFUNC_BLEND)
+		id.cached.texBlendColor = gstate.getTextureEnvColRGB();
+
 	*id_out = id;
 }
 
diff --git a/GPU/Software/FuncId.h b/GPU/Software/FuncId.h
index b505e88412..6c372a0c4f 100644
--- a/GPU/Software/FuncId.h
+++ b/GPU/Software/FuncId.h
@@ -171,6 +171,7 @@ struct SamplerID {
 			uint16_t w;
 			uint16_t h;
 		} sizes[8];
+		uint32_t texBlendColor;
 	} cached;
 
 	union {
diff --git a/GPU/Software/Sampler.cpp b/GPU/Software/Sampler.cpp
index bc997f7bf0..f346aa9759 100644
--- a/GPU/Software/Sampler.cpp
+++ b/GPU/Software/Sampler.cpp
@@ -509,7 +509,7 @@ Vec4IntResult SOFTRAST_CALL GetTextureFunctionOutput(Vec4IntArg prim_color_in, V
 	case GE_TEXFUNC_BLEND:
 	{
 		const Vec3<int> const255(255, 255, 255);
-		const Vec3<int> texenv(gstate.getTextureEnvColR(), gstate.getTextureEnvColG(), gstate.getTextureEnvColB());
+		const Vec3<int> texenv = Vec3<int>::FromRGB(samplerID.cached.texBlendColor);
 
 		// Unlike the others (and even alpha), this one simply always rounds up.
 		const Vec3<int> roundup = Vec3<int>::AssignToAll(255);
diff --git a/GPU/Software/SamplerX86.cpp b/GPU/Software/SamplerX86.cpp
index c7d6388eeb..cf650c0d67 100644
--- a/GPU/Software/SamplerX86.cpp
+++ b/GPU/Software/SamplerX86.cpp
@@ -1649,19 +1649,19 @@ bool SamplerJitCache::Jit_ApplyTextureFunc(const SamplerID &id) {
 			// We divide later.
 		}
 
-		X64Reg gstateReg = GetGState();
+		X64Reg idReg = GetSamplerID();
 		X64Reg texEnvReg = regCache_.Alloc(RegCache::VEC_TEMP1);
 		if (cpu_info.bSSE4_1) {
-			PMOVZXBW(texEnvReg, MDisp(gstateReg, offsetof(GPUgstate, texenvcolor)));
+			PMOVZXBW(texEnvReg, MDisp(idReg, offsetof(SamplerID, cached.texBlendColor)));
 		} else {
-			MOVD_xmm(texEnvReg, MDisp(gstateReg, offsetof(GPUgstate, texenvcolor)));
+			MOVD_xmm(texEnvReg, MDisp(idReg, offsetof(SamplerID, cached.texBlendColor)));
 			X64Reg zeroReg = GetZeroVec();
 			PUNPCKLBW(texEnvReg, R(zeroReg));
 			regCache_.Unlock(zeroReg, RegCache::VEC_ZERO);
 		}
 		PMULLW(resultReg, R(texEnvReg));
 		regCache_.Release(texEnvReg, RegCache::VEC_TEMP1);
-		regCache_.Unlock(gstateReg, RegCache::GEN_GSTATE);
+		UnlockSamplerID(idReg);
 
 		// Add in the prim color side and divide.
 		PADDW(resultReg, R(tempReg));